Slashdot Mirror
What AI Elements Could Improve the Web?
DavidpFitz asks: "I'm entering my final year of my Artificial Intellgience/Computer Science degree at Birmingham Uni. (UK). The trouble is that I can't decide what to do for my final project. I'd like to do something of practical value delivered over the web (things like an intelligent Slashdot filter spring to mind :-), but I always come up with reasons against everything I think of. Can anybody think of ways they would like a web site to react more intelligently that they currently do. Clever shopping carts? More targeted news? Both of these are rubbish, I think - so more interesting and complex ideas are welcome! The main thing, is that it has to have a strong AI element in it, not just appearing to be clever."
Interesting thought. So if we were to apply more AI to the web, what areas should we target? And I feel this is a valid question even if someone may be using these ideas for their school project. These are still just that: ideas. DavidpFitz will have to finish and implement his final project regardless of anything said in this forum, so why not take his line of reasoning, brainstorm a bit and have some fun with it?
Have an AI that keeps track of sports scores.
Well, that's what's selling all these damn web-enabled cell phones I keep hearing advertised.
So, for example if you give it a photo of a person, it gets you all photos in which that person appears
That's a hard AI problem. If you can solve it, I can assure you you'll be famous.
I know of a program that scans images (from the web or other places) and picks out porn. Don't laugh, it's real. The program, I don't remember its name, selects pictures containing nude bodies. It works by recognizing skin tones and, I think, not the absolute color values, but rather certain color gradations.
Face recognition (which is what you are talking about) is being actively worked on now. One of the applications is being able to automatically identify people observed by the ubiquious security videocameras. Would you like to live in an aquarium?
Kaa
Kaa
Kaa's Law: In any sufficiently large group of people most are idiots.
An agent for each program, and for each developer. A classification system that learns about the kinds of bugs and updates there are, and learns about the abilities of the developers, and bridges the gap between them. That'd be awesome.
It not only searches out developers to fix (or work on) code, it searches out programs that developers could make a big contribution to.
Also, possibly, an agent for each bug reporter... And for each "reviewer" - people who verify bug fixes / enhancements.
The system would learn, over time, for instance, that I like to complain about weird interfaces and constantly demand good help systems. It would then know to use me as a "reviewer" of fixes to those kinds of problems. "Dear VikingCoder, the AFE system has received a fix to a help system problem in emacs, would you be willing to review and grade the fix?"
Also, it learns that I'm a very apt OpenGL coder, and learns that if it asks me to write some code, other people are likely to grade my code highly.
HUH? Not bad, eh? This would be a cool, cool, cool system.
You'd probably be able to pawn it off on SourceForge, for instance...
Good luck with your plans!
Education is the silver bullet.
Two thingsb (neither of which I made clear in my original post):
.25 lb". Then the software has to find ALL models that match and price compare them for me.
1) I'd like to be able to say "Digital Camera, holding at least 36 1024x768 images, no LCD, weight less than
2) Reliably finding the price of a given item for even ONE site would be non-trivial (for some sites). Finding it for many sites would take AI (until we get smart and start using some standard XML for this kind of thing).
--
Have Exchange users? Want to run Linux? Can't afford OpenMail?
Linux MAPI Server!
http://www.openone.com/software/MailOne/
(Exchange Migration HOWTO coming soon)
It strikes me that perhaps the part of the web most in need of intelligence is search engines. It would be extremely useful (not to mention entertaining) to, for example, give web crawlers a more human-like 'personality', giving it the capability to dislike a site becuase of apparent lack of content, number of hits on a hit counter, flashy graphics, etc.
:)
But of course, AI should be dynamic... perhaps in such a situation you could program the crawler to use click-through data from the parent search engine (with the users' permissions of course) to try and formulate an idea of what people like in a website, thus helping the page rankings.
On the other side of things it would be an interesting AI challenge to try to work out what kind of thing people actually want to find when they search for certain terms, in terms of the content and function of the site.
Of course, this may just be a silly idea, but I suppose it might help
---------------- Take the red pill
Me: i'm thinking of going scuba diving in monterey this weekend
AliceBot: What is this "thinking"?
The defense rests your honor.
((lambda (x) (x x)) (lambda (x) (x x))) http://www.endpointcomputing.com a scientific approach to custom computing.
There have been many suggestions for pre-emptive content searches, but what about browser and content history? Something akin to an automagically accumulating everything.com.
For example, use semi-structured data parsing techniques (based on content, tags, etc...) to loosely organize the information on the web sites that you visit. Later on, you will be able to search, browse and correlate all this information that have "seen", but not retained.
To me, this would solve several very practical problems: Where did I see X the other day? Isn't X like some Y that I heard about?
In essence, an automated bookmarking utility based on content rather than location.
I suppose I'm not too threatening, presently, but wait till I start Nautilus
Any other features?
One thing I would like to see is something that lets you shop for the best price on more than one item at a time (with both items being from the same website) - if, for example, I wanted prices on a specific digital camera and the accessories to go with it, then I would be able to search for websites selling both, sorted in order of lowest combined price.
Of course, that isn't needed as often as a general price search for one specific item, but it could save a lot of time when it is needed.
If the poster's at Brum Uni, he's probably working under Professor John Barnden, who's doing plenty of research into the use of AI in responding to metaphors. So yeah, the whole Autonomy/Bayesian/remembrance thing: being able to deal with the fuzziness of human language and concepts, to get beyond the rather geekish world of the algorithm and of mathematical certainty.
Better yet, The Diamond Age, or A Young Lady's Illustrated Primer by Stephenson.
The problems of the web that can have AI-type solutions are generally semiotic in nature. We could have intelligent agents, able to pay our bills and do our shopping for us, if we had a system of symbols that mapped clearly and unambiguously to meanings, or, a non-symbol processing system able to decode the web the way a human does. Most likely, we'll get something midway between them.
:^P
The GOFAI idea that human cognition is a matter of disembodied symbol processing is dead, and good riddance. However, our computers remain most useful as symbol processing engines, not as the more complex kinds of massively parralel connection engines that most people think brains are. We can get computers to emulate that kind of brain functionality, but only on a very limited scale compared to the human mind. Human-equal parralel machines are not just around the corner, so we have to augment the connectionist systems we have with symbol processing facilities.
The kind of project that interests me is a system that uses XML formats to provide clear semantics on the web, and connectionist methods to make judgements about how to act in response to those symbols. A system, for instance, that can scan an XML resource for information about rock concerts or movie listings, and having learned in more connectionist ways the preferences of the user (both in terms of costs, scheduling and personal taste) can inform them of events they might like to see, perhaps even going so far as to make tentative reservations when it's very confident.
The same kind of system could be used to solve library research problems. An XML document structures data semantically enough that a connectionist system can make quick, fairly superficial judgements about the contents and how they relate to the research needs of its users. It can then do more indepth readings of the highest confidence documents, leading to better sources and new documents. In the end, it can provide the documents to users and assist them when there are gaps in their knowledge by pointing them to the document that fills the gap.
The killer ap for AI would be automatic translation. Since that's my field, I don't think it's somewhere you ought to go without a strong knowledge of linguistics, and of the past failures in the field. I have some ideas, but that's what my PhD is going to be about.
Let's say a combination of AI, robotics, and asteroid mining makes open-source hardware projects realistic to the point where its just as easy and cheap to create your own fusion-powered sattelite as it is to, say, come up with a Linux distro.
Then, someone comes along, and creates an open source web-controlled nuclear-powered death bot. Only a totally international, mirrored site set up with good encryption so users could be totally anonymous. In other words, a web site that was not subject to any government that could be used to kill someone.
Something like that is still a few years down the road, thankfully.
No, Thursday's out. How about never - is never good for you?
What about the COPPA? you can't keep any info about people younger then 13, which is definatly
the targte of most educational sites.
Cait Sith
The Alicebot is a real piece of junk. All the responses I got were seemingly unrelated to what I typed in.
Another thing to think about is doing something similar and tying in forward or backward chaining (don't know how viable it would be though, its been a while since I've looked at AI its all fuzzy logic at this point..)
I know a couple years back, there was also a lot of work being done with putting web interfaces on intelligent agents, NLP engines, knowledge bases, etc. Might be worth looking into that also...
-- A computer without COBOL and Fortran is like a piece of chocolate cake without ketchup and mustard
Technonerds: coming soon on z1nc.org.uk
http://www.doublezero.uklinux.net/
Doublezero: like Slashdot, only less useful.
Basic UI design should have no use for AI in the UI itself (ugh, that sounds ugly), and I don't know how useful it would be for the design process (it would have to be damn good). However, in customization processes, good AI could be very useful. Of course it would need components that would be very useful elsewhere as well, so they could be shared... What I mean is, flexible user interfaces have great advantages but are costly in time and other management resources. If you could have an AI that did the work of configuring the GUI for a non-technical user to his preferencies even when he or she is not very clear about what they are, that would go a long way toward intuitive interfaces. Basically it would do something analogous to the... (argh, ugly analogy)... interior designer. (there, I said it!) in creating an environment that is "home" by guessing, without a lot of data, what someone would imagine "home" should be like. This is probably the worst analogy ever, but what the heck, it's an Anonymous Coward so no one will read it.
How about the AI takes misspelled URL and compares them to a list that you have previously visited and pick the "closest one" or have allowable misspelling as short-cuts i.e. /. => slash.org
Something like a plug in.
TKrabec Pahh
"UI AI" is, IMHO, an ill-concieved idea that has had way too much work done on it in the past decade. The problem is very simple: if I spend a few minutes (or hours, or days,) learning a new interface, I want it to stay the same! I don't care if I never run "Backup", or if I visit Slashdot so much that it may as well be my home page... I don't want those settings changing unless I tell them to.
MS Office is a notorious example of this. In the newer versions, if you don't use a menu item frequently, it vanishes, so users aren't "confused" by too many options. I used to work tech support, and believe me, having your menus change for no reason is far more confusing than having "too many options"... and it is frustrating to new users and experienced users alike.
MSK
Or perhaps make a coding assistant which looks through the branches of the code and creates error messages which are helpful and distinctive. So when the message from a particular branch of the logic is still not understood well enough to solve the problem, you can at least search for that error message in the Knowledge Base and find the more detailed explanation...and the AI could create the skeleton KB also -- with KB alterations by humans being fed back into the AI...
Rest assured, there's no animated paper-clips or any other gratuitous (and useless) nonsense in my desktop UI vision!
:-)
Just think "adaptive behavior"
it almost sounds like your describing a genetic algorithm. which i think can work pretty well in this situation. let see you could have each member of the population moderate with its points and then have people meta-moderate determine the fitness of the moderation. you would probably want to do significant training of the system before it went into public use.
hopefully the algorithm would evolve to be complex enough to avoid being taken advantage of by first posters. it also increases the possibility that comments get moderated fairly. it also backwards commpatable with the current system and can be slowly phased in, by slowly increasing the percentage of algorithms to people.
Yeah, like something were it can tell if an MP3 is a Metallica song or not? :)
The Economics of Website Security
Isn't this what Google does, and does very well?
;)
;D
Google? What's a Google?
g0o0G13 5UX! L1NX2g0 rUl3z!
George Lee
I guess this could be used to really make sites like Slashdot personalized. This would go further then simply comment moderation. The site would learn what stories, comments etc that the user reads and thus is intrested in. New stories and comments that match those that the user spent time on earlier would receive a bonus and would be presented to the user prior to those that don't match.
This would go further then simply checking the section the news belons to, it would also check the contents of the story. And regarding posts it could check things like contents and who made the post.
Phobos - Greek word for fear or flight
Thats completely true. I have been also a part of a project that was started to implement AIinto the OS in different levels. Unfortunately the project was cancelled after a year because most people thought it was too big to be collaborated through the net. Anyway, If we could of AI for a system, not in the web, then yes ! UI's do need an intelligent way to track users' interactions.
Constants aren't, Variables wont
I've been fairly impressed with the Jump Start series of educational software for my son (now almost 6). He's had a couple of them. The latest is jump start phonics, which features a limited version of IBM's via voice speech recognition. My son plays it, 'cause the novelty of saying your answers appeals to him. Note to Mac users: Had to install this on on my pc rather than my son's mac because the Mac version doesn't have the speech.
It actually works pretty well and has done a good job of getting progressively harder as he improves. I can tell, too, that it does little things like keep drilling him on letter combinations/sounds he has trouble with. Also, when it's doing this and he gets several wrong in a row, it'll drop back to easier stuff (or ones that he knows) so he doesn't get too frustrated at getting like five in a row wrong.
---
DO NOT DISTURB THE SE
If we had AI moderators, it would take about a week before someone found a way of exploiting them so he could always post stuff they would moderate to +5. Wait, Signal 11 already did that and we don't even have AI moderation yet...
The bus came by and I got on
That's when it all began
There was cowboy Neal
At the wheel
Of a bus to never-ever land
I'd rather be lucky than good.
additionally this sort of thing would be useful for making playlists. just hit the 'fast driving music' or 'i'm feeling depressed' button on the mp3 car stereo and let the AI take over. ;)
-- This and all my posts are in the public domain. I am a lawyer. I am not your lawyer, and this is not legal advice.
I think that AI will have a limited use. Basically controlling content that people want to see,i.e filters. Also i suppose that marketing depts. could further customize ads and offers for surfers.
It's your kind of elitist "I got there first so anybody who came afterward is a loser" kind of attitude that drives people away from even trying to use Linux and insures that it will never be a mainstream operating system.
Actually customized news would be a nice thing to have if it really can work. Sorta like building a profile of the type of news you like to read, and giving you more stories like that, plus showing you follow up stories on stories you had already read.
Or even a program which can crawl all the news sites and do this for me. Sorta like how slashdot links to all sorts of news sites. So during the day the program can continue to scan the net and return to me links to stories it finds that match my profile.
It should be sufficient to improve an existing application (such as add smart graph layout to Dia or the network program TkIned/Scotty), as long as it is done well and the research methods are properly followed and explained.
Oh, you're using IE? In the upper right of Netscape is a "What's Related" button...
we should have artificially intelligent spam-bot blockers to get rid of all this dang spam. they can fight the spam-bots and challenge them to duels and stuff. it can be a new sport.
Actually, someone is working on this, and they plan to have a Linux client. It has an AI that figures out what kinds of songs you like, not based on genre but on how the AI figures out that the song sounds. You give it feedback on songs, and it also learns from what songs you skip, repeat, etc. It's currently in Beta, and the name of the thing hasn't yet been decided. I think the working name is Personal Media Agent.
It's a closed beta, but you can try to sign up. www.etantrum.com. Currently it only plays local MP3s, but their plan is to have it stream. It can play you songs you've never heard before that you will probably like, based off your profile that the AI builds over time.
Essentially, build me something like a customized version of Yahoo, based on my extended browsing history.
The only truly inteligent AI scheme would be one that was Client Side. If only for privacy issues. There would be many other benefits as well. If the client side AI engine monitored a user's activity over a period of time it would be better able to "predict" the next type of information being sought and then could query various search engines and compile links to those predicted sites. Client side AI would also be able to rate the effectiveness of the various search engines in relation to the information sought and retreved. Feedback to both the client AI engine and the search engines would guage the accuracy of results and then be able train the search engines to better serve the community. The final clincher in my book is that this formula would empower the fully distibuted computing model.
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
http://www.allmusic.com has something similar to this. It's not too bad in its current form.
I'd like to be able to speak into a cell phone "What's the weather like in New York?" and get a response from my favorite weather site. Or "Read me today's Slashdot headlines." The agent could also give you a call to report certain events. "I found a Linux server on dell.com for $699. Interested?" it might say. It could even call you on special occasions: "Call your Mom. It's her birthday."
Yes, this is very ambitious, but maybe some small part of this is feasible for your project?
-jimbo
"Hold me Bob!" "I would if I could man!" -Larry and Bob in VeggieTales
I agree.
HTML parsing != natural language processing
Unless you're implying that you should be able to enter a random URL, like http://www.joescomputers.com/, and have it discover the price for a 10GB Seagate hard drive, or some specific item, without anyone having told the software the format with which Joe's Computers displays its prices, or even which page they're on, then you're just talking about searching through HTML, tables, etc., which is most certainly not AI.
What if I entered "to be or not to be?" and I got back a link to the middle of a video containing that phrase. Not because the webmaster put in a marker for that phrase, but because the search engine analyzed the audio stream and found those words.
Or maybe I enter a picture of Harrison Ford and the search engine returns a list of videos/images which have him.
I hear a lot of talk about the web becoming a kind of interactive television. Unfortunatly, I don't think this will be as useful as the text because it's hard to find information in audio/video streams. Until you can do that, it will only be entertainment, not information.
-j
I think the world needs an automated agent that will scour the web for porn and retrieve pictures that meet certain criteria. For example, maybe today I am feeling the need for red-headed dominatrices wearing glasses. Perhaps tomorrow I will want to look at busty Asian women who are tied up. There are already methods to automatically detect pornographic content in pictures, just make it more advanced and embed it in an agent. The world will thank you!
Scuttlemonkey is a troll
What is the AI element of this? Moderate this down.
I believe there is enough information about people out on the web that you ought to be about to find people in lots of different ways. However, there isn't one big people search to do this. Most require certain fields that I dont have. Someone should create one big all encompasing, Master People Search, that draws data from all known people databases, maybe straight off the web, and maybe from its own database. This search shouldn't require you to have a last name. It should allow "like this" match phrases in the firstname, lastname, and email fields (example you know their email name but not hostname). It should allow you to look for phone numbers, email addresses, addresses, just a name, web pages, ICQ, AIM... I want to search on any info I know or might know.
And if you can come up with a way of having this work, let me know. :)
Gauntlet
Really, what you do with your AI project depends on the model of AI that you are using to model intelligence. In my little AI project here, I am using a system that is based fundamentally upon language (among other things)...if you are doing something similar, it would be useful to have an AI program that combs multiple sites, following links as appropriate, and forms a synthesized synposis of news that it has learned that you find interesting. I know this is been tried before, but most "expert" programs lack the learning capability to pick up on the major subjects that you follow, and then can't make the leap to finding related (but not random) stories, being that they lack even minimal comprehension of subject matter.
The last time AI was la mode, which was about 15-20 years ago, one of the computational walls AI hit was in computer vision. So, 10 Moore's Law cycles later, ...
If a Web page knew you had a camera pointed at yourself, could it watch? As in, see what you're looking at, where you point (forget those mouses!), what you spend time reading?
So instead of "Big Brother is watching you!" we'd have Tim Berners-Lee is watching you!"
... an idea, the fugitive fermentation of an individual brain ... -- T. Jefferson
Yes, I believe that AI can be used to enhance natural intelligence. I recommend a book called 'The Diamond Age' by Neil Stephenson. In it, a young girl is essencially raised by an AI-based, interective, self-modifying, book. It teaches her lessons in many topics, test her understanding, evaluates her progress, builds the next lesson based on her performance, provides an outlet for creativity, and so on. You might find it a thought-provoking model. If you could use that great education you are getting to build tools that would allow us to help make our children smarter or more emotionally healthy or just better people, you will have performed an amazing service. And I'll bet it's more fun than creating the next programmable shopping cart. There is a huge market for this already. 20 years from now, something akin to this will exist. Maybe you will be the person to get rick bringing this to market. Good luck and aloha, =brian
Well... you may want some kind of UI for the AI itself. That's more or less one of the functions of the paperclip. 'Course the value of having little characters is debatable. Apple did some research on that sort of thing a while back, when they were working on Bow Tie.
-- This and all my posts are in the public domain. I am a lawyer. I am not your lawyer, and this is not legal advice.
I'm not sure if this could actually be full-blown AI, but what about some sort of overarching adaptive browsing environment? Something that learns where you go and prefetches/digests information. It might even "learn" that since you like Slashdot, it should parse the Slashdot headlines and provide a ticker for you. How it would know that I have no idea. AI magic. Or perhaps it might learn what type of sites you do not like cookies from...perhaps it does a lookup on doubleclick, fuzzily figures out what doubleclick is about, and subsequently by default denies cookies from other doubleclick-like sites. Same for ads.
Still that's admittedly AI-weak. How about the adaptive (dare I say even neural-net-ish), FreeNet project? Could you do some work for them in perhaps detecting "cancerous" nodes?
It's 10 PM. Do you know if you're un-American?
along those same lines I would an agent that would rember what links I have visited and type of content at that link. I would like to be able to ask this agent quesions like: "What page had the pizza recipe on it? Where was that howto on building a low power amp?"
This may seem as a good idea as you first look at it, but there are some small chances that it could be abused.
Everything would be fine while people simply play with the bots, have them run around in circles and crush small stones and things like that. But I'm sure that someone would eventually come to think about using the bots for evil deeds (someone always does, even earplugs can be used for evil). It is not clearly obvious what evil deeds the nuclear-powered death bots could do, but a twisted mind can think of anything...
Phobos - Greek word for fear or flight
To a limited degree this is indeed possible, and there has been past work on this. Most of it has been based on the analogy of intrusion detection being similar to our immune systems (detecting unfamiliar proteins/traffic and use patterns). Wish I had the papers handy, but I don't.
It all depends on what you define as intelligence. More neural interconnects? The ability to clearly connect ideas together? What if you can't communicate those ideas? What if you weren't even aware that you had those ideas? Is that intelligence or would that make you an idiot-savant?
I suppose this is more of a question for philosophers than computer scientists (but if you're in the AI field you're surely dosed with lots of metaphysics). However when we're talking about search engines we have to realize that their inaccuracy is due to the fact that that there's no way of them to understand the intent of the user, because all that is communicated though a very complex context we call "human life." It really takes the sum of all your human experiences to understand the sentences that you read from a page (and not some freaky formula as early works of Bertrand Russell / Wittgenstein would put it) and really
- understand
.Surely we can create a massively parallel, multiprocessor neural network computer ("Daaayyyy-sies, Daaaay-sies...") with enough power to surpass the computational abilities of tens of human brains, but the thing would still have a hard time understanding people. But make two of them, let them talk to each other, and suddenly the AI's develop identity politics and launch the nukes.
[pink beam of light]
An AI that can recognize the different meanings of things _In_Context_. That way I don't get "HOT PORN SEX!!" when I'm searching for "Heat sinks".
Since I'm soon going to be installing hardwood floors in my computer cave, I can relate. Seems any search with "hard" or "wood" in it turns up mostly Viagra vendors and porn.
But I doubt that any AI widget could fix that. Not without angering multitudes of limp guys seeking porn. "What's this? I didn't ask for pictures of red oak planks..."
I can only imagine what sort of lewdness a "tongue in groove" search would have gathered.
I've researched stuff like this in javascript. Basically, my system associates a string of digits "sA" representing categories on a 0-9 scale. There is a variable, "C" incremented every time a song is accessed. There is another variable, "uA" which is a string identicle in format to a given songs string "sA". When a user listens to a song, sA and uA's individual digits are sent to parseInt() and stuffed in arrays. The variables in uA's arrays are multiplied by C.
Then uAArray[i] = uAArray[i] + sAArray[i];
in a for loop. C is incremented, uAArray[i] is divided by C, and uAArray is sent back out as a string and written as a cookie. Strings are used rather than sticking to arrays for persistence reasons. uA is then used to search a database of songs whose sA matches uA within tolerences, and the result is a means of pesenting the user with an evolving list of songs they may like.
So what we have here is a feedback loop that does what you're describing, but it isn't an AI.
"Let him go, Ralph. He knows what he's doing." --Otto Mann (simpsons)
Uh no:
artificial = manufactured
I think you meant virtual...
Regarding the rest: Please, never accept a position at Microsoft R&D. Distributed Windows Scripting Host anyone? (yeah it's a cheap shot, sorry)
I strongly believe that trying to be clever is detrimental to your health. -- Linus Torvalds
why not allow us to be REAL lazy. it could not only decide which links we'd click, but also read the pages for us, follow any links on that page, etc. and dont forget it would receck any pages we constantly reload, email, comics, slashdot and even post on slashdot witty comments and first post claims. We could just sit there and have it surf for us, hell, it could even read the pages out loud. no, then we'd have to listen to it... it could just go through the whole web and sum all the information up for us.... would it say 42?
- "yes but can you hit someone over the head with a rolled up internet?" -Foxtrot
A good deal of thought and research is going into how people translate ideas between languages. This would make one heck of an AI project and could prove very valuable to the web. As far as books on the subject go, I'm thinking of Le Ton Beau de Marot by Douglas Hofstadter.
I work for a company that builds products which use AI technologies to assist web and e-mail customers in getting around a company's web site. Rather than use standard pattern-matching and vocabulary databases, we use natural language processing components to determine the intent (or intents) of a message, then use the intent(s) to select a case in a case base.
It looks to me like they're mostly considering the colours used in the images rather than the shapes and you get very fuzzy matches. Still, it is something in the direction of what you were thinking.
--
Fuck the system? Nah, you might catch something.
Aaaaarrrrgh!
That was my first thought - based on what M$ have given us in the name of AI user-interfaces (the paperclip and friends).
But it occurs to me that it needn't be that nasty, if you put a bit more work into it. For example, I get really annoyed when Word changes 'teh ' to 'the ', and I just notice after going left,backspace,left,'h',end. A more intelligent 'agent' would have remembered why it did what it did, noticed what I was trying to do, and sorted everything out.
So, bringing back the Internet theme, how about this for a project: a browser which can learn about sites that the user visits regularly, and provide various access shortcuts. For instance, if I regularly refresh the Slashdot index until a new article appears, the browser could notice this behaviour and offer to take care of it and notify me when the page is updated.
Or it could notice that you often click on a link labelled 'download software', and this takes you to a page with a high legalese quotient, which you always skip straight to the bottom of and click the 'I accept' button... the Intelligent Browser could bypass this page for you (this sort of thing could really piss off any DMCA lawyers).
-- Andrem
There has been a major scientific break-in
Yea. We wish.
A book that will teach you to read if you don't know how...
-- The act of censorship is always worse than whatever is being censored. Always.
or hack tevo to the same ends
"Let him go, Ralph. He knows what he's doing." --Otto Mann (simpsons)
I think it would be neat, not because we really need a computer program to moderate for us, but just as a challenging problem to write an AI for.
Probably genetic algorithms would be a good way of going about it. The AI could train by comparing itself to human moderators, there's certainly enough data out there for it. The big problem, the one that makes it near impossible to do though, is having sensible criteria for how the program will evaluate a block of text as being 'insightful' or 'funny.' If you can do this, then you have made a huge breakthrough in the field of AI.
In fact, a computer that has true artificial intelligence will 'do better than it's programmed,' as it will be able to learn as it goes, provided it gets feedback on the success or failure of its options. Also, who's to say that we wouldn't have multiple AIs each with a different set of values, combining to do moderation just like us humans?
When you look at a fresh page of comments, which do you think more often: "These comments were definitely moderated by humans." or "What were these moderators smoking?"
--
share and enjoy
With this in mind, I suggest reading Neal Stephenson's The Diamond Age", since the "primer" in the novel represents probably the ultimate teaching tool - one that takes Nell from first words to computer programming and martial arts without needing adult intervention!
(ceci n'est pas un
Me: Well, I just had lunch.
Alice:
> Well
Well what?
> I just had lunch
Only just? Do you still have lunch ?
Me: Now that response wasn't very intelligent
Alice:
Why now? "Do you still have lunch" response was not very intelligent ? I get it.
-------
No Alice, I don't think you do...
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Consider the ideas of collaborative reviewing and scoring, along with third-party annotations like like what Third Voice does. Imagine a "moderated" WWW...
---
As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
how about an intelligent caching server?
this assumes the user is behind a squid cache.
a client would scan the user's browsing history (from ~/.netscape and ~/.mozilla directory) and then query a server to see what else a user would probably download next. then download the predicted web pages (via the squid proxy).
the server would need to keep track of browsing patters in some fashion (that would be the ai part).
alternatively you could use the squid cache itself to do the predictions, but then you're dealing with multiple people's browsing patterns: experimentation might come into play here to see how well that works.
anyway it would be an interesting project in terms of speeding up web access by utilising "downtime" in net connections.
US Citizen living abroad? Register to vote!
I wont reply to flames with flames.. its a waste of my time and for everyone else's time but I will say this, it looks like its that sirwhatever the one who doesn't know anything about anything..
Constants aren't, Variables wont
ASG is a relatively uncharted field of Artificial Intelligence, with possibly unprecedented value with respect to Rapid Application Development and Software Verification procedures. It has industrial as well as theoretical value.
Perhaps attending the AAAI conference in July/August in Austin Texas would also be of help in providing some insight to what field would be most suited to your desires.
Cheers!
Brian
OK just a gripe about many of the search engines I use. Maybe I'm not using proper filter or search techniques, but what about an AI controlled web spider that searches out links, and then is able to filter the findings by classification--ie if you want to filter out web board messages for instance.......
What? Me worry? NEVER.....
Imagine a site like mysimon that compared millions of products from thousands of stores.
A laudable goal, but in order for this to work on that kind of scale you need to establish a standard, and then you need the retailers to participate. This is a classic bootstrapping problem, because retailers hate price transparency--no seller is going to sign up for a system where it has to do extra work to compete against potentially millions of other retailers based on price. However, if you can make the system large enough that consumers start to go there first, all of a sudden the retailer will need to participate, since it will start to lose customers to the service otherwise.
Then, instead of making you do all the work of signing up and purchasing from each store, the program could use a single shopping cart, and negotiate the sale from each of the stores that had the best prices, using the same user information
The single shopping cart has the same problem: in a world of increasing price transparency, one of the few "sticky" features an e-tailer has left is that you've already gone through the tedious job of filling out a user profile on their site, and not someone else's. Are you really going to go take 15 minutes to fill out a profile at Egghead.com to shave a few bucks off the price of a digital camera if you already have a profile at PC Zone? Probably not. Plus, there are perhaps some privacy issues.
Of course, it would be a truly wonderful thing to have if you could find a way to make it happen. I just wonder if it's realistic...
Maybe there is a lot more GUI work than AI here, but it would be nice tool to have.
Pre-programmed? I think anything with any shred of intelligence could come up with these responses on its own!
Summary (Where's Waldo?):
Develop a protocol to intelligently look out the cameras attached to, or the "windows" of, the web.
Preface:
My guess is that the planet, as well as the solar system, will soon be 'covered' with a large set of randomly distributed CCD cameras attached to the web,-- managed by many different people/orgs.
At the same time:
1.) It would be nice to be able to take a virtual walk that preserves physical continuity, by visiting camera after camera: maybe even "walking up" into outer space, or "down" into the oceans when the available window-space of the web makes that possible.
2.) It would also be nice to be able to visually locate, as well as visually follow someone,--let's assume that they want to be located and followed and that they are aiding us in our attempt at finding them and 'sharing their day' by carrying a portable GPS that is also attached to the web. (Super ICQ.)
--in the future, image recognition would become an order of the day--
More Specifically:
Perhaps a fun AI project would be to a.) develop a web front-end and internet protocol for the cameras in WWW-space (that wanted to participate in the window-space of the web) that would allow these cameras to be query-able as to location and pointing, as well as b.) develop a traversable network of these nodes which could be correlated with (real-time) positional and GPS data (or other types of data,--but we're just starting out so let's make the problem well-defined).
quick aside:
--Could you build camera-anonymity into the project?
In any case: there is
1.) A theory component:
a.) Developing the protocol. (which would need to be complete enough to take into account our outer-space, under-water, etc., nodes.)
b.) Developing the traversal algorithms (which can be your AI component).
2.) A technology component:
a.) You can get five, or 10, or a hundred cameras, and one of those cute SONY GPS devices and implement a 'toy' system as proof of concept.
3.) A web component:
a.) As well as being inherently web-centric, there are multiple web front-ends that could be put on a system like this.
4.) ...The ever-present commercial component:
a.) Realistically, building a new, ordered, visual means of traversing the web will have plenty of commercial spin-offs.
Best wishes,
--Forgotten Password.
(1) Learning Symbolic Email filtering rules, implemented in a Eudora plugin (Mac, Windows, Unix). Man, I can't maintain my email rules anymore. Non trivial problem. There has been a bunch of work but none's made it to prime time, and most is subsymbolic (so you can't edit them). MAXIM was one system developed at MIT. Not good enough though.
(2) Automated Web Scrapper Tool - V.Hard. Build a tool that with minimal user input (preferably unskilled), will produce a program that can repetitively mine the site for data. (hint, DOM and REGEXP). Ariadne is one such project, but there's no source available. (USC-ISI look for Craig Knoblock, Steve Minton).
Winton
Develop a bot that accepts a search query, or from a selected list of thousands (millions?) of products already listed and categorized (another AI project all in it's own!).
From there, this product will be searched for at shopping sites, online auction houses, or anywhere this product is found for sale.
This info must be kept up to date (updated once a day?), and not an affiliate service (not paid for listing), and is a 3rd party service to bargan shoppers.
There are sites that exist, but most SUCK!
computershopper.com - prime example.
Using computershopper.com as an example though, a wizard which could help a shopper look for what they want based on their uses would be another very useful feature.
Also this could be expanded to items other than computer hardware/software. clothes, cars, etc.
That would be one hot site!
Guess I replied to the wrong one :) Like I said in a previous post, there's already one out there that does this, only it's smarter than that even. You never have to manually rank things. You give feedback as you listen, or you can ignore than and just let it learn when you skip songs. Still in beta, www.etantrum.com.
I LOVE that netflix thing, and I was going to mention that as well. I actually saw a thing about that on TV before I found it on accident on the web. They can also apply that to CDs. Their plan is to make "soulmate" groups. Basically, the theory is that someone who shares your tastes for a lot of things will share your tastes for most things. They call these "soulmates" and keep them in a big database. They want to divide these out into groups of 4-5 of your soulmates, then you can talk about CDs, movies, etc. and recommend them to each other. I think it's a great concept. An interesting AI that uses existing patterns to make recommendations.
What does it mean: "Document contains no data?"
--
share and enjoy
http://www.cs.nott.ac.uk/Research/ltr/index.html
If you can make our automatic marking systems more efficient ( friendly? ) , Nottingham students for years to come will worship the ground you walk on.
Yeah! And filter out virus hoaxes too. You can always check the archives and see how good it is at recognizing them.
Scriv
1) An AI that "looks" at a page, graphics and all, and verbally describes to the user what the page is saying. This could be very useful for visually impaired folks. Back when the commonly used HTML tags were semantic tags rather than formatting tags, this would have been a trivially easy assignment. But now, one almost needs to parse an image in order to view a typical Web page.
2) Along those same lines, a more "intelligent" filter for converting HTML to WML, so that the myriad Web pages currently formatted for 15" graphical displays could be translated better onto small text-based displays...or even Braille-based devices.
3) An enhancment to Google's "I feel lucky" mode, but applied to Jeeves, so that a Jeeves-type natural language query would generate a succint, single-sentence reply. One problem with Jeeves right now is that even when it understands your question properly, its answer is an entire Web page! not a simple sentence. An AI that parses the Web page to generate a single sentence that directly answers the original question -- that could be an interesting challenge.
I hope that after I die the one word people use to describe me is "resurrected."
It seems that the creation of a system like the one you described would entail not only clever AI, but also clever teachers. In fact, the described systems draws not only on AI and sound pedagogy, but also on psychology... truly a grand undertaking.
Such a tutoring system would have obvious positive aspects. Yet, in the world of lazy and short-sighted humans, there exists a clear danger. It seem obvious that such a system, no matter how well implemented, would pale in comparison to a good fleshy-type teacher. But once we have "The Young Lady's Illustrated Primer," how many real teachers would remain? They aren't portable, they are expensive, and the pay just isn't there.
Ideally, this software would be just what the OP described, tutoring software -- a supplement to learning. Yet, what's ideal rarely happens.
Glazik
That's a joke. Get it? There are a bunch of witty responses like that built in. It's really not that dumb..although it is still pretty dumb...
Yeah what we need. Every /. reads intelligent browser constantly reloading the page because that's what people mostly do.
DDoS! Bigtime!
It's called new wave but it's just the same.
I happen to work at a startup, Links2Go, Inc., which approaches the "better search engine" problem from a different direction than most engines. Instead of farming huge numbers of web pages and doing greps on them for relevant text, our server sorts these pages by topic automatically and rates a page's relevance, not by the number of keywords on the page, but the number of times the page is referenced from other pages.
Users can then search on our servers by topic *or* by the URL of a page. What the user gets back is either a list of the most relevant pages to a specific topic *or* to a specific URL.
George Lee
That's not AI, you idiot; that's called Mad-Libs.
You could make an intelligent Slashdot-posting agent. It would require only a limited vocabulary, like "Microsoft sucks", "Natalie Portman", "hot grits", etc.
Possibly a better (and more lucrative) idea would be to write an intelligent computer-industry rumor generator. The great thing about this is that it would require virtually no input to generate an infinite amount of output.
That's "Mr. Soulless Automaton" to you, Bub.
Oh, hang on a minute, I already have my degree and I did all my own f***ing work for it (and bloody hard work it was too).
Clear off, slacker.
Rich
Damn Right!
"Who says a funk band can't play rock?"
Such a tutoring system would have obvious positive aspects. Yet, in the world of lazy and short-sighted humans, there exists a clear danger. It seem obvious that such a system, no matter how well implemented, would pale in comparison to a good fleshy-type teacher. But once we have "The Young Lady's Illustrated Primer," how many real teachers would remain? They aren't portable, they are expensive, and the pay just isn't there. And for that matter, why would anyone ever go and pay to see a movie again when they have the joy of television in their own house? Just because a new technological development comes along does not mean that prior technologies are thus outdated and discarded. Human teachers will always have value.
Devolver's Homepage... more fun than a box of crackerjacks.
I was actually thinking in the other direction.
How about a AI engine that can determine whether a picture is porn related/unappropriate material. It could then be expanded to mpg/avi/etc. Then their wouldn't be an issue of filtering educational text. The erotic stories would have to be filtered another way. Visual interpretation would be an excellent project.
I wouldn't call it canned. I would give some examples but START is kind of slashdotted right now. But START has given me information about distances between two completely arbitrary cities on earth, simple math, and even the answer to the universe. If you are asking for actual *thought* on the part of START, then I am afraid not. I don't think you can ask it for the highest prime number or for it to perform logical deductions, although it would be cool.
Ahh, there's a good AI project (though not web related). Make an AI program that can find out how to figure out the highest prime number! Easier said than done right? Heh...
I don't know much about search engines or AI but what if when you search for something it looks up the definitions of the words and then uses the context of the meaning of the words. It could use everything.com for phrases and if it finds that phrase in everything.com it could then use the definition in to weight search results.
He who knows not and knows he knows not is a wise man. He who knows not and knows not he knows not is a fool.
... that not everyone gets to ask Slashdot what to do for their final year AI project...
;)
Jammy. Very jammy. Hope it works out.
--
Google has been doing this from the start, and seems to do it pretty well. All without annoying banner ads, portal front-ends, lag, etc. that make the other popular search engines a real drag to use.
Screw Micro$oft.
It might be a bit ambitious but, what about an intelligent language translator? You could choose your native language and all sites that you visit are automatically converted to your language.
How would you understand my words if you don't understand my silence...
The problem with interface architecture is that it's geared towards making the most effective interface for the most people. So what we end up with is the most common mode for all users rather than the most effective mode for individual ones.
It's interesting that you're starting with the web, which uses text and visual communication modes. You can widen your scope to have your AI monitor the data and datatypes that flow to and from the user's IO media. Given an evolving model of the data and datatypes that are most efficient in facilitating comunication with the user, the AI can derive the communication modes that are most effective for the user (text, audio, lecture, video, pictures, graphs, comics, scent, braille...etc), and then dynamically model the internetworked data experience to best leverage those modes. This can be manifested as a web app, but it can also be a window manager or a kernel module that views the user as an i/o stream and optimizes the stream to maximize throughput in the available bandwidth.
The short version is a human-machine interface that self optimizes to most efficiently communicate with a particular user.
"Let him go, Ralph. He knows what he's doing." --Otto Mann (simpsons)
This is an idea that was talked about in Discover Magazine long before the internet revolution. In this article, they called them "knowbots" and at the time it was a revolutionary idea. Now we have spiders and portal sites that have almost arrived, but in a push fashion and not pull. I've also seen this idea in cyberpunk and sci-fi.
The idea was this: You tell the agent information you're interested in, and it's always looking for it. The Knowbots of the article would get you the stock quotes you wanted and all news articles on emerging technology, for instance. This has already arrived, with web portals. Again, push, not pull and I would expand on that, make them smarter.
What if you had an AI on your computer. It's yours. It is learning about you all the time. You don't just say, "I want news from this catagory, and the latest movie releases and the following stock quotes". You would say that at first, but then it learns over time, based off of what you use and don't use, what you are actually interested in. It is always searching the entire web, news syndicates, etc. It uses a combination of AI technolgies, and things like Netflix, to determine all of this. It wouldn't tell you ALL new movie releases, it would tell you the ones it knows you like. It would tell you the day your favorite band releases a new song. It would pull up an obscure news article you never would have found on your own about a recent fire in your old home town. The longer you use it, the more tailored to you it becomes. It learns not only through your direct activity with it, but also all other activity on that computer: web browsing, software you install, CDs and DVDs you play.
This is very vague, but I think helping the internet to intelligently detect viral infections and stop their spread would be a very useful project. Perhaps some way of monitoring when a message/attachment is sent to a site, it gets sent out again in multiple copies. As a side benefit, it might kill chain letters.
Also check out The Great AIP (Artificial Intelligence Project)
Note that the plan files are old and things have really changed.
Just because you think of problems with the ideas you've already come up with doesn't mean they're not worth implementing (depending on the problems of course). Regardless of what you do, you're going to have to cast a critical eye upon it when you're done.
My area of interest is access to and organization of information, so things that smarten up indexes and semi-automate classification of resources would be what I'd go for. Maybe something to recommend Dublin Core tagging data, but then, that's probably too small.
what could help us more than a bot that analyzes the needs of slashdot users and moderates messages you would like up for you.
of course this would stop any new ideas getting to your closed circle of ppl.
wonder if that happens now?
Parents Against Kuro5hin
One of the big headaches for the new distributed filesystems like Gnutella is keeping the level of traffic reasonable so the network is not swamped with bad packets, pings etc, and at the same time keeping the distributed connectedness at the best level, auto-optimising the network would be a really cool project and be of tremendous value to everyone whos working on projects like Gnutella and FreeNet.
Any sufficiently advanced man is indistinguishable from God
Some friends of mine and I have been talking about intelligent moderation schemes, in terms of individual interest. Slashdot's moderation works pretty well, as far as it goes. But out on the news servers, you've got a truckload of information in a tankerload of waste, and that's in the topics you're interested in.
Having a good personal assessment scheme without asking me to categorize (a little bot that watches me read, and guesses how much interest I have based on time spent reading, which titles start me reading, and a real simple subjective scale at the end, a "1 to 10, how'd you like this post?"). The little AI tries to make connections -- are my preferences author based? Topic/keyword? Length? Flame content? And develops an evolving rating scheme. If it was successful, pretty soon I'd trust that I could ignore everything rated 2 and below, and spend less time surfing and more time getting information.
Of course, the idea applies to weblogs, too. Generally speaking, any filter that got rid of stuff that I consider crap, without me having to figure out WHY it's crap, would be enormously handy.
http://www.artificial-life.com
Most people here want better search engines. Forget that. That's a crowded arena.
I'd like smart web prefetching and advertisement filtering. Basically, I'd like my browser to figure out which links I'm most likely to follow on a page and start prefetching those links. I'd also like it to block content which I'm not interested (but still leave a tag so that I can 'correct' it if it's overzealous).
Essentially, a combination Squid + Junkbuster, only proactive.
--Joe--
Program Intellivision!
As a general improvement I would like to see embedded mp3s and/or streaming mp3s. Is it possible to embed mp3s yet? As for AI? i would enjoy seeing an improved search engine that would take into consideration the bowtie effect and give a more accurate search result.
My biggest problem is finding things that I
know exist somewhere on a site. If you
could improve searches, you would get a big
win.
Another thing would be a simple change to
browsers so the do not open a new page as
soon as you click on a link. Just change
the color of the link to indicate the
download is in progress. In fact, the link
could be made into a small progress bar
showing how much of the download was
complete.
That way, a viewer could select several
pages he wants to see, and go to the first
one that is ready.
This would save a lot of time on long
SlashDot pages!
The web is growing every day with more and more content that is dynamically generated. What we need is something that will at least give search engines a grasp of what's buried under all of that.
Sure, many sites with dynamic content provide an engine that will allow THEIR dynamic content to be searched, but that doesn't help if you're using a major search engine to find ALL the sites with relevant information, not just relevant information on ONE site. We need a way for the engines that search dynamic content to report back to the big search engines what they have in their databases.
And then we can deal with all the security and privacy issues that will probably come with it.
The entire point of the Internet is to relay information. Information must, by definition, be meaningful to its recipient.
I'm sure you all remember the study done a year or so ago reporting that even the best search engines hit only 16% or so of the sites that are actually on the web. Clearly, there is a need for a good AI agent to look for information relating to a query and present that information to its client. Ideally, the client would be able to ask a question like, "Who was the fourth Pharaoh of the Nineteenth Dynasty?" and receive a weighted list of answers (e.g. 85% of sites consulted say it was Seti I).
The data is there. What we need is the means to collect it and turn it into information.
www.alarmist.org
I was discussing something like this the other day with my friend, on the subject of the ILOVEYOU virus that recently went around.
How about some AI to detect patterns in emails being sent in and out. Say, an email comes in with the subject "ILOVEYOU" with an attachment of a certain size, and a specific body. Then, suddenly, there are 25 more emails that come through with the same exact content, or very similar. Then there's 100 more of them. It would be great if the mail server could detect this type of pattern or irregular activity, and trigger some action, such as alerting an admin, or even going so far as checking the emails against an anti-virus authority, and blocking them.
Of course, AI would come in extra handy here if the virus was implenting some morphism, and not every virus email being sent had the same exact contents.
This is probably not the most creative use of AI, but I think it would be a great feature anyway.
Well, we searched for intelligence in the sky (SETI) we searched for intelligence on this planet and we concluded that there is none of that anywhere in the entire Universe. So, we decided to introduce it into the nature and call it AI.
Book of Joshua the new
History of Silicon Valley, year 2159
The Dooms Day
You can't handle the truth.
How about this, forget the web, and apply your AI expertise towards building a decent desktop UI!
In the Windows world (forget about stability, I'm talking USABILITY here), we have windows that pop themselves to the foreground whenever the hell they want, and where you click START to SHUT DOWN or LOG OUT.
And in the Linux world, we have a new window that gets created but doesn't get FOCUS, and we have the very UNoriginal Windows95 look and feel, but without scroll-wheel mouse support. And the rarely implemented-properly cut/copy/paste features.
All in all, I have to say, a decent desktop UI with some AI (or even just 'I') features would be just dandy. So forget the web stuff for now, and give us a decent UI !
Create a daemon that sits on a server (running with root privilages, important servers preferrably which get a lot of traffic such as an ISP), and monitor traffic. It could monitor raw network traffic, or it could scan e-mails of all the users and text files in their home directories. If you really want to get into the AI, disect audio and visual data.
Use AI to attempt to find trade secrets, bomb threats, and itricacies of international espionage. Figure out the source of it, and then report it to a central server (or your local government) which probably should do some more sorting.
It'd be quite useful, and I'm sure if we all install the server/clients we'd harness enough computing power to learn everything!
yeah..
I'm sure you could make alot of money if you created some AI that could reconstitute the real e-mail address from slashdotter's anti-spam addresses, taking contextual clues from their .sig files if necessary.
Or perhaps you could design some stuff to help out the fellas at Echelon or Cybersitter, since they've had 10 years or so to figure out what pr0n reall means, and haven't been able to do it yet.
Remember, spam is exactly the kind of freedom of speech that needs protecting.
- My password is slashdot
"Alice" is not even an attempt at intelligence. It simply analyses speech patterns without any regard to content or context or previous sequence of conversation and regurgitates replies that were hard-coded in beforehand and designed entirely and simply to "sound realistic" and not at all to mean anything.
if you are looking for a web-based Artificial Intelligence which actually solves problems and attempts to in some way synthesize the information given to it based on context, i suggest you look at
http://www.forum2000.org/
I assure you, you will be impressed.
Irritable, left-wing and possibly humorous bumper stickers and t-shirts
One of my favorite professors here at UC San Diego does a ton of interesting work regarding AI in search engines. Some are document based (like find things related to this document) and others have to do with merging to delated document heierarchies. And still some others use GAs to do protien recognition and docking. Very cool stuff. His web page is here , but the page looks a little out of date. And if you want any more info, send me or him an email.
Best of luck,
Ben
bshapiro@ucsd.edu
I don't think AI is going to be able to handle that very well. If I you tell me you want to swallow my carrot, I might be quite aroused and HAL, the webified-porn-detector might not even notice.
This would be useful for blind web surfing so that the user, for example, would listen to the content first before getting to the sidebar. This could also make web sites more PDA friendly.
You could split pages into content and links, catagorize web sites into those and present accordingly, extrapolate summaries of pages for faster surfing, simplify tables, strip unnessessary design to get just the meat of the page.
I know there are static websites that do this already (pricewarehouse.com comes to mind), but how about an agent that searches for the best price on a given item. Other options include:
-automatic sale ending date detection
-automatic score-lowering for companies the user doesn't like (i.e. give me the lowest price that isn't at WalMart)
-automatic score-lowering for companies with "bad practices" (i.e. give me the lowest price that isn't from a company with slave labor)
-couple those last two with automatic parent company tree-walking (lowest price that isn't from company doing bad stuff AND isn't owned by a company doing bad stuff)
-full generality: I want to price toilet paper AND houses
Any other features?
--
Have Exchange users? Want to run Linux? Can't afford OpenMail?
Linux MAPI Server!
http://www.openone.com/software/MailOne/
(Exchange Migration HOWTO coming soon)
An AI spam filter that could trustfully remove the garbage, and for stuff it's not sure about - send back a reply with a question that only a person could answer effectively.
I think a search engine that could use its inputs to figure out what pages are "crap" and what's "good" would have such a profound impact on the web.. I can't imagine a bigger advance.
What about an AI that takes a given pool of data (e.g.current news headlines, a collection of images, etc) and creates an electronic image *inspired* by these. I don't mean a random mishmash of whatever it was handed, but rather, identifies some underlying theme in the data and interprets it in an artistic fashion... ok, maybe it's weird or impossible, but I thought it would be cool... probably fairly challenging too.
---
I hope you're not pretending to be evil while secretly being good. That would be dishonest.
I have always thought about what it would take to build a program like one of those old doctor programs, but one that would post to a newgroup or IRC channel. I know this has been done, but not very convincingly. This program would remember who it talked to and about what. It could pick on key words being talked about and got out on the net to "learn" about it before posting. If something good enough could be developed, it would be fun to watch a few of these have a conversation and see how far they could take it.
"Those who make peaceful revolution impossible, make violent revolution inevitable" - JFK
how about something to detect probably malicious javascript?
I have several ideas. First, weblogs really annoy me in that they only log the numerical address and not the domain name of a site. Second, it would be cool if you could look up an area by a person's IP address, figure out where it is located, and specify news for that region. That would rock.
What ever happend to Pete Townshend's "give me your bio and we'll make a song about you" AI routine?
How about a search engine that when it comes across a new topic, lets say it crawls to a John Tesh appreciation page, it sends a questionnaire to the webmaster that looks something like this:
1. Is this site informative about John Tesh? Pick one: 1,2,3,4,5,6,7,8,9 (High number is yes, low is no)
2. Is this about John Tesh's music?
and so on, it could use an AI routine to come up with appropriate questions based on bios or definitions of the topic. The search engine would be question based, I would type in "Who is John Tesh" and I get the most obvious hits listed but I also get a series of links asking me things like: John Tesh's Music, Why people don't like John Tesh, Photos of John Tesh, etc.
All done by AI and webmaster feedback.
Essentialy you'll get an informed series of specialized topics and their hyperlinks for every search. Sure it would spam the hell out of people, but the better it works the more webmasters will want to fill out the form to get a more accurate listing.
Project: a study of the current open source AI projects that are begging for input and are floundering for lack of direction. Get in touch with a broad spectrum of these projects with a web response and write up the pros and cons of each. Plan to spend time on your preferred project to enhance it in the best state-of-the-art tradition. Do it in a way that others can also put themselves in the study to find their preferences by making your investigation into a development story and not a re-invention of the wheel. For example, I have an AI project that is a special form of linear and non-linear PCA based AI primitive, called WinGrid. WinGrid is INGRID for Windows and you can use it to write your story. Right now I'm trying to get it into night clubs as a wall to wall mind machine. Essentially, imagine that I build multidimensional music kaleidoscopes. When running WinGrid really fast with music interrupts, you would see a spirograph transform pixels as a fast screen saver photo album from a central point inside a multi-faceted diamond that leaves traces of light as it turns. So that all this light doesn't all turn to shit, oscillators need to be added to make it run in time to the music. To do this I need help to write a background image loader that will pre fetch the next image without missing a beat. Interested? aroha, j. http://ingrid.netpedia.net
Something better than a Babelfish will require IA to work. It should have neural networks to learn more vocabulary and better its syntaxic and grammatical knowledge, can receive feedback from users about the quality of its translation, and suggestions about the meaning of some words in a given context. Eventually, with users' feedback and suggestion analyzed and stored into its neural net, it would be able to translate as good as an (apprentice ?) human can do.
Furthermore, if it works from the conceptual meaning of the text analyzed rather than from its list of possible litteral meaning, it would be able to understand and translate better.
sigmentation fault
Finding, for instance, data that is more related to a user's OS when seraching would be a nice feature. The problem is, for anything that demonstrated an appreciable amount of AI, you would have to go beyond simple searches. My recommendation would be to create an automatic moderation system for a weblog. I'd be curious how an AI would moderate a posts involving Natalie Portman and hot grits. :P
--
We may not imagine how our lives could be more frustrating and complex—but Congress can. – Cullen Hightower
AI built into a web browser that could automatically restructure a web page before rendering based on readability and usability. The browser could monitor your browsing (perhaps with feedback, i.e., asking what you like about various pages) and learn how you like information to be presented. Or, perhaps an intelligent text-to-speech engine designed specifically for websites and their varied designs, rather than reading all text in a linear manner. This would be great for blind people, or people who are just busy and would like a page to be read to them. Plus, it would cut down on eye strain. Another cool idea is to have AI software that could learn how a person uses the net, what they are interested in, what they are likely to be doing on the net, and what their hobbies and work are like, and then generates a personal portal when you launch your browser. The coolest use for AI in my opinion would be to create an OS or software that integrates with an OS and learns what your computing skill level is, and completely hides all complex features above your current level of knowledge. The AI would monitor your computer usage, and gradually reveal more of the complexity to you as your skill level grows (or hides stuff if your skill level doesn't grow). Something like this would seem to me to be the holy grail for tech support people, since it would truly enable a computer to become "foolproof".
An engine that takes an HTML file and removes all the fancy precize positioning, tags and what not, leaving standard-comliant, readable, simple HTML that gives you approximately the same layout when viewed in a browser.
Hell, just about ANY sort of intelligence would be an improvement, IMO. ;)
this may have already been said, but if you ever get a chance to read tad williams otherland series.. a good ai project would be implementing something like the character beezlebot.... an intelligent assistant that works within the framework of the web
Yes, I've written such a program myself. (Try to write it as the shortest possible shell script!) But some method based on meaning - or at least on something a bit more sophisticated than word frequency - might produce better results. Eg you could ignore not only stopwords such as 'and' and 'the' but also hackneyed phrases such as 'if you thought that X, think again'.
-- Ed Avis ed@membled.com
Web-controlled nuclear-powered death bots.
But you'd better invest in a good server b/c the site would see a lot of traffic. Well, until enough people had used the death bots anyway.
-- This and all my posts are in the public domain. I am a lawyer. I am not your lawyer, and this is not legal advice.
What about an intelligent program that would be run either on the server or client side that reads E-mail that you recieve, and then keeps track of which mail you open, which you delete, which you reply to, and how long you read each E-mail. The program could then recognize which mail is most important to you by subject or sender and sort them for you so that you do not have to waste time going through spam to find the most important E-mail.
After enough time, the program could begin deleting certain E-mails that the program recognizes that you would delete anyway. So if you recieve a fair amount of random porn site spam at work, and you delete them every time before reading them, the program would then recognize similar porn spam and delete it before it got a chance to clutter your inbox.
Many people use filters to do similar tasks, but an AI driven mail sorter could save time and it could recognize trends in your E-mail reading that you don't recognize yourself and respond accordingly.
Hipokrit
"I'm not an artist, I just hack art"
I think your project could go hand-in-hand with mine... Take a look at http://www.gwolf.cx/wrap
As it is right now, it exists merely in a oblivious theoric state only... But I think that pattern matching can actually be pushed beyond Perl's ability... And it can be VERY benefical to projects like mine.
One thing I hate about Netscape 4.X is the Smart Browsing. It never seemed to find what I was looking for, so I junked it. How about a truly intelligent agent that prefetches based on a history of usage rather than just the current page? If I had just completed several searches for subject X, then there is a high probability of me looking for more of X.
-- The Hollow Man
-- The Hollow Man
Non illegitimati carborundum
It doesn't take that much intelligence, artificial or not, to get a "first post".
/.-centric, would be something like Ask Jeeves that actually works.
Now, a duplicate story detector for Slashdot.. that would be interesting. Make a 'bot' that reads all of the articles ever posted on Slashdot, and acts as an extra level of screening (after "meat" approval by Taco et. al). If the article looks familiar, it could warn about it.
Another idea: an artificial moderator that would read all posts, and based on the way other things have been moderated in the past, tries to figure out how these new posts will be moderated. You probably don't want this doing actual moderations, but it would be interesting to see how well it's able to predict the behaviour of the Slashdot moderators. It could also potentially be used to check for moderation abuses... things that differ significantly from the prediction are likely to be abuses (assuming the prediction was very accurate).
One more idea, but one that isn't
All of these ideas are pretty heavy on the NLP aspect.
Quick Algorithm:
1) Rank all words in a selection by frequency of occurrance.
2) Throw all out pronouns, connectors, prepositions and other too-frequent words that are not nouns, verbs, adjectives, or adverbs.
3) You now have the gist of the article, still organized by word frequency.
4) Go back and find the sentences in the article that contain a large number of high-frequency terms. Print them.
5) You will find that you have just effectively summarized the article.
Actually you will find that you have merely listed a bunch of sentences with high-frequency terms. Use your AI skills to determine how to arrange these sentences so that the top ones *do indeed* summarize the article. (Directed graphs? Semantic nets? Internal references?)
DNA is a Turing machine. You, however, being dynamic and emergent, are not.
But part of the point of that story is that the
pure-AI editions of the book, even in the age of
ubiquitous nanotechnology, internetworking, et
al., pales by comparison with the version that
a real human actress is behind, even though she's
"just" doing what the AI tells her to. It's
hardly a testament to the power of AI -- more a
nuanced argument that, yes, it can do good things,
but humans can always do better, and some of the
consequences of AI are scary. (Actually a pretty
convential viewpoint, at that.)
It could be a good idea to make a robot that upon reading a Slashdot article would search the web, using appropiate search engines, to post interesting links related to the story.
You could use moderation as an evaluation of the quality of the strategies. Keep in mind that early posters have more chacne to be moderated (up or down than late ones).
__
__
Men with no respect for life must never be allowed to control the ultimate instruments of death.
GW Bu
How about some intelligent tools to help organize and put large volumes of information on the web. It would be nice to have a system that could take a large volume of information purtaining to a common subject, say docs from a court case, and parse through the information, organize it, and cross reference recurring information. This could all be published to the web automatically
I always thought it would be interesting to train a neural net to be able to detect when someone is attacking your computer. But I have limited knowlege of NN to know if this would be possible.
Several thoughts of mine...
You'd have an AI program with a web based interface.
Or
You'd have an AI enhanced web interface.
One of the former:
A program that digests and characterizes an mp3. Say there is a store of music on a sister server that people can download and listen to, and then score in several ways. Think Cinematch at http://www.netflix.com where people can rank their preferences and get statistically collated with other people who rank their preferences. In this case, though, you correlate tastes of a person with the music. So you ask the person who listens to rank on 1 to 5:
Slow . . . . Fast
Heavy . . . . Light
Sad . . . . Happy
Tense . . . . Relaxed
Simple. . . . Complex
Loved . . . . Hated
Where complex is taken to mean that the song is *both* sad and happy at places, tense and relaxed, etc. So the individual who ranks creates this 6 part characterization of the music, which is fed into some sort of NN and correlated with the music itself, somehow. The end goal would be to feed music into the system and be able to characterize the music correctly *and* decide with good certainty that a person would love a song or not.
It's a selfish goal of mine because there is too much music out there, and I know what I like, but of course I don't know what I haven't heard. Having a device that filters out 70% of the music I like correctly, with the remaining 30% left for variety and error, would be very interesting.
Just one idea!
-AS
-AS
*Pikachu*
The classic internet webcrawler with some more intelligent heuristics.
Have it search not on keywords but on content so a search for breast examinations will not return a bunch of pages for Human Anatomy 105 final Exam or a college girls P0rn site.
Have heuristics which determine which links on a page are most promising and which should be delayed until later. This could use domain name, url info, even a HEAD read to get meta tag info. But once again not a keyword search alone.
Build a database so it will try to find similar topics a.k.a a page dealing with only the term mamogram would also be noticed.
--- Linux... a college project gone horribly right
It might prove interesting (and lucrative) to create an app that has natural language processing to spider through webpages looking for comments/press releases/etc. about publicly traded companies and let it loose with a mock trading account. At least you have an easy metric with regards to success..
How about a blackboard expert system that provides personalization services for web sites. You could post objects to a blackboard that represents events and objects from the web site's domain. Then multiple knowledge sources could use those objects to post other objects on the blackboard that could then be used to deliver personalized content. for example, say a user visits several pages on your site, each page visit being "posted" to the blackboard. A knowledge source(s) could have rules that fire on these events to determine that, for example, this user is a good candidate for some product. The next time the user requests a page a link or whatever is served up. You could use jsp/servlets that interact with the blackboard to serve up the content. The key benefit of using a blackboard and expert system technology being that the rules can be dynamically changed and new rules added without re-coding a zillion lines of code. that's my 2 cents worth - maybe that's all its worth.
The difference between Links2Go and Google is that Links2Go learns a topical directory of links from an analysis of web pages. See, for example, links and topics related to Linux. Links2Go chose these topics and links because web page authors as a whole tended to use those topics and links to organize information on the web (not because a single human somewhere decided that they were the most interesting links).
This allows Links2Go to organize links by topic, allows users to navigate the topic hierarchy (topic "drill down") as they would with one of the hand-created directories, and bias keyword search based on topics and links related to the search terms. These are all features that are lacking in Google, because Google has no taxonomy of topics or method for classifying links by topic.
I like Google and use it all the time when I'm looking for specific information. But, when I want to research a topic in general and explore links related to a topic, I use Links2Go.
A portal site which interfaces with your web browser. It looks at which sites you visit most often and puts in excerpts and links to sites which would most appeal to the user.
I've heard that Salford University here in the UK does a lot of work with intelligent Agents. Their CS webpage is at http://www.cms.salford.ac.uk/web/home.nsf/34a0a1c4 dccd32d180256549004f44f4/4e388a29a352c50 7802565da0078fe1c!OpenDocument
I think you might have to register for this one.
I realize your objective is to apply AI to web pages or a web site, however...
I would prefer to see AI put to work in the form of a personal web search entity. Your preset web browsing interests are fuzzy-logically enhanced by an artificial entity's observation of your web browsing patterns. That is, when the entity sees that your interest in the subject of 'computer games' leads to you web pages with certain words or phrases or brand names, that information gets added and subtracted to/from the defining characteristics of that subject by fuzzy/intelligent deduction. During periods of inactivity the entity looks up more web pages of interest to you by way of search, news, and discussion sites, also by way of a server for entities to meet and exchange information. Older, smarter entities help out the new ones. You get a report on demand or an alert when your entity has some info it thinks you might want. Of course, a 'personality' on it would make it fun. Like the computer doggie from that old AT&T 'You will...' commercial, "I have a lead on the tickets you wanted."
"Thank you Fido."
You could have an artificially intelligent agent compare the content in the list of pages the user has visited without backing up. Then have the agent hilight or bold links to pages whose content relate to or inductively amount to the subject the user is looking for.
This can be done very non-invasively; it augments the surfing process rather than redirects it.
Furthermore, there is the opportunity for the agent to learn to better assist the user.
The agent has a vast knowlege base, the internet itself. The agent is not an active program which need process the information in that knowlege base, the agent does not require massive computational
capacities. The agent is smart in that it knows the patterns of the user, not because it has massive
cross-referencing capabilities.
---J-Cow--
Well, as for me, the only thing, which I miss a lot in Internet - MORE intellectual filter in search engines. Example: I have tired to balance on search "narrowing". If I supply more words - often responce is "no result". If I keep to supply just few words - the result is enormous in terms of resulting entries. So, it could be nice (at least - in my opinion) to have a filter, which implements more intelligent search without interface complication, with feature "never show me that bunch of stupid/porno/advertisment/promotional sites". More detailed about last feature: I would like to have something like "additional button" in every entry of found results with functionality "NEVER show me such kind of web sites in search results - it makes me crazy" ;) You could ask me - what is common in this feature with AI. My answer is here: it is too difficult for simple program to guess what I'm thinking about.
this field has been intentionally left blank
Here are two related ones. First, you could try a browsing profiler. The program monitors what you are looking at (over time, so single or occasional visits are filtered out). Then it checks pages regularly for updates, and notifies you when something new is added.
The other is similar, except that it is targetted more at news sites than at static pages. It is the opposite of the search engine. It monitors what you are browsing, generates a series of characteristics (say you are into roleplaying games with werewolves, news sites talking about Burundi, and movie reviews of action movies). Then it generates searches based on those characteristics, and can notify you as new pages pop into existence.
Just a thought. Looking back, it looks somewhat like the 'what's related' button in Netscape. hmmmm....
I have a vague recollection of someone trying to do this a few years ago - IIRC it tried to detect a certain percentage of um.. pink cylinders (limbs, usually) in certain configurations. Didn't work, of course.
... but what the web can do for AI. Obviously, client-side scripting begins to resemble neural processing. Consider training GAs with web-site-visitor input regarding, for example, aesthetics. Try applying ideas in heuristic search to links in web sites: just for kicks, come up with a situation in which you want something other than the shortest path or goal state. If at a loss, go for the gusto & do Perl:LWP crawls over newshub.com and try translating the results to FOPL. Do standardized requests from all the usual search engines, and based on the usefulness of the results, create a fuzzy controlled meta search engine deciding which to use for a given search. good luck.
- - -
"this world's been too many for me."
One up Amazon. The smart no click shopping cart tied in to doubleclick would anticipate your purchasing desires based on sites you visit. It would extract your credit card number from your hard disk and make the purchases it deemed necessary. Best of all you wouldn't have to click to sign up for the service. It would automatically purchase itself for you.
I'm off to the patent office.
134340: I am not a number. I am a free planet!
The hoax part could be pretty simple like this: cross-reference with the 5 best sites listing and explaining hoaxes, then add to the first lines of the mail (this could be shared as a joke too) that this is a hoax and not to be takes seriously and few links to sites explaining the hoax. It would require some intelligence to detect variations of a theme. Since we have language barriers it's next to impossible start judging the senders intentions so it's not good to just delete the mail outright.
Few sites:
Spam doesn't have the language problem (it's mainly english anyways) but it is certainly hard to recognize without some kind of AI (spammers will surely come around any simple script).
It would be good if this could be implemented as some kind of plugin/API so that any mailserver could use with as little variation as possible.
This would be ultra cool and nail the goddam spammers for good, it would also significantly reduce damage from hoaxes and even (hopefully) educate the masses of hoaxes (by telling what they are).
That's why I think we'll end up with hybrid systems when we start to see real AI agents on the web.
:^) that can try to figure out if these pages come anywhere near answering the question.
With XML we can provide some semantic clues. We can find websites that claim "Drew Barrymore" as a major topic, or celebrity interviews with Drew Barrymore listed as an interviewee. We can check the website content to find sections with some bearing on Drew, and we can even use fairly simple language models to make good guesses at the kind of content that website has. Then we can pass the data to some more connectionist kind of program (this is where the magic happens
I think that's a viable, useful approach to these kinds of problems. We can't provide full semantic markup with XML, but we can get part way there. Hopefully, it can be close enough that CPU intensive processes like neural networks can go the rest of the way.
What we really need is an AI widget that can actually find WHAT WE WANT off the net. But that's probably 10-20 years away.
An AI that can recognize the different meanings of things _In_Context_. That way I don't get "HOT PORN SEX!!" when I'm searching for "Heat sinks".
This is what we need.
--Remove chicken to e-mail
What about Love? Does it too? Fun? Hate? Good/Bad? Ugly?(may not be opposite of Beauty!)
If you can quantify what these things are, you are well on the way to the human condition. If you can make the results meaningful, try interpreting economics to benefit the most, not the privileged.
Try to invent a universal meaning engine for people who context things differently, not just subject/verb/object, but mean something different. Have it discover insights into authors, and find hidden meanings in political speeches, etc.
The list goes on ad infinatum, but noone has done any research on these. Philosophers, maybe, but not bit heads. Good luck.
This e-mail will self destruct in five sec..*
This mind intentionally left blank.
The KKK a bunch of sheetheads? You decide!
What about icons that expand or contract depending on how many times you've clicked on them? Or the level of audio output automagically compensates for changes in ambeint noise sampled from the mic?
Any sufficiently advanced man is indistinguishable from God
An AI that would provide news, hints, & reminders is all very well and good, but I'd prefer one that could learn. Ok, everybody knows about targeted advertising, all that, but that's crap. Kiddie Scripts do that.
What I'd like to see is one that knew WHETHER OR NOT to offer me these tidbits, these reminders about meetings, based on my behaviour within a page or a site... I'm talking about typing rates (to see if I'm being leasiurly in my browsing), modify its offerings based on click-paths (does he have seem to have a specific thing on his mind, or is he browsing at random? Short attention span today, or a longer one?)... that sort of thing. Immediately, it might work on that sort of speed-default setting, but over time it might learn to do some fun pattern matching withing my mouse-movements, tab-uses, etc., to provide more and more an experience based on what in real life we'd think of as non-verbal cues. THAT's be fun.
AHHHHHHH! I'm burning with goodness again!
- Reakk, Sluggy Freelance
More and more content on the web is being stored in databases with interfaces that are poor at best. By poor I mean that there is no way you can see all of the information on the site by following links. To see the information the site has to offer you need to intelligently enter keywords or phrases into text fields, or chose the appropriate combintaion of items from drop down lists. To further complicate the problem, you don't necissarily WANT to index the information on all such sites.
The next generation web search tool needs advanced AI to decide weather to index database driven sites, and what input is intelegent to get appropriate output from such sites.
I hope that the scope of this problem is not too great for what you were asking. Good luck.
Well, it seems that you are not really thinking of web sites. You are thinking about what used to be called "intelligent agents", specifically those which are running remotely and with which you can communicate in HTML. Another word for this would be personalized service with a Web front-end. A search engine is a trivial example of such.
So think about what cool thing can a remotely running program do for you. Find you stuff on the web? That's passe (do you want to code another shopping agent?). Filter news for you? Academia has done some interesting stuff here, not sure it it went anywhere.
You might also want to keep security and privacy in mind when designing your agent.
Kaa
Kaa
Kaa's Law: In any sufficiently large group of people most are idiots.
I have always liked the idea of forward searching web browsers. Especially for slower connections. It is arguable that as connections get faster, this will be unnecessary. As connections get faster, html coding just gets crappier, and content gets BIGGER. An intelligent forward searching mechanism would be a good project. Perhaps you could use the mozilla source, and then add this. The idea is to forward retrieve all of the links, sort of like wget, but dynamically. Supposing that you were to do this. The AI could determine which links are meaningful to you based on your browsing style. If you are skipping over porno banners, it obviously wouldn't follow links to porn sites in the search. If you spent a couple seconds looking at a page with many links, it could assume that this is a portal page, and you will be returning to it, and therefore choose to preload all links off of it (useing heuristics determined by your prior surfing of course). It could even make suggestions as to which page to visit next, which would be useful if you were using a search engine.
Eh...
I believe I have seen something like this before (but then again that's never stopped thesis authors before :). We have lots of search engines for web pages and their text. However, a crawler which goes through and indexes only images and has an attached AI unit which classifies those images somehow (that's your department) is relatively original and potentially very useful.
Maybe something like Ask Jeeves, except that you may want to make yours actually work.
Some companies (including Dell) have systems that reply to email tech support questions based on content. It would be interesting if you could create a general purpose system that chewed on a bunch of old requests and figured out some rules to classify them. It could work on spam too.
---
Have you ever noticed that at trade shows Microsoft is always the one giving away stress balls...
I would love to see intelligent spam filter, for
mail and browsing.
For browser, it would be a program that
would filter unwanted banners (and other ads) by content, and not necessarily by address.
I think that the mail part would be a lot
easier, but I'm sure it would be very challenging.
(sorry about my english)
When you smile, the world laughs at you.
I think a good use of AI would be to reduce the amount of information presented on the web. A lot of people get information overload (natural effect that reduces how many days you spend online in one sitting) and reducing that much information would be helpfull.
:P
An AI might also be able to reduce the amount of bandwidth something takes. Integrated into a browser or proxy, it might automatically lower the image quality of banner ads (I'm not in favor of getting rid of them) or marking them as ads. With browser integration, the user might be able to mark a specific image as "ad" and have the AI figure out what related sites/links, etc are also ads.
Another use would be to use the AI to help set the security settings for the browser automatically. In theory, if I were to wander warez or adult sites (notice I don't use numbers in an english word), the AI might be able to determine that most of the popup's, links, etc are all false or misleading and automatically mark them as such. This would speed up the navigation process considerably. Or, given some ethics, have the AI agent go those those voting sites and vote for the site without having to show the user the links. This is all theory, since I don't go to those site.
Another use for AI is to produce a client-side news service that pulls from various sites (slashdot, freshmeat, register, etc) and produces a simplifed listing that points to those sites (automatically stripping out duplicates). Yes, decent scripts can produce the same thing, but the AI might be able to handle changes to the page or even different types of information.
An AI could keep track of the pages you visit (automatically going up if there is user login, etc) and let you know when they change content, not just graphics. The AI might also be able to track moving pages and update your bookmarks as needed.
Another use for AI (get the impression I thought about this before) is to expand on the auto-login feature. IE remembers what password/login you use for a specific site. An AI might be able to automatically login into the site also. But for people who use multiple accounts at the same site, a low-level AI might be able to bring up a list of those login/password combinations. Some sites attempt to defeat these features and an AI might be able to recoginze them.
Some of these can be used with simple scripts, perl, etc but the AI might add a little bit more that makes life easier to you. The primary one is reducing information overload. We can process only so much information before we get "full" and a good summarizer might increase how much knowlege we can handle before we hit that point.
--- My novel, The Mummy's Girl is now for sa
So hey, after reading over a lot of really sharp comments, it seems to come down to this: whats the REAL role of the web once everything's settled? And how interactive/smart are we going to make multiple user server-based software, and for what reason?
.but it all comes down to what part of your web "experience" you want controlled for you.
.or maybe the 'smart houses'), I can't really see a role for AI.
;o)
I can think of excellent applications for AI in data organization, etc. .
Were it me, better searching, more intelligent queries, yadda yadda sounds great, but until the web starts filling a greater role than it does (like perhaps when all these 'internet appliances' come out. .
My two cents!
S
http://students.washington.edu/steve0/
steve0@u.washington.edu
- - - - - - - -
Don't worry, being eaten by a crocodile is just like going to sleep in a giant blender.
Since the web is a world wide community it would be useful to have a good translation system. This my not seem like an application of ai but it is a really complex system that I believe could benefit from ai. Another good project is a search engine that would actually find exactly what you are looking for first time, no problem. That would be nice ... I would even pay a monthly fee for that service! Regards. "The pinnacle of success is not reaching the goal but the road to reaching the goal."
Keywords and blacklists are too blunt, wouldn't it be a challenge to make a *useful* filter?
I don't mind people looking at naked bodies, but I would very much like to be able to do a "sex AND NOT [porn]" search for example.
(and if it drives the snake oil salesman known as Cyberpatrol at al out of business, I wouldn't mind)
All opinions are my own - until criticized
An "information manager" type application that works with the web AND has data-analysis intelligence built in. The features I'm personally after would look something like these:
- integrated into browsing (Mozilla?). Sort of like Pyra's Blogger -- if I find something nifty, I can just click & add the URL into my info-base, without going to another page, without bringing up a dialog box, and without having to choose what "category" it fits into.
- data-mining and text-analysis. The info manager should be continuously trolling through my collected links, retrieving & saving remote URLs, analysing their contents and making suggested links based on the analysis (connected to Google?).
- The analysis engine should be able to refine itself over time, observing which recommendations I agree with, and which I ignore (like a super-powerful epinions.com).
- The analysis engine should also be able to pay attention to my personal classification schemes, and start organizing things accordingly.
- as an option the analysis engine should be able to check my current surf patterns, and notify me of related URLs that it has found on it's own.
- all of the data it uses should be encrypted.
- it should be extensible, so that I can add filters to other (non-HTML) data formats, enabling the analysis engine to read PDFs, etc.
- it would be nice if the system was able to store it's raw data on a remote server (encrypted) and use the server for the heavy-duty stuff, passing down the results to my client.
- it would also be nice if the "remote server option" I just mentioned was an option -- I should be able to have everything run on my desktop if I want.
I could keep ranting, but I guess I'll stop here. I want this sort of app, and I'll pay for it. Anyone got any recommendations? I can see these functions in many disparate apps, but haven't found anything that sews it all together.I have no
You should go ahead and make a massive intelligent agent that cruizes around the web, gathering information it finds "relevant", "interesting", and eventually contacts a similar agent somewhere in Centuari.
---
How long have you been listening to the world's famous?
'Bout six weeks.
Six weeks!
This sounds like the Giant game that Ender played. _Ender's Game_
May i suggest a little different objective? In www.worldforge.org is a project of making a free ultima-online-like game, and one of the objectives is making the NPCs (Non Playing Characters) intelligent. There's a good overview of the objectives at here As you see, is a very interesting project. For the lazy, the idea is making NPCs that behave realistic, not that LOOK realistic, so that they have goals and ambitions, transmit them to their offspring and make decisions based on the environment. Very ambitious, but that's the cool part, isn't it :)
"Now you can see that evil will triumph, because good is dumb!"
If you could more accurately predict user behavior in a browser, you could preload links and cache more intelligently. (of course, the former is internet-community hostile, the latter internet-community friendly)
You could also do this kind of preloading on a larger scale by monitoring the server loads, and dynamically changing the content that is preloaded on web pages to anticipate user clicks.
Thank you for not thinking.
Consider the way that Google can identify valuable (or at least popular) websites without any such clumsy user input. Is there a way you could identify a valuable slashdot posting by looking at user reading patterns? There's a lot of different kinds of data you have to work with: How many people read the thread, how much time people spend before moving on, numbers of responses, clickthroughs on posted links, and so on... perhaps all weighted by karma?
You could also try and evaluate a posting based on certain heuristics, though I suspect that would rely a lot on obscurity.... e.g. if people knew that a posting with three URLs was always given credit for being informative, you'd see a lot more suck.com style linking.
On the other hand, you might be able to do about as well as a lot of slashdot moderators.
simply someone(thing) to talk to. it can remember your days schedule, talk about the news with you, read some funny news highlights, all the while learning your behaviour and interests... different web site would have different personalities, unintentionally because of their fundamental structure, but consequently appealing to different groups of people.... this used to be popular in pre-GUI days and seems to have dissappeared now that we've "graduated" away from the command line. the whole idea could be simple (as does appear on the net now) but should really be leveraging the latest AI theories. sounds like a great grad-project. publicity, fun, cutting edge..... abraxas1
1) AI to scan images you load & compare to previous ones, keeping only one copy of identical images with different filenames in your cache.
2) Instead of centralized search engines, develop personal spiders that seek out topics of interest.
3) Agents to fill out useless marketing questionnaires with amusing input.
4) A personalized page full of personal slashboxes, dynamically generated from the sites you visit most frequently, self-correcting based on a neural net which tracks whether you use the slashbox.
5) A neural-net visual recognition of satellite images which can warn you if bad weather is approaching.
6) A voice-controlled browser which allows you to 'click' by reading aloud the link you wish to select and responds to simple commands like 'scrollup' 'scrolldown', 'back', etc.
Just throwing some out there.
These people looked deep into my soul and assigned me a number based on the order in which I joined.
Check out the Agents Group for this and other projects.
Any sufficiently advanced technology is indistinguishable from a rigged demo
--Andy Finkel (J. Klass?)
I don't really know how much it pertains to actuall AI or not, but a detector/tracer type utility for DOS attacks might be a good idea.
I saw a paper pertaining to that the other day if I remember correctly.
Driving backwards on the highway of life
in order to ultimately begin to make intelligent asscertians re: what probably will get mod'ed down or up
You realise that if you could do that you could use it to play the stock market as its just a giant moderation system except its denominations are dollars rather than Karma.
Any sufficiently advanced man is indistinguishable from God
On the other hand, I can think of useful AI tools on the web, namely good search engines. Some of the clever ones try to match ideas rather than just simple text pattern matches. Perhaps you could work with that area, but it's nothing new.
But this still seems really silly. If you can't think of any good new innovative uses of AI to use over the web, then why are you asking us? Why focus on that small area of potential for AI applications for your project? More importantly, what are you going to do with any new ideas that do show up here? Are you going to give credit where it is due, namely to whichever /. reader gives you the thesis idea? And is this really what your profs have in mind when they ask you to come up with a project to prove what YOU can do? Is this going to impress whoever has to judge the value of your project?
You know what to do with the HELLO.
You know what to do with the HELLO. ...
Help create an open-source world
How about a web browser that attempts to determine which link we're most likely to click next, and then goes out and caches that page for you. If you click that, it comes up FAST. If not, no big loss (kill that transfer instantly and begin a new one). I don't know how feasible it is to guess which links the user will access, but that's the AI part.
Visit me on #weirdness on the Galaxynet.
Yeah, that occured to me while I was posting, but I though I'd just seen Pi one too many times.
Much Love,
"S"HM
*****
(I refuse to spellcheck out of contempt for your belief system)
Also, DirectHit is what's powering the beta version of HotBot - which incidently uses the Cyc Engine & KnowledgeBase.
The Cyc project seems very cool - but I've not even seen any of their stuff running yet. (Despite attempting to implement an inference engine myself around the core of their KB!)
Whilst we're on that subject, the core of their KB is available for free download & usage - even in commercial apps - providing you give them the necessary credit in the app (etc). Check out their website for full detalia...
HTH,
fRoGG
The best application of AI, I can think of to enhance my web experience would be for an AI to be able to recommend new pages that would be interesting to me based on other pages I have viewed or based on a profile I filled out. The system could create a new section in my bookmarks called "Recommended Sites" as I surfed the system would automatically track the sites I visit and determine what my interests are. It would then search the web for other pages that fit my interest profile and add them to my bookmark file. It would also alert me to new sites with a pop up window when it finds a particularly relevant page. For example if I were looking to purchase a new computer and I began visiting site for DELL, COMPAQ etc. the system would realize I was interested in purchasing a computer. If it could not determine what price or specification I had in mind the AI could ask me a few questions to completely specify my request. Then it would look for computers meeting my specifications all over the web (auction site, small vendors etc.). When it found a number of sites it would then format the list of computers by price and features and then pop up a window allowing me to jump to the site to obtain more information or purchase. I understand this is a very hard project. The AI would need to have the ability to make assumptions and inferences based on a built in real world knowledge base (like CYC). But all of AI is hard; if it were easy everyone would do it. Good Luck
There is a piece of software called AI RoboForm.
AI in the name of the product does really mean
Artificial Intelligence. It's one-click form filler.
Why is it AI? It imitates user reading and filling the form. It reads and matches field captions, not field IDs.
Sounds like simple AI to me.
This could patroll newsgroups and/or forums, and have the FAQ for that forum/newsgroup, it could read messages, decide if that message is asking a question which is answered in the FAQ, and then autmatically answers the persons question.
this would be nice since most forums I've been in have people asking the same questions over and over, and people tire of answering them, so sometimes they dont bother after a while.
-I go to Rice, so figure out my email address
I think the most obvious expert system application would be tech support.
There are thousands of tech support pages for every possible thing from Networking software to portable CD playeres.
Most of the suck, a few are pretty good.
There have been many attempts to provide some good engines for this.
If you succeed in creating something which works better than what's out there you will certainly make my life easier.
Question Reality
I've wondered if a neural network could be developed to do more efficient routing of IP packets. I hypothesize that such a network could approach the performance of human-tuned QOS systems withut requiring the explicite tuning.
Feel free to steal this idea - I don't ever intend to develop it!
-sam
I perfectly agree wida idea....but it defers from the basic question being asked. It's specifically stated how AI is incorporated into web. What ever the educational/fun stuff can be incorporated for that matter to any software. But on the web the most important issue is information retrieval.
Or, this could end up being a filter that an individual user could apply to his/her Slashdot viewing, so that moderation reflects his/her tastes, data-wise.
Much Love,
"S"HM
*****
(I refuse to spellcheck out of contempt for your belief system)
Computer Associates already does this. They use neugents which are snmp agents (and thus integrates into CA's TNG managment tool).
They just dusted their old OO jasmine database Engine and throw it to the face of the world like the E-Business solution
You should look closely at what's already available, to see if you can do the same kind of things, or yet better ones, using free tools.
none Yet.
I don't know if it really qualifies as an AI project or not, but I'd like a search engine
that looks at what I've clicked, then picks the most likely candidates that I might pick again
(realizing this make take a couple of clicks for it to develop the pattern), then in realtime
starts checking sites to see if the other URLs I might choose actually still exists or not.
I'm really sick and tired of AltaVista and WebFerret returning thousands of broken links.
I've maintained for a while that the first 200 (or so) hits returned for each search each day
should be put in a list and checked for validity.
This is slightly-OT, but I thought it was interesting. The current "state of the art" in web-based AI is the "AliceBot." It's a conversation bot that won an international competition for AI.. it was dubbed the "most human computer." It's pretty interesting, but as you play around with it for a while, you'll get the impression that it really isn't all-too advanced. While I realize that AI that generates conversation may be much more difficult than any AI you're planning on coding, I still think it's interesting to look at the current leader in the field. I'm sure trying to make a computer seem "human" and trying to use AI to tailor web experiences are two different things, but I still think there's a cross-over.
--
--
Entropy isn't what it used to be.
I've just today finished the last of my 3rd year project at Sussex Uni.
I'm not sure how you'll be assessed, but the thing I found most important was that you are assessed on your report, not necessarily on how well what you do works.
If you can try and target your research to something that will allow a good write up, then you're on to a winner. For example, someone did an email client that attempts to learn what you do with emails. The thing that made the report good was that he was able to test it on different people and collect data and evaluate it.
FOR EXAMPLE I need to know how many switch hit home runs did Mickey Mantle hit against the Dodgers, and compare that to the total home runs hit by Mark McGwire against the dodgers. The user speaks this request (once again only needing to learn the "Guide" and the guides user interface (beign vocal).
The Guide also knows me. Knows what type of movies I like and why I prefer theatre 1 vs. theatre 2. I then come home and ask my "Guide" "What movie should I see tonight?" and not going to a website and checking to see what movies are available to see. The Guide would do this for me.
Where does this scifi suggestion fit into the web.......www.ask.com on speed. They have the idea using todays technology...give it a few years (ok maybe 10) and we will have our very own "Guide" that knows us and will ultimatly, SURF THE WEB FOR US
An auto-Napster account maker/changer would be useful for when Lars catches up with you! :)
Any sufficiently advanced man is indistinguishable from God
for extra credit, write the code gen utility that parses the application's pages and introspects the app server's classes for tunable bottlenecks (maybe it finds places to insert a queueing interface or an object pool) makes adjustments to the architecture as necessary, sets up timing triggers, configures the genetic algorithim, generates the first population of config files, and iteratively launches the application under one individual against a load test tool....
ok, so its not the Internet at large, still, it could be fun!
E.G. You're reading an article on quantum mechanics on CNN, your links on the bottom bar (I'm thinking where the netscape icons are on my browser) would link to the university and department front page of the group doing the research. Additional links to other University research postings, trade magazine sites, government contracts for research, maybe the home page of the head of the department mentioned in the article.
A step further, take an article on earth smart cars, links would go to sites about companies that offer them, research about, and fact pages on polution, viability, materials, and history of mechanics.
Could help for when you finally find a site (out of the 1 million hits returned) that you actually wanted, and want to know more.
Ctimes2
My cube. My friend. My solace. My prison.
Oh yeah, lets have an Eliza-ish convo-bot,
Bot: Do you like Hot Grits?
Me: No thanks
Bot: Do you like Natalie Portman?
Me: No!
Bot: Not even petrified?
Me: No!!
Bot: How about Natalie Portman, petrified with hot grits down her pants?
Me: Noooooooooooooo!!!!!!!!!!!
Any sufficiently advanced man is indistinguishable from God
There needs to be a real push into getting intelligence in educational software (for kids). Most of what I've seen is drek - while some of it is very slick and good-looking, it lacks real educational content. You either do not learn anything, or you learn it once and then repeat it endlessly.
Here's a challenge for your AI - adaptive educational software. Most software today requires the child to 'log in' so it can keep track of their saved games. Go further. Keep track of what the child does, how successful they are, and tailor the next experience accordingly.
Give rewards for progress. Reduce the rewards for continued success at the same level (gradually). Prod them into more difficult problems / puzzles / challenges. Eventually remove the lower, introductory, levels all together. Give different rewards.
Do all this while keeping it fun, and keeping them coming back for more. Pop quizes to keep them sharp - reward those accordingly. More advanced information (kind of like sidebars), when they are ready for it, can appear as options. Almost a tutor / friend relationship.
Teach the young how to learn - what could be more challenging for an AI project?
Personally I would like something that could trawl the net and report to the webmaster of a site all it's broken links with possibilities of the correct (new) links.
I even have a timescale for you:
If you have 5 months for this project:
First week: get excited about it - feel that you can really make a difference.
5th month: Realise you haven't got anything that resembles a project
Last two weeks: Find something similar that was submitted 3 years ago to another prof. and pull all nighters until you have a half mediocre project.
5 months 1 day: Go out and get drunk - because you feel you deserve it.
(even though you clearly don't)
Why break the tradition.
:-)
I'd really like a search engine that allows me to put in a natural language question, and parse it intelligently. AskJeeves doesnt even come close...
I don't know very much about AI, but I seem to remember something about an AI having the capability to learn and draw conclusions from experiences. If that is so, would it not be possible to make a program that scans all the pages the user visits and then based on that information 'improves' the hits on searches to be more intresting to the user, and maybe goes out on the web and finds pages to recommend to the user.
The recomendations and hits would then improve the more the program learns about the user. It would also see what recomendations the user follows and learn from that.
Phobos - Greek word for fear or flight
A little advanced for your average user but this might be effective. Lets say you are a day trader and want to know all the news having to do with your large portfolio of stocks. An internet App that could find sites that deal with that subject and continually check it as well as continuing a general search would be good. Of course you would need to include a number of different AI concepts in the programming. First being language comprehension so that the program would be able to acurately identify good information. Second would be learning so that the user could identify sites and content that is not wanted, and the program would learn to throw out any other information "similar" to the rejected information. The program could then sit and watch its prefered sites (ie sites that it found good info on) and notify the user when there is a change on that site. This though might be a little more complex than what you are wanting to do.
I realize that neither of these is particularly original, but then neither have they been perfected by anybody; there is still lots of room for improvement, and plenty of scope for adding as much natural-language AI as you like:
:-) Seriously though, it would be useful for busy people who haven't got time to read things from top to bottom. It could have some way of determining which links are relevant, following them, and adding their information to the precis generated (with attributions if necessary). Or you could use it to cut out types of information you don't want - to filter out marketingspeak, if you're a techie, or to filter out techspeak, if you're not.
A precis program which will condense long websites or discussions. Jon Katz's articles on Slashdot might make a useful set of tests, also try to precis the comments posted to them
The other suggestion is a remembrance agent which looks at the website you're reading and suggests links (from your browser history, from search engines or from some big collaborative database) which might be relevant. This might finally be a use for those sidebars that recent browsers seem to have spouted. Again, this has been done before, but it's not something which has been done perfectly. You might also be able to use it as a fact-checker for postings you make to Usenet - although that would be rather difficult to implement, I imagine.
-- Ed Avis ed@membled.com
...................
...................
...................
I thing that the main difficulty/need on the net is to find pertinent info ! Search engines exists but indexing isn't enough for me, in particular the info has to be 'rated'(evaluated) to be truly usable. In fact I'd like an agent which would provides me some links and enhance his evaluating function with my feedback ! but to be even more efficient this agent should reevaluate his function with other agents of the same family... Let's imagine that Agents exchange 'links' and Evaluation function parameters... (GA comes to mind...) Each agent should be able to evolve and make its evaluation function better using other agents parameters... I mean If agentA gives another agent (AgentB)'good' sites (sites well rated by AgentB user feedback) , AgentB should use AgentA parameters (combine with its own) to ehance the evaluation function... GA for the evaluation function tuning Neural network for the evaluation (Neural network should be efficient for site rating -> Input=request , output=URL : evaluation=user rating...) Don't you have enough to play with ? BTW : When I say evaluation function you could even think indexing functions...
This is not a plug, but people want to tailor the net to the individual. Here's a link to one company which proposes to do EXTREME PERSONALIZATION of the net. Only time will tell if this is a winner. http://www.rediff.com/us/2000/feb/14us.htm
Basically something that can be taught new data formats, and extract information for me and collect it... heheh ;) How's that for some vaugue things? Well... maybe hopefully it'll get ya thinking towards something cool!
MP3.com already implements a system by which it looks at the artists you download and comes up with a list of other artists you might also like. This could DEFINATELY use improvement, as some of the choices are based only on the name of the group, and not the style of music.
Now this suggestion might get me marked as a troll, but it is definately a money maker! The Adult industry is one of the biggest money makers on the internet. If you could develop a system that would monitor the sort of types of pornography or the type of girls (brunettes, blones, racial preference, etc) that people tend to look so a company could target certain pictures at different people, any site would be begging you for a copy of your software!
-- From my Best Friend (Written to me over ICQ): "i was gonna go to a party...but i had to reinstall windows"
The immense gulf between interacting with humans and interacting with computers is that computers lack context sensitivity.
During a conversation with a human, you can communicate very efficiently because you both share the same knowledge of the context of the conversation. Thus, two or three word sentence fragments can substitute for entire concepts, like "There" instead of "the information you are currently looking for is located over there."
With a computer, you need to explicitly feed it all the information it needs, in the correct context before it will return a satisfactory result. Misspell a search entry and the computer will gleefully return all instances of "Slashgot" it find. (the exception is certain search engines that incorporate spelling traps as part of the parse function, but that falls under the heading of "clever" rather than "artificially intelligent.")
Of course, for a computer to react "intelligently" to your context, it will either have to laboriously form a worldview based on the limited information you provide (tracking the sites you visit frequently, forming a list of preferred search strings) or we'll have to provide expanded sensory inputs for the computer. Currently, if you delete a file you needed, you say "Oh darn," but you don't notify the computer that you made a mistake that should not be repeated. A future, sensory-enhanced computer could perceive from the tone of your voice and from the image of your face in your hands that it should, in the future, not REALLY delete the Photoshop file you've been working on all day, but rather move it to a hidden location for possible recovery, or force a backup to remote media before it deletes the local copy.
So, to answer your question, I think AI has great value in expanding the machine's understanding of your needs and wants outside of the current, severely constrained communication channels.
Marc Siry || interactive media professional, motorcycle enthusiast ||
I'd like to have an intelligent agent that would post Interesting=1, Insightful=2, Funny=1 messages to my /. account.
--
Sheesh, evil *and* a jerk. -- Jade
Mix
- natural language parsing
- web crawler / discussion group logger
- intelligence
Get
- persons (id by nick/links/style)
- topics a person discusses
- depth and linkedness of topics
Provide
- lists of specialists on a wide range of topics
Probably nothing too new, but the reqs/specs should be adaptable for something useful and implementable.
I think, therefore thoughts exist. Ego is just an impression.
I think there are a couple of interesting things to do out there. Your indecision seems te me to be the problem.
I would suggest you to pick something that previously caught your attention and try to add a touch of AI in it. It is not of the uttermost importance to be something that will change the course of the world. Remember - you are only one person, and setting a huge goal may end up by doing nothing.
A long journey begins with a simple step (no guys, this is not a chinese proverb - it's just common sense).
I can't tell you what to pick - I have no idea what your interests are. But I can tell you that I found a very rewarding "landscape" in the field of computer vision. I had to make such a project myself when I graduated, and I wrote an OCR
But as I said - it doesn't really matter to write something huge - it is more important that you learn from it.
And, yes, I'd like though to make a suggestion somewhat related to the previous answers. The paperclip in Word has some good AI behind it. Don't flame me for this - I usually don't put the words Microsoft and good in the same sentence. But in this case it's not the AI that's bad - it's the idea of having a paperclip dancing while I am doing some work. I mean, why the hell is that thing stretching on my screen while I am trying to do some work ?
Conclusion : do something, don't set a goal higher than you could possibly reach (for now), and don't make it annoying.
For years, I've wanted a system for my house to make it smarter.
In particular, I've wanted to keep my complete music database on-line, and as tunes are played I can assign them a rating. "Computer, never play this Metallica crap again!" The system would scan the existing database and any new music added, and would assign a probabilty to it as to how I would like it (based on rhythm, preponderance of bass, etc), and what category it might belong in (background MP3 downloads come to mind).
Songs with a lower rating would be played much less, and if I specify a genre, the music would be drawn primarily from tunes match the genre. Then, assuming that my mood changes over time (having learned what I "normally" do), the system might start adding tunes from other genres over a period of time, until it's playing my normal rating.
If it sees me doing something totally out of character (say, playing crappy Metallica...), then it would know that ridiculous excursions shouldn't be added to it's "knowledge" of me.
It should also know that if it detects a high quantity of airborne THC particles then it should only play tracks from "Dark Side Of The Moon". Oh yeah, and automatically dim the lights.
"an intelligent Slashdot filter".
;-)
Hmm. If an ad filter filters ads, what does a Slashdot filter do?
How about trying to improve on the Eliza bot. I'm fairly positive Eliza bot would have been covered quite significantly in you AI class. For some more info check out http://www.msu.edu/user/gilber63/paper3 .htm.
I've always been intrigued by this bot and since (according to the above article) she was created back in 1987 I'm sure there would be lots of "updates" you could make.
OTOH, you may have covered this extensivly and this could be very unoriginal. Whatever the case, Eliza is a prime example of AI at work.
Good luck,
LiNT
i have a couple ideas:
1. if you can make a bot that i can talk to in between slashdot posts, that would be great. of course it would have to pass the turing test, but that's not too hard. often times i am constantly hitting refresh, and getting very lonely at the hours staring at the same post. if i had a person that i could talk to, someone who i can relay my feelings too, that would really make me happy.
you could call it 'slashbot'. but i don't wanna think i'm talking to a bot, so maybe 'slashthisreallyisapersonnotabot' is better. but then you're lying.
2. here's another idea: maybe make an AI being that could find the question to life, the universe and everything. i built one in my AI class that solved it (contrary to popular belief), the answer being 42 of course; but i neglected to find the answer.
hope this helps. IANAL.
I think the biggest challenge for any artificial intelligence is still to fool humans into believing they are human (the opposite is quite common)
In the forseeable future, that's a thing that will probably happen. Also, the presentation of the web will change, just like it has changed before. Greater data-speeds and improved hardware will make things like VR common.
When it happens, these AI-machines can and will be used in the entertainment industry. Croupiers in virtual casinos, virtual girls for pr0n, stand-up comedians that only exist in cyberspace.
I don't know if that's far fetched, but it will allow you to combine web and ai, and at the same time give you a social angle.
----------------------------------------------
the pun is mightier than the sword
How about being able to simulate a user? I know UseIt says that you don't need very many users to do testing, but most web pages can't afford to pay any, especially if they make a lot of revisions. At it's simplest form, the program would check that the fonts are big enough to read, none of the basic guidelines are violated (pop-ups, etc. check the questions for Jeff Zeldman for more ideas) and so on. Then, as the program progressed, it would watch users surf the web and try and figure out what kinds of things they did.
This way you'd be not only studying users actions, but creating a program that would learn from them to either formulate rules about usability or simulate a user itself.
On the client side there are lots of places where current technology could be expounded on. Internet Explorer (and other Microsoft software) tries to implement some basic tricks like URL completion, automated form completion, menu optimisation, evil paperclips, etc. Any one of these could have an entire thesis devoted to optimising it.
Granted, all of these can be done with statistical models. If you're not so interested in the math, figure out what neural nets or evolving software would be optimal at.
I would really like some sort of intelligent agent that would run my life, or at least the boring tedious bits, filter news would be good, but also i think that an agent that interacts with a home network could be good, say doing the shopping for me when i run out of caffeine, regulating temperature, playing random music from a jukebox every morning, i know this can be accomplished individually, but how about doing them together, maybe even with a XML interface that can be accessed through various internet connected devices, like if i want to record X-program, i just ring in and tell the device and it does it, damn it, i would even put speech synthesizer "You have no clean pants left", or "Yet another microcrap post at slashdot".
-I can only program my video,ahh, I am not a gook, but a joook -The World is a theatre of the absurd
Beyond that, in order for this to be an interesting project I would have thought that you need to exploit the information (or at least data) richness of the Internet.
What I would like to see would be a navigation companion. Not merely a history list, but a service which watches what you look at and then goes looking for similar content which it reasons you might be interested in too.
The service would operate in three phases:
- tracking - where it would effectively be that dumb history list, albeit more useful because it would build up a network
- discovery - this is where running on a server becomes useful: after you've logged out, this companion will run its inductive process across your behaviour, your preferences, and your navigation history to look for things you might be interested in next. Then it goes off to search engines, or crawls itself, or watches web sites you've shown persistent interest in, to find new content which you might want to see.
- presentation - next time you come back, there's a little flashing light or other widget saying that there's new content. You read.
Now I confess that I'm not sure how phase (1) would be implemented since this might be better approached as a client side tracker (maybe an application within Mozilla?) but once you've got that data then you have a fair amount of material to work with.Start again at phase 1.
--
Dunx
Converting caffeine into code since 1982
One of the things that comes to my mind ponder AI and the web is smarter search engines: use the power of classification systems (say something like self-organising maps like WEBSOM in order to get something like useable semantic nets. For example, I'd love to see a web engine that, when searching for "Serpent" or "Blowfish" would ask me "Are you looking for an animal or an encryption algorithm". Also, this would make it possible for a search system to produce hits that don't use the literal search term(s), but only synonoyms.
You could do a search engine, I don't think AskJeeves really counts as AI. I'm sure you're looking for less saturated ideas, though. Web-based AI may be a bit premature right now. It seems as if most of the stuff done on the web relies almost exclusively on user-input, and I don't see a lot of need for a server-end AI application. Perhaps a client-based application that would work with a user's tendencies involving web-browsing and such. Say, for example, a program hat would know what pages to open up for you to read when you asked it (or would prompt you to open said pages, so as not to be too invasive?), what directory to download certain programs to - it'd be really convenient to not have to type g:\quake3\ every time I'm downloading a patch - the client could take context from the page I'm downloading from (or trace my usage from a certain page to known mirrors - i.e. from a page that mentions a q3 update on fileplanet.com to a cdrom.com mirror) to know where to put its files. I can't think of any really large duties for a web-based/oriented AI program, but I can imagine a lot of small ones.
Moo
...clients are. The obivous answer that everyone wants to say is customized news. The information on the web needs to be in a wonderful XML type format and your have intelligent clients that learn what you want to be informed about, what off-the-wall news bits you want, then go out to sites and find the news items you'd like and deliver them to you. This allows these AI clients to be much more in tune to what you want since you'll go thru a single one for almost anything. Rather than have AI's on dozens of sites that are really no more than preferences (like Slashdot) that don't get the chance to know you since the site needs to have room on their computers for 1000's of sets of likes/dislikes for the end user.
<--#insert file="witty.sig"--
Tim Berners-Lee and James Hendler is working on DAML- The DARPA Agent Markup Language. Most of the AI guys (that I know) that are doing web stuff are writing spider/crawlers. The interesting ones are using RDF and XML spiders to be able to search by content, rather than keyword.
"There's so much left to know/ and I'm on the road to find out." -Cat Stevens
I'd like to see a system that rates music by "listening" to it and determining basic things like melodic content, key, bpm, etc. in addition to the usual artist, genre and so on, and starts recommending music for me based on my past choices. It's really getting to be old hat to have music recommended based on what other people liked and how they rated it. I want to see a more impartial, objective method that works... maybe even better.
I say this having previously worked at a company* that did expert system stuff (I was in the web dev. dept. though). There are tons of potential uses for expert systems on the web - we had most of our success with advice-giving type stuff. Our software allowed a "knowledge engineer" to work with some sort of subject-matter expert to distill their expertise into a set of rules - which could be simple if/then, or could have a complicated weighting system. The upshot of this is that a user can surf over to your website and, after answering a few questions, be given information that was specific to his/her concerns, rather than being forced to wade through a heirarchical filesystem full of stuff that may or may not be relevant. I don't know how interesting a research project this is - but economically speaking, if you can throw together a fast, scalable, reliable expert system tool that integrates with databases & webservers - you could do very well.
*note that said company is now out of business due to venture capitalist stupidity
This could give you some ideas for a project.The idea of chronologically listing search results according to some rankings is a limiting factor, especially so when more and more useful information gets onto the web. Some kind of visualization or ordering technique could be designed which will give the user more control for searching. A better human computer interface (HCI) if you like, for a web search engine would be a useful project. The term "three dimensional" seem to crop up. Who knows, you could be the next Jerry Yang :)
How about a little AI browser plug in that logs where I go and from that information determines what I like to see on-line and gets it for me. For example, if everyday I get to work and fired up Netscape and surf over to wired.com and read some things there, then I got to CNN.com and read the US News and tech news and finally I head over to Slashdot.org. Have the agent log that information and when I get to work in the morning, Netscape is already open and displayed on the screen at the headlines from wired, CNN US and tech news and Slashdot. That is a somewhat involved project but it would be a cool one to work on.
Keep us posted!
Despite some of the negative things written about the use of AI down the road... I think it will be huge. The uses for AI will grow slowly but surely, and I'm a strong believer that one of the biggest applications for it will be "smart agents". You know the deal... have it run your life for you when you're not on-line. Looking to purchase a particular antique vase? Have your agent search the Internet over a period of time and notify you when one goes on sale. Looking for news article relating to war in Bosnia? Have your agent notify you when news becomes available.
The applications for smart agents is unlimited. Write a smart agent that performs a particular task. Maybe make an agent that scours slashdot for articles/posts of interest so I don't have to read the crappy ones. (No offense you anonymous cowards.)
This is a really shitty idea. We definitely don't need a computer moderating for us. If we need more moderation, then assign more people as moderators. A computer won't do better than it's programmed, and it's programmed by one person - which means our filtering is the result of a single set of values rather than a distributed set.
Plus, we don't need to help those who want to write filtering software for other purposes.
I'd rather he worked on a slashdot posting agent! Something that combed the web for potentially interesting Slashdot stories for us to read and comment on.
First, make it work, then make it right, then make it fast, then, make it bloated!
I was playing around with a pre-caching engine that would build page responses for a user based off previous traffic, then I got too lazy :-). eg.
If a user always goes to slashdot.org then to slashdot.org/index.pl?section=apache the apache link is dynamic and must be created when the user asks for it. You might be able to use AI to predict user patterns and increase response time by pre-caching that page. (Just my $0.02)
Idea 1: The WebScreen.
I know with the rapidly expaning of the web that finding new sources of good information about any particular topic has become more difficult. Search engines and news sites have become generic ad-agencies, yet I know that there ar very good sources of information available. Getting to them is the hard part. The process I ussually follow invoves going to the few selected trusted information sources. When that doesn't work I usually go to one ow two types of search engines. Those that index the web by keyword (Google, Raging, Altavista, Snap, Infoseek) and try to think of unique keyworks that will produce the best results. The second is a web index that categorizes the web by topic/ subtopic such as Yahoo or the ODP. Then I review the result summaries for promising sentences.
Idea2 Security Abuser.
It seems to me that most security holes in servers and other software fall into specific categoeies. 1)Backdoors intentionally installed during development 2)Workarounds not accounted for in the programming logic 3)Weak spots in the software infrastructure that can be exploited 4)Using a tool/command/software for other than its intended purpose. If a smart program could be written to scan and debug software with this in mind it would save a lot of heartache. The debugger would have to scan and debug the software code and "abuse" the running program to check for weaknesses.
The world needs a cold, intelligent browser plugin capable of ruthlessly terminating the lives of those that get in the way of "the mission". In this case, it could kill off the people who design awful web pages, all while singing a lovely little song!
Only through hard work and perseverence can one truly suffer.
You're too late. The program already patented itself. You now are required to leave that computer running constantly for 20 years.
This isn't to imply that such social spaces should be public - having a gradation of public-private and well-known, unknown spaces is more multidimensional, more what we really do socially. And it's probably not a MUD - playing social roles with rules isn't for everyone or every activity. And it's not a dating service. And not just a toy, and probably not based on avatars - minimal gimicks, maximal versatility. An advance that allows us to 'just connect' better, like the perfect bartender, maybe without the hangover.
"with their freedom lost all virtue lose" - Milton
I hate people who say that. When humans can't even moderate decently (metamoderation) we are supposed to let an AI do it? I read at -1 because I don't want people to decide for me what is good or bad.
Geez, why don't we let a 2 year old hit a button to decide.
That's inflammatory sounding.
"I disapprove of what you say, but I will defend to the death your right to say it" - F. Voltaire.
Below is what I initially wrote, but the more I think about it, having my computer us it's cam to see it's environemt and react to it would be really useful. Anyway, here is my original message: I have a webcam on my site and I'll be adding more and there are two things I would like to do with it. 1. I want to archive pictures, but only if they change. When I'm gone, the pictures change only as the sun moves overhead, so most pictures I don't need. But if someone walks in front of the webcam, I want to archive that. 2. I want to know who walks in front of the web cam. If I'm on cam, I want my users to know that I'm on cam. I'm sure all the other cam site operators would think this was cool, anacam, planetconcrete, jennicam. I'm right in the middle of updating my site and scripting most of the daily tasks I need done with python, but I can't really script archiving the cam images because I would have too look at them. I really don't want to do that. Right now it's 1440 images a day. If I add 3 more cames, that's over 5000 images per day at 1 minute refreshes. Way to much both to save or look at. The other thing that ties into this is that I paint pictures, and I often would like to take elements out of my pictures, but not have to manually select cut and paste. I would like to drop a bunch of pictures into the app, tell the app what the template image is, and let it find pictures with that kind of image inside, and let it cut out the matching image and save it in a new file. I think something like this would be neat for a variety of other reasons...imagine finding more shots of President Clinton and Monica? But seriously, it would be really neat for me to archive all my digital camera pictures and keep a database of which ones have trees, or my kids, or naked women, or whatever...or find images with certain colors etc. That way I could search my digital camera images with templates, or through a database after they have already been checked for template images, and I could do the same thing with my paintings. As an example, I took about a 1000 images on a recent trip to London (my first, I wanna go back) and I would love to know which ones contain images of my grandfather, or the westminister clock tower, or an underground sign. (you can find my london pictures here: ) Okay, that's what I think might be useful. Especially for me. -calvin
Hmm..lets see..
A Search Engine to serve up content to any device, whether its a Desktop or a Wireless Device.. oops.. Google been there, done that.
An intelligent utility sitting on the server to serve content matching the user preferences, through intelligence acquired by watching what the User bought, selected, and chose over the past months - Oops again...Vignette, Broadvision are already in that problem space..
Build an Evil Empire over two decades, write millions of lines of code and deliver half of the functionality promised, crash twice a week and still manage to eat up half of its competitors.. Oops again.. M$ has already done that too.
Build one of the best desktop UI for Linux to make it more easy and accessible to John Doe... Andy Hertzfeld did that too
Build a thin protocol for transferring data on the wire,making seamless integration of components possible over a distributed networked architecture using XML - SOAP(Simple Object Access Protocol) is already doing that..
Create a good online teaching utility which is intelligent enough to understand the topic that its supposed to teach, based on the complexity grades assigned to each topic, grade the student over a period, serve content to him based on his preferences, watch how much time is spent on each topic and based on that, decide to serve up more topic of related interest or switch to something else. All in all, make the online learning experience easy, intelligent and which requires least effort from the user, and make him feel that theres a real person on the other side, watching his every move, grading him etc..etc.. Now that could be something..
I hate sigs..
www.hackorama.com
Rapid Nirvana
This is just the kernel of an idea here so I am not sure how it would be implemented, but perhaps some sort of agent which would help the open source world find who is using various pieces of code in their software. This would allow for better policing of the GPL. It might constantly search any downloadable software packages it can find for files with a certain name or if the source code is offered perhaps the code itself to create a list of suspects for the author(s) to look into.
Another idea might be an agent that allows you to track all references to you, your website, e-mail address, personal info, etc. This would help people interested in preserving their privacy.
I would like to have an expert system that classifies and archives email for me. It should automatically detect what email are private, what email I can safely delete, what is spam, and also present it in an easy to browse fashion.
I've always found it disappointing that systems which try to profile customer preferences aren't smart enough to understand that people can like the same thing for completely different reasons. A smarter system should be able to model the motivations and intentions of consumers to better match them to products and services. It would need to be able to store partial information, which may not make sense initially, but which could provide meaning after sufficient accumulation.
I think consumers would be very willing to answer questions like, "Why did you buy this product" or, "Click the attributes you like/dislike about this product." Most people who browse the internet are often actually looking for in-depth product information. In fact, the ideal way to collect this information would be to interact intelligently with the user when they are using a search engine, trying to find a specific piece of information. It would be great if AI software could help them find what they are looking for and be able to suggest truely similar alternatives.
This may seem nefarious, but I don't think advertising would be an intrusion if it were driven by true interests of individuals rather than the sales goals of marketing execs.
"What I cannot create, I do not understand."
try wine.com
Blar.