Slashdot Mirror

From https://aaai.org/ in the description of next year's conference:

AAAI-18 welcomes submissions reporting research that advances artificial intelligence, broadly conceived. The conference scope includes all subareas of AI and machine learning.

Now, if you think you are such an expert in the field to say that the Association for the Advancement of Artificial Intelligence, which was founded in 1979 as an academic association, is wrong about the definition of artificial intelligence, I'd like to hear what contributions to the field you made that can back up the idea. If you did none, then just let the scientists working in the field define what AI means and contains, and accept it.

I'm fed up with those recurrent comments that statistical learning is not AI. Please base your opinion on evidences and not on your feelings.

For instance, go to the page of the major conference on AI: http://www.aaai.org/Conference...
Look at the topics, they include machine learning. There you have it, the professionals in AI consider ML to be AI.

We can now safely have that useless comment die and never appear again.

Artificial Intelligence is a field of study, just like Algebra.

The way ACM defines it [1], AI includes computer vision, NLP, planning, and knowledge representation and reasoning. All these tasks are today based on computational statistic/machine learning. You can also look at the last AAAI conference [2].

The way it is defined, your GPS is an AI system.

[1] http://dl.acm.org/ccs_flat.cfm
[2] http://www.aaai.org/Conference...

We now have game AI that's really good at tactics but not so good at strategy (chess AIs), and we have game AI that's really good at strategy but not so good at tactics (AlphaGo with its failure to spot tesuji). The next step would be to make game AI that's good at both. See e.g. On Adversarial Search Spaces and Sampling-Based Planning. The next step after that? I'd say incorporating the kind of strategic capabilities AlphaGo shows to make AIs for very large incomplete information games.

Just saw a talk about the Narcotweet project. The interesting part about Narcotweet is that it's documenting the emergence of a new kind of "journalism:" the "tweet curator" who aggregates local social media reporting. These people are routinely followed by bigger news media (CNN en Espanol) yet maintain extremely strong ties to the people witnessing these things first-hand. The power of this entire project is that it's a way of getting information from places where the conventional news sources have decided it's too risky / too expensive to send *actual* reporters.

“We found that six degrees actually overstates the number of links between typical pairs of users: While 99.6% of all pairs of users are connected by paths with 5 degrees (6 hops), 92% are connected by only four degrees (5 hops),” the Facebook Data team said.
...
The average distance between all people on the site in 2008 was 5.28 degrees, while now [Nov 2012] it is 4.74.

Twitter

Our optimal algorithm finds an average degree of separation of 3.43 between two random Twitter users,

From the first link I got to this page which has the delicious quote:

"we might want to stop thinking about jobs as the main aspect of our lives that we want to save. They may be a means, but they are not the ends."

outsourcing our drug use (and sex, apparently)

It's actually sex, drugs, and rock&roll.

a robot challenge that will test robots' vision and language understanding.
the robots/sobots must be able to recognise objects automatically and perform tasks like: get the "star trek" poster or get the blue dry erase marker. the final event will be held at the twenty-second AAAI conference on artificial intelligence in vancouver, canada july 22-26 '07 [taken from ofpblog]

some more follow up stuff:

Also see: Googling Your TV - Prototype software from Google Research could listen to your TV and send back useful information -- and ads of course. By Wade Roush. Technology Review (August 24, 2006). "A system recently outlined by researchers at Google amounts to personalized TV without the fancy set-top equipment required by previous (and failed) attempts at interactive television. Their prototype software, detailed in a conference presentation in Europe last June, uses a computer's built-in microphone to listen to the sounds in a room. It then filters each five-second snippet of sound to pick out audio from a TV, reduces the snippet to a digital 'fingerprint,' searches an Internet server for a matching fingerprint from a pre-recorded show, and, if it finds a match, displays ads, chat rooms, or other information related to that snippet on the user's computer. ... When word of the research first appeared in the media, some bloggers and other technology watchers reacted with horror; many assumed that the background conversation picked up by the microphone in Google's system would be uploaded to Google. But the technology makes it impractical; at four bytes, the fingerprints don't contain enough information to reconstruct the original sounds in a room. 'Some people did get the impression that we had an open microphone that was going to listen in on them,' says [Peter] Norvig. 'Clearly, that was not what we were doing. We are transmitting a key that can be matched but not reversed. That said, users are giving up some information -- and that's something they have to decide about.'"

from
http://www.aaai.org/aitopics/assets/AIalerts/curre nt.html

see also http://www.technologyreview.com/read_article.aspx? id=17354&ch=infotech

>I guess that lawsuits based on ordinary language would be a disaster

...for much the same reason that software written in natural language can have difficulties.

Documents that describe how something should work out and the reasons for it, whether in the legal or the engineering realms, necessarily require technical jargon and precise structure, if they are to have predictable results. The legal "programming language suffers the grave disadvantage of having been crafted over centuries by thousands of people. Some of them were dickering in court, who were often interested in dealing with their particular case, and others were working in legislatures, who are often interested in something else entirely. The result is a language with the clarity of Assembler and the efficiency of COBOL.

All this effort, and the results may still not be substantively just, but after all engineers too can have difficulty making clear specs conform to what the customer wants. What can ya do?

P.S. your "pro se/prose" observation was delightful!

That was part of my point, and why I was high-lighting specific passages while referring to it as the only real content of the press release. This kind of market-droid crap detracts from the legitimate study and research of AI.

Plus, to your point, I'm not a big fan of the SEC, constantly portrayed (similar at times to the rep of the Fed) as some kind of glorious governmental overseer of all things laissez-faire, when the truth is a little murkier.

&laz;

CP is actually a vibrant, active area of artificial intelligence research with a large, strong community. If you go through the AAAI conference proceedings (available at http://www.aaai.org/), you'll see that, indeed, it had a number of papers this years. Additionally, they have their own conferences and researchers with strong interests in the area, including my research advisors and several of my colleagues.

Machine learning, also, has come a long way. Popular methods at the moment are support vector machines, and bayesian (graphical) models. You'll also see meta-learning taking hold, in the form of bagging, boosting, and ensemble techniques.

Building on this, you'll see systems that do a bit of both. Check out MaxSAT as a problem. You can easily see where researchers are driving toward systems that place probabilities in the weights of a MaxSAT solver, and then solve based on this. I attended a talk yesterday on this by Dan Roth. Very cool stuff.

+5 ahead of your time.

The coolest part of slashdot is that they log all your comments. The negativity I get is funny because all inventors and developers get it when posing something revolutionary. But I'm fortunate enough that I can point people back and go,"Just look at how most people didn't understand the concept."

Anyway I'm writing a scientific paper for: An ai conferance

Its amazing that theres an AI conferance in a city only 45 min away from where I live!

AARON is generally viewed within AI (along with the rest of this topic) as more an issue of creativity. That is (to quote Margaret Boden) : the generation of novel and useful artifacts.

Novel... can indeed mean randomly generated in the simplest case (as per "Typogenerator"), but generally other factors guide such processes in human endeavours (combinatorial, heuristical, transcendental).

Useful... aye, there is the rub. How do judge art?

See here for more.

I've been doing research (non origional) recently on the subject and would suggest looking into Intlligent User Interfaces (IUI). Most of the research in this field is more future oriented but there is always speech recognition software. Of course this technology has a long way to go to reach conversational proficiency. I believe that at some time in the next few decades computers won't have files to open or hard drives to format, but there will be a personal assistant 'living' in the computer. I would suggest reading the Winter 2001 AI Magazine.

With all due respect to Backgammon, Chess and Id, the best game I've ever played is Go. It's easy to learn but takes a lifetime to master. As well, it's deep enough to force a player to actual think all most all of the time. With it's handicap system, even beginners can offer masters a challenging game enjoyable to both.

American Go Association
http://www.usgo.org/index.asp

International Go Federation
http://www.nihonkiin.or.jp/igf/index.htm

The Interactive Way To Go
(excellent tutorial)
http://playgo.to/interactive/

Tips for Learning Go
http://go.kestrel.nu/

Interestingly enough, it remains the one game that cannot be won by brute force number crunching. Even an average player can beat the best Go programs. As such, I conisder it to be a useful tool in the search for meaningingful computing.

Go is a whole new challenge
http://www.aaai.org/AITopics/html/go.html\

If you don't Go, you'll never get anywhere! ;~)

Turing test anyone?

Pathfinder software archive

CLIPS

: A Tool for Building Expert Systems. Maintained by Gary Riley.
fuzzyCLIPS

Some other NASA soft:
COBWEB/3 (ptolemy.arc.nasa.gov) ?
AUTOCLASS AutoClass is an unsupervised Bayesian classification system for independent data.
PRODIGY cs.cmu.edu Integrated Planning and Learning System

What I don't see is how random changes in DNA can eventually create more quality information for new processes.
I mean it's like taking a software package (eg MS Office) and randomly modifying / deleting / and inserting code segments and expecting the code to actually work, let alone work better. Just think about it, could a random process produce a better 2.6 kernel? would you trust it in a production environment?

Yes, a random process could produce a better 2.6 kernel. You're not going to get it with just a single copy mind you. Evolution works on the order of billions or trillions of copies (eg: insects) changing over millions or billions of years. I'd suggest you familiarize yourself with genetic algorithms and genetic programming.

Genetic programming is just what you suggest, actually. Take a population of programs, see how well it performs a task, then randomly change those programs and see which ones perform better. Keep on going through the generations until a good solution emerges. For something like a kernel, that would take an inordinately long amount of time to evolve, but it could be done.

I met David Hanson two years ago at the AAAI conference in Edmonton, Canada. He hung out with our robotics team for a couple days during the conference where he was demonstrating his (really freaky) robot heads and we were competing in the robot host competition. He's a very artistic guy, and about as enthusiastic as they come. I'm glad to see he's starting to make it big.

Funny thing is, the Ray Kurzweil (who was also at the conference) quote in the article sounds like a conversation I had with David. Our robot, built to serve hors d'oeuvres in a coctail party environment, was designed to look like a table, rather than a butler (Although it had a pan/tilt/zoom camera for a "head"). The idea was to improve on people's expectations of a table rather than disappoint people expecting a real human. Kurzweil's quote sounds like something I probably said to David: "Better to build a smart piece of furniture than a stupid human."

the American Association for AI

The way that AI is headed, I think that there may be robots that will be able to to menial work, e.g. cook, clean, build, etc. I don't think that AI will become an actual intelligence for some time.

An artifical intelligence is able to invent new things is something that may take much longer.

Although, I think that some researchers have realize what it is that made humans want to leard; Physical and Emotional needs. For all we know one of these needs driven robots could develop a more advanced intelligence than humans.

• Kismet
• Meet Kismet...
• Robot Learning Driven by Emotions
• AGI Project Phase-Overview
• Emotional and Intelligent: The Tangled Knot of Cognition

You may want to read this book and see it yourself whether data mining would make a breakthrough in the future.

This sort of stuff goes on at the conference as well. The conference featured a Robot Rescue event, where robots were sent in to save hypothetical victims from a rescue scene. Unfortunately, none of the robots made it into the most difficult zones. aaai link

The value of these competitions and robot demos at AAAI is somewhat controversial, as AAAI is not very keen on actually publishing papers about robots at that conference. The robots get lots of press attention, but the delegates won't hear many robot-related technical papers

At the main international robotics conference the subject is treated a little more seriously.

Actually you can get a set of Adobe Type 1 fonts for LaTeX from the AMS. You can then make TeX use these standard PostScript Type 1 fonts in the PDFs it produces. More details here and here. You can also make LaTeX use the standard PostScript fonts for its body by using the \usepackage{times} directive in the preamble.

Copy of a message I sent the editor :) I can't believe they couldn't predate 1971 for AI (see Sci Fi Word List)

Hi Mike,

Science predates Science Fiction :) Next time I see him [JM], I'll mention it :)
Winton

AI or Artificial Intelligence

Coined by John McCarthy [in a SCIENCE setting, not SCI-FI!], 1956. Seems to be fairly unanimous.... concept goes way back.
" He [JM] invited them to Vermont for "The Dartmouth summer research project on artificial intelligence." (reference)

1956 John McCarthy coined the term "artificial intelligence" as the topic of the Dartmouth Conference, the first conference devoted to the subject. (reference)

RoboCup is only one part of IJCAI. Another interesting event taking place at IJCAI is the AAAI Mobile Robot Competition for urban search and rescue (USAR) robots. They have to navigate three courses developed at the National Institute of Standards and Technology (NIST). The courses have proved extremely difficult for autonomous robots to navigate.

The American Association for Artificial Intelligence AI Magazine Summer 2001 has a paper by John E Laird and Michael van Lent "Human-Level AI's Killer Application - Interactive Computer Games". The title says it all (and I don't want to get in trouble quoting bits of it).

AAAI has (or points to) lots of good introductory material on decision trees, and machine learning in general.

AAAI has (or points to) lots of good introductory material on decision trees, and machine learning in general.

AAAI has (or points to) lots of good introductory material on decision trees, and machine learning in general.

Even this little robot cylinder thing doesn't do more than navigate a path through a virtual field. Nothing most slashdot readers couldn't code in under an hour. What ever happened to breathroughs in AI? Why are the serious researchers leaving it for other fields? Should we forget the dream?

cye is not intended as a general ai solution. i saw the cye base in action at this year's aaai conference, and my understanding that the base was a proof of solution to a traditionally difficult problem in robotics - that of dead-reckoning (estimating your position based only on measurement of previous movement).

dead-reckoning is difficult because carpeted floors, which are the default for most office environments, wreak havoc on traditional wheel-and-shaft-encoder combinations - that is because the fibers in the carpet introduce very tiny amounts of variation into robot's movement, but those variations add up really really fast. the insight in cye is the wheel design, which helps solve that problem (i don't remember the encoders or brains being anything out of the ordinary, though).

btw, regarding "nothing that most slashdotters couldn't code up in an hour" - for dead-reckoning code that may be true. but i wouldn't extrapolate to other problems in robotics. i've seen many highly competent programmers getting their egos crushed by the sheer difficulty of interfacing a machine to the real world. it's a very difficult, and sadly largely underconstrained problem.

r

ps. and regarding the sad state of ai in entertainment - the times, they are a-changin'. check the july issue of game developer magazine (treebased medium, unfortunately) for a cool article on the state of ai in games, which also includes a few notes on the work between the entertainment industry and the academia.

So unfortunately, teaching it to compile your kernel is out.

oh really? :)

don't know whether this applies to the commercial version, but our robotics group at northwestern implemented the scheme language for the dogs, and our pets are now fully programmable. you can even have them crunch equations for you in their spare time. :) (yeah, yeah, i heard all about beowoof clusters. stop that! :)

btw, the dogs make a really cool robotics research platform. if you're in orlando this summer, i'd check out the aaai conference - several universities have had the dogs for a while now, and i hope there will be some interesting applications of the dogs to Real Robotics Research (tm) exhibited at the conference...