Slashdot Mirror
Executive Secretary In Every Computer
An anonymous reader writes "BusinessWeek Online just ran an interview with a researcher from Sandia National labs whose team has developed an alternative approach to artificial intelligence. They have come up with a software program that models a computer user's behavior and gives the user advice, corrects his errors or saves files according to the user's own logic. The idea is for computers to learn how to use with users -- instead of vice versa. The software has already been tested with air traffic controllers."
Comment removed based on user account deletion
What happens when the user is a sick, twisted and sadistic person. Will the computer adapt to that kind of user?
Not exactly comforting, if you ask me! I expect air traffic controllers to know their systems and how to use them. What happens when this software has learned to compensate for one traffic controller's particular errors, and then suddenly another traffic controller takes over his/her station?
zWhat would an EWOULDBLOCK block, if an EWOULDBLOCK could block would? -- me
Indeed, this sounds exactly like Clippy. I read an article on Clippy a few years ago. Clippy was a great idea, that was supposed to help in just these ways. During R&D it worked very well.
Then MS marketing got involved. They decided that Clippy didnt get activated enough. Clippy in its research version might have popped up once a month when a user really needed help. However, once a month would not justify the expense of development and marketing, nor could it be hailed as a great new feature if the users almost never saw it.
Enter the new and marketing improved Clippy any MS office user over the last decade has had the misfortune to experience. Junk the I part of AI, and just make an annoying paperclip instead of a helpful tool. I can only imagine how the researchers felt about having their nice idea turned into something like what Clippy got to be.
Maybe we'll see a real implementation of this kind of technology at some point in time. But I'll bet any commercial application of this is more likely to get written by popup ad companies, and jog the ATC guys elbow by suggesting which airline he should be using or something...
Remember oliver, the electronic personality extender predicted by Alvin Toffler in "Future Shock"
There's an interesting passage about olivers in John Brunner's excellent novel, "The Shockwave Rider":
"... so-called olivers, electronic alter-egos designed to save the owner the strain of worrying about all his person-to-person contacts. A sort of twenty-first-century counterpart to the ancient Roman nomenclator, who discreetly whispered data into the ear of the emperor and endowed him with the reputation of a phenomenal memory." (pp. 41-42)
-kgj
The next killer app, in my opinion, is the application that allows you to not only save content, but also the context (or contexts, even - human beings don't keep things in their head under one strict association - there are multiple pointers to the same information) behind that word doc, picture, etc.
I would love to be able to quickly find items that I need that were saved years ago. Almost every day I have to find such things on my disk, and having a searchable interface (particularly for binary encoded files, such as executable or graphics files - which have little searchable text inside of them) that works would save hours every week.
Instead of only having a limited amount of information, filename and directory, you would be able to search over multiple hierarchies as well as descriptive text - even for binaries. This would put the user in the driver's seat, allowing her to build relationships within the data that have meaning to her.
Lodragan Draoidh
The more you explain it, the more I don't understand it. - Mark Twain
I want my computers to present me with clear and unambiguous output. In return, I will give them as much unambiguous input needed to get the job done. Save the "clever" AI for Doom 3 and let me get back to work.
Ita erat quando hic adveni.
I remember reading a while ago about the comparison between the computer-learns-human style of doing things vs. human-learns-computer.
The examples I believe were the current Palm OS with its logical if somewhat odd "grafitti" system. It was compared to the old Newtons which attempted to learn the user's handwriting, as well as the new tablet pc's.
Basically the long and short of it was that the order of % correctness went newton > tablet > palm. Although the tablet pc's do a pretty good job interpreting, they still "make mistakes" when someone's writing gets really sloppy. On the other hand after a minimum of time the average user can use graffiti with a high level of accuracy and can understand the malformations of a sigil that might produce an error while being made.
All in all though it seems most of these attempts to "learn" what a user may do are misplaced. I try to keep my "websites" directory very well organized, as well as my "print work" directory, but both vary in structure from each other, even before my own mistakes and idiosyncratic files. And my applications directory is a completely different story... and lets not even get started on consumer media. Shouldn't this all be handled by XML soon anyway?
We've still got the world's best massively parellel computers in our noggins. Pattern recognition OWNZ.
You could also direct it by voice command. I had this program back in the day, heady stuff at the time.
Here's a pile of other stuff on Software Assistants.
May be someone can write a module for Alice integration with Slashdot. I think the dialog-based parsing engine would work just great after some tweaking.
/. spirit).
An alternative approach would be to first parse Slashdot archives to get a lot of posts, articles and moderation data and then use Bayesian theory to decide which sentences/keywords should be included to produce highest moderation based on the words in the blurb (or the linked article, but parsing that would be against
It can be further enhanced using the poetry evolution engine. If we limit the system to very short posts (cliche jokes or smartass oneliners), it might work quite well (feedback, of course, would be the moderation).
Any volunteers?
Future Wiki -- If you don't think about the future, you cannot have one.