AI development?
-
I have just spent aprox 2 weeks surfing the web trying to come to grips with what the current AI trends are, and a suitable area to start studying in. What I discovered was that AI has branched out into a multitude of different fields and almost every Institution is going of on ther own tangent. Eg If I study Chemistry here in NZ it is very similar to any similar course in Europe or the USA. But with AI there are still no axioms and lots of differing thoughts. There is an interesting dev kit called Visual Prolog and there is a lot of Java source code about, not so much in C++. The main are I was looking at was Natural Language Recognition and even that is very diverse. Regardz Colin J Davies
Sonork ID 100.9197:Colin
More about me :-)
Your findings are basically the same as mine. There does not appear to be any clear direction. It seems that the 20+ years of research have not lead to signifigant standardization of concept or approach. It is a difficult problem to solve, but surely more progress has been made than it appears. All the AI research/development work I have seen so far has been highly specialized into one particular aspect of how it will be used. ie. games, natural language, military applications, etc.
-
Dabble is the correct word. I bought a book a while ago but I have yet to find one that doesn't completely lose me after the first 3 chapters or so. As far as market analysis goes, they use NN for prediction (believe it or not) and pattern recognition. I'm sure they have other uses, but these are the two I'm interested in. I've seen a demonstration of a NN that predicted the S&P500 pretty damn close after churning through a literal ton of data. I don't expect to get very far with this, but I'm sure I'll learn a lot implementing it. Mostly they work at pattern recognition I think, which is what the majority of technical analysis is anyways. There are certain patterns that historically have indicated certain moves in the market. By building and training a NN, you can assist and automate the recognition of these patterns and hopefully make gobs of cash. That's the plan at least. I have to cut my lawn first though... J "I am wise enough to therefore not spout my ill informed opinion as if it were remotely related to fact." - Christian Graus
Jamie Hale wrote: I've seen a demonstration of a NN that predicted the S&P500 pretty damn close after churning through a literal ton of data I wonder how much "manual pre-training" was needed to achieve this. As I stated in the thread below (to Colin Davis), it seems that all the AI work being done is highly specialized for its particular use. I have not seen anything that is just generic intelligence.
-
Your findings are basically the same as mine. There does not appear to be any clear direction. It seems that the 20+ years of research have not lead to signifigant standardization of concept or approach. It is a difficult problem to solve, but surely more progress has been made than it appears. All the AI research/development work I have seen so far has been highly specialized into one particular aspect of how it will be used. ie. games, natural language, military applications, etc.
Matt Gullett wrote: All the AI research/development work I have seen so far has been highly specialized into one particular aspect of how it will be used. ie. games, natural language, military applications, etc. i think it has to be. since we have such a limited understanding of natural (human) intelligence in general, the best we can do now is to limit AI to very narrow fields, where there is a limited number of factors to consider. -c
ABSURDITY: A statement or belief manifestly inconsistent with one's own opinion.
-
Yep. Diagnostic expert systems, and more recently, intelligent agents. The agent stuff is experimental, the expert system stuff is for a product I'm building. /ravi "There is always one more bug..." http://www.ravib.com ravib@ravib.com
That's very interesting. What has your expeience been with the results acheived by employing AI? What form of AI are you using? Did you develop it yourself or use someone elses work as a base for getting started? What kind of agents?
-
Matt Gullett wrote: All the AI research/development work I have seen so far has been highly specialized into one particular aspect of how it will be used. ie. games, natural language, military applications, etc. i think it has to be. since we have such a limited understanding of natural (human) intelligence in general, the best we can do now is to limit AI to very narrow fields, where there is a limited number of factors to consider. -c
ABSURDITY: A statement or belief manifestly inconsistent with one's own opinion.
Speaking of "limited number of factors"... How many variables do you think is reasonable for a customized AI system with current technology? From what I have seen it appears that many of the AI projects being developed are force-fed a solution along with the data to find the solution. Chris Losinger wrote: i think it has to be. since we have such a limited understanding of natural (human) intelligence in general No argument from me. Our understanding is apparently extremely limited. The complexities of human intelligence seem to be beyond simple mathematical machines at least for a while.
-
That's very interesting. What has your expeience been with the results acheived by employing AI? What form of AI are you using? Did you develop it yourself or use someone elses work as a base for getting started? What kind of agents?
Matt Gullett wrote: What has your expeience been with the results acheived by employing AI? I'm probably dating myself here :), but from 1987-94 I worked with John McDermott, et. al. on XCON (a forward chaining expert system) at DEC's AI Center in Mass. XCON (for "eXpert CONfigurator") was used to configure the company's flagship line of VAX computers and saved the company a lot of time and money by being able to guarantee delivery of properly configured hardware to customers in a much shorter timeframe than ever before. We basically canned the knowledge of many "ace" hardware designers into a rule base that eventually grew to about 15K rules. XCON was written entirely in OPS5, with small portions written in Lisp. My interests have since shifted to diagnostic (backward chaining) expert systems. I published an article in Australian PC World (April 1987) that explained diagnostic expert systems in layman's terms. I can send you a photocopy if you're interested. I'm currently working on a diagnostic expert system framework in the hope of making it easier for non-computer-literate domain experts to use backward chaining reasoning to solve problems. The agent stuff I'm doing is pretty basic. You can some of my early efforts here. I'm currently developing agents to selectively retrieve news and financial information. I hope to meld the 2 technologies together, whereby the diagnostic engine uses agents to glean facts, and is simultaneously used by these very agents to determine how to locate and identify information. Another big AI area is robotics (of which I know nothing). Most labor-intensive and user-hostile manufacturing environments are almost completely robot controlled. Without machine vision and robotics, we wouldn't be able to sell cars and chips at the price we do. /ravi "There is always one more bug..." http://www.ravib.com ravib@ravib.com
-
Wow! You're experience seems quite extensive. Ravi Bhavnani wrote: I published an article in Australian PC World (April 1987) that explained diagnostic expert systems in layman's terms. I can send you a photocopy if you're interested. I would love to have a copy of this. It sounds like an excellent read. I really appreciate the feedback as I am contemplating spending some time researching AI and your feedback has helped me better understand some of the areas of interest.
I'd be happy to mail you a copy. You can send me your snail mail address to ravib@ravib.com. /ravi "There is always one more bug..." http://www.ravib.com ravib@ravib.com
-
Matt Gullett wrote: What has your expeience been with the results acheived by employing AI? I'm probably dating myself here :), but from 1987-94 I worked with John McDermott, et. al. on XCON (a forward chaining expert system) at DEC's AI Center in Mass. XCON (for "eXpert CONfigurator") was used to configure the company's flagship line of VAX computers and saved the company a lot of time and money by being able to guarantee delivery of properly configured hardware to customers in a much shorter timeframe than ever before. We basically canned the knowledge of many "ace" hardware designers into a rule base that eventually grew to about 15K rules. XCON was written entirely in OPS5, with small portions written in Lisp. My interests have since shifted to diagnostic (backward chaining) expert systems. I published an article in Australian PC World (April 1987) that explained diagnostic expert systems in layman's terms. I can send you a photocopy if you're interested. I'm currently working on a diagnostic expert system framework in the hope of making it easier for non-computer-literate domain experts to use backward chaining reasoning to solve problems. The agent stuff I'm doing is pretty basic. You can some of my early efforts here. I'm currently developing agents to selectively retrieve news and financial information. I hope to meld the 2 technologies together, whereby the diagnostic engine uses agents to glean facts, and is simultaneously used by these very agents to determine how to locate and identify information. Another big AI area is robotics (of which I know nothing). Most labor-intensive and user-hostile manufacturing environments are almost completely robot controlled. Without machine vision and robotics, we wouldn't be able to sell cars and chips at the price we do. /ravi "There is always one more bug..." http://www.ravib.com ravib@ravib.com
Wow! You're experience seems quite extensive. Ravi Bhavnani wrote: I published an article in Australian PC World (April 1987) that explained diagnostic expert systems in layman's terms. I can send you a photocopy if you're interested. I would love to have a copy of this. It sounds like an excellent read. I really appreciate the feedback as I am contemplating spending some time researching AI and your feedback has helped me better understand some of the areas of interest.
-
Is anyone out there doing any side projects or serious work on articficial intelligence? If so, what aspect of AI are you working on? Is it for commercial business type uses, games, etc?
Some years ago I write a "life simulator". I built a world and small creatures that fight, eat, reproduce, get older and die. They was divided into families that figths for the food. But was only a game. Sometimes I think to rewrite it and use some more sophisticated rules (I was thinking at Dangeons&Dragons...):-)) bye - SONORK ID 100.17421 : b0nu$ -
-
Grammar checker! Todd Smith
Is this something you have already written or are working on? Or, is it something you would like to see?
-
Stock market analysis...mmmm... that's interesting. To what extent do you see the AI performing this analysis? I have been spending some of my spare time lately trying to complete (?) my spell checking control. It's kind of a hobby so I use it as a base for experimenting with other things. Anyway, I was wondering how I could apply AI to the spell checker. I have been reading alot about it, but I haven't seen very much talk about it on CP. Good to hear that someone else is dabbling in it.
Grammar checker! Todd Smith
-
Is anyone out there doing any side projects or serious work on articficial intelligence? If so, what aspect of AI are you working on? Is it for commercial business type uses, games, etc?
AI has been a pet side project of mine for a long time now. I incorporated a bit of intelligence into something I worked on at the last place I worked at. The system I designed processes files from various sources - the files can be in binary (big and little endian), and various flavors of text. I have an agent that watches a part of the file system for changes and figures out what type of file changed. It was really difficult to figure out the text part of the recognition since the text files contain similar information and formatting. The system does not learn but is cool in that all you need to do is register a COM object that exposes a certain interface and the my system can figure out what types of files the COM object can work with and routes only the appropriate files to it - it is a cool design. I am a member of a couple of SIGs, subscribe to some publications, and buy AI-related books by the pound. A functioning application that uses fuzzy logic, or learns, or is capable of being trained has been rather elusive for me so far. One of my side projects involves analyzing the results of a search to figure out what a user is searching for. For example, if you search for "hockey" using Google, my intelligent app should return "hockey" based on its analysis of Google's results. While it may not seem useful at first, I think it forms a great basis for summarizing the contents of a document - that plugs into another idea I have, which I cannot comment on right now :suss: Enough about me - what are you working on? Essam - Author, JScript .NET Programming
...and a bunch of articles around the Web -
AI has been a pet side project of mine for a long time now. I incorporated a bit of intelligence into something I worked on at the last place I worked at. The system I designed processes files from various sources - the files can be in binary (big and little endian), and various flavors of text. I have an agent that watches a part of the file system for changes and figures out what type of file changed. It was really difficult to figure out the text part of the recognition since the text files contain similar information and formatting. The system does not learn but is cool in that all you need to do is register a COM object that exposes a certain interface and the my system can figure out what types of files the COM object can work with and routes only the appropriate files to it - it is a cool design. I am a member of a couple of SIGs, subscribe to some publications, and buy AI-related books by the pound. A functioning application that uses fuzzy logic, or learns, or is capable of being trained has been rather elusive for me so far. One of my side projects involves analyzing the results of a search to figure out what a user is searching for. For example, if you search for "hockey" using Google, my intelligent app should return "hockey" based on its analysis of Google's results. While it may not seem useful at first, I think it forms a great basis for summarizing the contents of a document - that plugs into another idea I have, which I cannot comment on right now :suss: Enough about me - what are you working on? Essam - Author, JScript .NET Programming
...and a bunch of articles around the WebEssam Ahmed wrote: system I designed processes files from various sources... Sounds cool. Essam Ahmed wrote: A functioning application that uses fuzzy logic, or learns, or is capable of being trained has been rather elusive for me so far. Same here. However, my current work project could easily benefit from some AI/fuzzy logic. Essam Ahmed wrote: Enough about me - what are you working on? My endevours are pretty mundane. I developed a spell checker a couple of years ago (it's here on CP) and I have used it several times and expanded on it since I published it. It's a pet project of mine I use to give me an excuse to play with various technologies. Right now I am reviewing AI (fuzzy logic too) and considering trying to incorportate some AI in my spell checker. Also, at work I have been developing an object persisitence and logical representation system. (I think it's pretty cool.) One of the long term goals of this project would be to do some analysis of data sets and extract patterns, etc. I am planning on using AI (neural networks) for this when I get the go ahead.
-
Is this something you have already written or are working on? Or, is it something you would like to see?
It's something that could be added to the ieSpellChecker Todd Smith