Comments on: Texai - The Year in Review 2007 http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/ An open source project to create artificial intelligence Sat, 30 Aug 2008 12:39:19 +0000 http://wordpress.org/?v=2.1.2 By: Jean-Paul http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-957 Jean-Paul Tue, 01 Jan 2008 05:53:14 +0000 http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-957 well-done - Good to see someone taken a different approach and not just talking about it but getting down to coding. You're a motivating example. Keep up the good work in 2008! PS Why not add also ConceptNet to your KB? well-done - Good to see someone taken a different approach and not just talking about it but getting down to coding. You’re a motivating example. Keep up the good work in 2008!
PS Why not add also ConceptNet to your KB?

]]> By: Steve Reed http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-960 Steve Reed Tue, 01 Jan 2008 11:37:23 +0000 http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-960 I looked at <a href="http://web.media.mit.edu/~hugo/conceptnet/" rel="nofollow">ConceptNet</a> three years ago when I was at Cycorp. The first issue is that it is not truely open source - if you are a commercial user there is a commercial license involved. The second issue is the quality of the KB. When I looked at it there was a substantial amount of joke or sarcastic assertions that would entail a manual filtering before combining with the much higher quality OpenCyc, WordNet and even Wiktionary content already in the Texai KB. And the final issue is the lack of precision in the assertions because the KB was gathered from filled-in web forms that were not subject to semantic well-formedness checks. Once the Texai bootstrap dialog system is operational, it should overcome these objections that I have with ConceptNet. -Steve I looked at ConceptNet three years ago when I was at Cycorp. The first issue is that it is not truely open source - if you are a commercial user there is a commercial license involved. The second issue is the quality of the KB. When I looked at it there was a substantial amount of joke or sarcastic assertions that would entail a manual filtering before combining with the much higher quality OpenCyc, WordNet and even Wiktionary content already in the Texai KB. And the final issue is the lack of precision in the assertions because the KB was gathered from filled-in web forms that were not subject to semantic well-formedness checks.

Once the Texai bootstrap dialog system is operational, it should overcome these objections that I have with ConceptNet.

-Steve

]]> By: Arthur T. Murray http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-993 Arthur T. Murray Fri, 04 Jan 2008 17:29:41 +0000 http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-993 The "Year in Review 2007" is an interesting read. I was struck by how much this project tries to gobble up enormous corpora of lexical and conceptual information. In my own Mind.Forth project, I start out with a minimal innate knowledge base (KB) and let users add simple Subject-Verb-Object (SVO) factoids. Then the AI Mind tries to maintain meandering chains of NLP-generated thought about the contents of the KB, and also in response to user input. Most recently (in 2008) I am trying to get the AI Mind to back out of generating thoughts where insufficient information is known in the KB -- specificially, where a subject noun is proposed but no verb is available to complete a statement. Then the AI is supposed to go by default to a question asking the human user for more information. If the human user makes no response, the AI is supposed to continue thinking -- either self-referentially because of an Ego module, or in a traversal of the KB (not yet coded :-). In sum, it is interesting to see such an ambitious AI project and the approach being taken -- apparently an approach of organizing massive, Cyc-like data, whereas my own approach is to _grow_ the artificial mind from a starting point of minimal knowledge, adding more mental functionality as the mindcore is debugged and permits tack-on functionality. Otherwise -- Happy New Year 2008! -Arthur T. Murray The “Year in Review 2007″ is an interesting read. I was struck by how much this project tries to gobble up enormous corpora of lexical and conceptual information. In my own Mind.Forth project, I start out with a minimal innate knowledge base (KB) and let users add simple Subject-Verb-Object (SVO) factoids. Then the AI Mind tries to maintain meandering chains of NLP-generated thought about the contents of the KB, and also in response to user input. Most recently (in 2008) I am trying to get the AI Mind to back out of generating thoughts where insufficient information is known in the KB — specificially, where a subject noun is proposed but no verb is available to complete a statement. Then the AI is supposed to go by default to a question asking the human user for more information. If the human user makes no response, the AI is supposed to continue thinking — either self-referentially because of an Ego module, or in a traversal of the KB (not yet coded :-). In sum, it is interesting to see such an ambitious AI project and the approach being taken — apparently an approach of organizing massive, Cyc-like data, whereas my own approach is to _grow_ the artificial mind from a starting point of minimal knowledge, adding more mental functionality as the mindcore is debugged and permits tack-on functionality. Otherwise — Happy New Year 2008! -Arthur T. Murray

]]> By: Joe Simone http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-1046 Joe Simone Thu, 10 Jan 2008 21:54:14 +0000 http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-1046 Hi Steven, I just happened to stumble on this web site! Last I heard you were at Cyc corp? I guess you are now striking out on your own. I have been following Doug Lenat on and off for about 15 years - since around 1992. I used to think that Cyc would eventually hit pay dirt. I am not so sure anymore. OpenCyc seems to be a bust as well as it is so poorly supported and lacks the vibrancy of an active community. I look at the posts on source forge and very rarely do questions get answered. Both the OpenCyc and Cyc web sites seem not to be updated frequently if at all. Overall I see a lack of progress and stagnation - but that's because I can only see from the outside and don't know what going on inside cyc land. Hopefully Texai will fair better. Have you looked at using/incorporating the OpenMind KB? I participate in OpenMind and it's enjoyable and amusing sometimes to teach new knowledge. I've played the Cyc FACTory game and it is pitiful. Oh well. If you have the time, would you be so kind as to give a brief summary of the current state of affairs in pursuit of a common sense AI? If you do not wish to post here, you can just email me directly. Is Cyc destined for failure? Are we even close to bootstrapping AI such that it can learn from dialog? Best regards for a great and productive 2008! Joe Hi Steven,

I just happened to stumble on this web site!

Last I heard you were at Cyc corp? I guess you are now striking out on your own. I have been following Doug Lenat on and off for about 15 years - since around 1992. I used to think that Cyc would eventually hit pay dirt. I am not so sure anymore. OpenCyc seems to be a bust as well as it is so poorly supported and lacks the vibrancy of an active community. I look at the posts on source forge and very rarely do questions get answered. Both the OpenCyc and Cyc web sites seem not to be updated frequently if at all. Overall I see a lack of progress and stagnation - but that’s because I can only see from the outside and don’t know what going on inside cyc land. Hopefully Texai will fair better.

Have you looked at using/incorporating the OpenMind KB? I participate in OpenMind and it’s enjoyable and amusing sometimes to teach new knowledge. I’ve played the Cyc FACTory game and it is pitiful. Oh well.

If you have the time, would you be so kind as to give a brief summary of the current state of affairs in pursuit of a common sense AI? If you do not wish to post here, you can just email me directly. Is Cyc destined for failure? Are we even close to bootstrapping AI such that it can learn from dialog?

Best regards for a great and productive 2008!
Joe

]]> By: Steve Reed http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-1060 Steve Reed Fri, 11 Jan 2008 15:47:13 +0000 http://texai.org/blog/2008/01/01/texai-the-year-in-review-2007/#comment-1060 Hi Joe, After working at Cycorp seven years, I was let go in August 2006 during a massive downsizing due to cuts from a major US government sponsor. I appealed to my wife and our investments for support so that I could pursue my own AI research. Perhaps as you know, John DeOliveira and I were the instigators of OpenCyc. After we left, and also as a result of the reduced funding, OpenCyc has indeed stagnated. In the meantime Johd D founded the Cyc Foundation and its mission is to augment OpenCyc via publically accessible tools. I attend their monthly face-to-face meetings in Austin. In contrast to Cycorp, which is a for-profit company that must protect its intellectual capital, my project is no-profit and open source from scratch. I plan to create artificial intelligence by writing, with least effort, a bootstrap English dialog system that intelligently acquires knowledge and skils, and then recruiting a multitude of volunteers to continue the work. Because OpenMind is related to ConceptNet, please see my above comment. I will write the post you suggest immediately. -Steve -Steve Hi Joe,
After working at Cycorp seven years, I was let go in August 2006 during a massive downsizing due to cuts from a major US government sponsor. I appealed to my wife and our investments for support so that I could pursue my own AI research.

Perhaps as you know, John DeOliveira and I were the instigators of OpenCyc. After we left, and also as a result of the reduced funding, OpenCyc has indeed stagnated. In the meantime Johd D founded the Cyc Foundation and its mission is to augment OpenCyc via publically accessible tools. I attend their monthly face-to-face meetings in Austin.

In contrast to Cycorp, which is a for-profit company that must protect its intellectual capital, my project is no-profit and open source from scratch. I plan to create artificial intelligence by writing, with least effort, a bootstrap English dialog system that intelligently acquires knowledge and skils, and then recruiting a multitude of volunteers to continue the work.

Because OpenMind is related to ConceptNet, please see my above comment.

I will write the post you suggest immediately.
-Steve
-Steve

]]>