AI Ramblings : Comments /blog/ en-gb This site is powered by e107, which is released under the terms of the GNU GPL License. jld@nospam.com (Kevembuangga) jld@nospam.com (Kevembuangga) Sat, 04 Sep 2010 20:50:11 +0200 Sat, 04 Sep 2010 20:50:11 +0200 http://backend.userland.com/rss e107 (http://e107.org) 60 AI Ramblings : Comments http://www.kevembuangga.com/blog/e107_images/button.png /blog/ 88 31 Re: Seeing the forest from the trees (of "open problems") http://www.kevembuangga.com/blog/comment.php?comment.news.6 paper which is available at his home site, nothing appeared of much interest to me.
The idea of iterating the meta-levels until something "stabilizes" is absolutely brilliant but he seem to have been lost in his own inspiration.
I downloaded his latest program Malice CAIA's solver with which I did run the preset demos on an AMD64 but beside the demos its an humongous useless piece of junk.
I emailed Pitrat for more infos and he responded but did not seem to want to share more, only to brag about his results:

Je ne sais si vous avez vous-même essayé de résoudre ces problèmes ou
d'écrire des programmes pour les résoudre. Si oui, cela m'intéresserait
de savoir si en tant que humain vous arrivez à des solutions meilleures
que celles trouvées par CAIA; le logiciel en donne la structure
générale, c'est à dire l'arborescence qu'il développe. De même avez-vous
écrit des programmes plus performants que ceux écrits par CAIA ? Cela
m'intéresserait aussi de le savoir.
Avoir mis en ligne ce logiciel permet donc de juger la qualité des
résultats de l'approche que j'ai entreprise et la valeur de CAIA en tant
que chercheur en IA. La plupart des problèmes que je donne sont
difficiles pour un humain,
etc.. etc...


Is he old fashioned and secretive or confused and lost in his own mess, difficult to say...

[ edited 10 Mar : 17:10 ]]]> noauthor@nospam.com (Kevembuangga) Wed, 10 Mar 2010 17:07:19 +0100 http://www.kevembuangga.com/blog/comment.php?comment.news.6 Re: Objects as epistemological artifacts http://www.kevembuangga.com/blog/comment.php?comment.news.10 nh for recanting and for your fine appraisal
As for relevant papers I don't have much more to suggest, I recently bought Meaning and Grammar but haven't really dug into it.
Some works by Reinhard Diestel on infinite graphs seem to me to possibly bring some insights into knowledge representation problems, because in the end I think the "key" to understanding what AI really is to be IS a matter of representation, somewhere at the boundary between the formal and the informal (maths/logic versus language).
An interesting note by Dick Lipton highlights the deficiencies of current math formalisms and by the same token the power of "informal" natural language: in spite of the informality they are able to meaningfully discuss the question in the blog post.
Actually I haven't done much about AI recently, having been bothered by other personal questions and willing to take a break by doing some more immediately rewarding hacking.

[ edited 10 Mar : 16:40 ]]]> noauthor@nospam.com (Kevembuangga) Wed, 10 Mar 2010 16:37:04 +0100 http://www.kevembuangga.com/blog/comment.php?comment.news.10 Seeing the forest from the trees (of "open problems") http://www.kevembuangga.com/blog/comment.php?comment.news.6 Artificial Beings]]> noauthor@nospam.com (Anonymous) Mon, 08 Mar 2010 19:23:14 +0100 http://www.kevembuangga.com/blog/comment.php?comment.news.6 Re: Objects as epistemological artifacts http://www.kevembuangga.com/blog/comment.php?comment.news.10
The summary you provided strikes me as particularly insightful, so I must ask if you could suggest some relevant papers or areas of study (besides the ones mentioned in your blog post, of course).]]> noauthor@nospam.com (nh) Thu, 04 Mar 2010 10:52:54 +0100 http://www.kevembuangga.com/blog/comment.php?comment.news.10 Re: Objects as epistemological artifacts http://www.kevembuangga.com/blog/comment.php?comment.news.10 Your "point" seems knocking down any and all approaches to AI without discussing their faults in specifics

Have you met my favorite quote about AI:
“If AI has made little obvious progress it may be because we are too busy
trying to produce useful systems before we know how they should work.”
This isn't my own prose, this is from an AI researcher in 1983, nothing new under the sun.

Actually there are no "faults" just delusions, it is not that any bit of existing or previous AI research cannot be of some use it is that by not trying to investigate what could be the core problem a lot of time and resources are wasted efforts with very poor returns.

Part of this due to the hunt for funding, in order to get some budget you have to convince the sponsors and alas the possible sponsors aren't necessarily the most enlightened, this results in "tragicomic" wastes like the Cyc project for which you certainly have seen my criticisms all over the place.

A more serious impediment which I will try to explain is the quest for the definitive silver bullet, there must be a "right way" to do AI, a theory of everything like in physics.
For some it's sequence prediction (AIXI), for others analogy, or quantum logic, etc, etc...
It's both neither and all of these, we should certainly be able to deal with all such matters if AGI is reached, isn't it?
Look at this thread at The n-Category Café, what are they trying to do?
To reach the "Holy Grail" of maths the ultimate "right" foundations.
No chance, yet what they fail to notice is that during this very discussion they are quite able to exchange valid arguments and make some progress despite the lack of formalization.
How do they do that?
The same goes for AI.
That's my point!

while simultaneously failing to suggest a more fruitful area of research


Not so, here is the best summary I made at some blog:
The problematic phase for a computer to do AI isn’t even to recognise an object or concept, a dog or a verb, but to come up with the “idea” that tagging a whole bunch of “inputs” with some unique label is a good way to organize the chaos and build a model of “reality”. As long as the AI researchers will skip this question by doing the conceptualisation themselves instead of leaving it to the computer, they won’t have any useful insights about “intelligence”. Before the computer can go ahead with any “logic”, it has to recognize concepts, and before recognising concepts, it has to discover them; it should not be told that there is a dog or a verb “out there”.

(not to mention repeatedly engaging in ad hominem attacks on AI researchers, calling them all "monkeys" in several comments).


You are mistaken, it's not AI researchers I call monkeys, it's all of mankind.



]]> noauthor@nospam.com (Kevembuangga) Sat, 05 Dec 2009 18:16:00 +0100 http://www.kevembuangga.com/blog/comment.php?comment.news.10 Re: Objects as epistemological artifacts http://www.kevembuangga.com/blog/comment.php?comment.news.10
At this point I'm willing to believe that you are not actually interested in AI, and are merely trolling AI internet discussions. Good luck with that.]]> noauthor@nospam.com (nh) Sat, 05 Dec 2009 12:13:42 +0100 http://www.kevembuangga.com/blog/comment.php?comment.news.10 Re: Objects as epistemological artifacts http://www.kevembuangga.com/blog/comment.php?comment.news.10 I'm afraid I can't add to the discussion, because I'm still trying to tackle the mathematical background necessary to understand Hutter's (and others') work in detail.

Though you "agree" with it you don't seem to understand my point either, so there isn't much we can argue about. ]]> noauthor@nospam.com (Kevembuangga) Thu, 03 Dec 2009 11:18:25 +0100 http://www.kevembuangga.com/blog/comment.php?comment.news.10 Objects as epistemological artifacts http://www.kevembuangga.com/blog/comment.php?comment.news.10
Ultimately, simply dismissing present AI work is worth actually less than the flawed research you dismiss, because you aren't suggesting alternate routes to pursue in actually building a model or theory of intelligence. Your comments essentially amount to "You don't know how intelligence works. Also, I don't know how intelligence works! See ya later!"

I've ran into your comments on three different blogs now, and in every case they've been nothing but distraction. If you actually wanted to dismiss Hutter's AIXI as irrelevant, you would need to discuss, in details, how his research does not and will not provide insight into the "ontologizing" phenomenon. Furthermore, you would need to describe into what a successful research program would focus on. If you cannot do this, your comments amount to a blanket dismissal of a large body of work, which certainly won't convince the people conducting that research, because you haven't given them any idea for an alternate path to pursue.

I'm afraid I can't add to the discussion, because I'm still trying to tackle the mathematical background necessary to understand Hutter's (and others') work in detail. I do think the field of AGI is the best approach we have at the moment, so if you're serious about actually seeing a working AI, please contribute. Otherwise, you might as well not even comment.]]> noauthor@nospam.com (nh) Wed, 02 Dec 2009 08:34:18 +0100 http://www.kevembuangga.com/blog/comment.php?comment.news.10 Almost right... http://www.kevembuangga.com/blog/comment.php?comment.news.9
Have you read Wolfram's A New Kind of Science? Among other things, he discusses what properties of theorems seem to be associated with our judgments of "being interesting" or "usefulness" in the context of logical operations.]]> noauthor@nospam.com (Caledonian) Sat, 25 Apr 2009 18:26:15 +0200 http://www.kevembuangga.com/blog/comment.php?comment.news.9