Marcus Hutter’s AIXI. . ..  I’m guessing Hans will be vehemently against this approach.

Marcus is a “what it does matters” type of AI guy, not a “what people call it”.

This is like my approach, but I think this approach is too much extreme the other way.

I said it before, and I’ll say it again, “the proof of the pudding is in the eating”, and although people like Hans are cocksure they’re 100% correct, and rest of us are all wrong, we’ll see… Maybe Marcus Hutter will give us TRUE AI !!!

http://vimeo.com/14888930

Either you don’t understand what I’m doing, or you didn’t understand the lecture your linking to; there is very little correlation between his work and what I’m working on.

A few hints for you:

- His approach is to build an intelligent agent that optimizes towards a specific task (based on a universal starting model). He is NOT talking about replicating human behavior. (Note: I don’t get why he talks about ‘general AI’ while his approach is clearly towards ‘expert systems’, as his examples do show without a doubt).

- He is totally discarding any psychological issues, therefor not taking things into account like consciousness, feelings and how these interact with a frame of reference and therefor influence the whole model. The simplification he proposes is great for expert systems but not for general AI (or ‘strong AI’, a term I like much better).

- His part on ‘prediction’ is not based on ‘analogies in experiences’ but instead on statistical pattern matching. While this works great for specific problems (stock market predictions), it is totally unusable in real world problems where the complexity of environmental influences rule out any empirical foundation for a statistical approach.

Victor Shulist - Mar 12, 2011:

...although people like Hans are cocksure they’re 100% correct, and rest of us are all wrong, we’ll see…

Actually, it is you who demonstrate in your replies that you are pretty sure you are right, whereas I’m constantly researching the validity of my model. My model is still very much in flux and adapting to new insights. However, your stance is perfectly understandable as you are in the implementation stage, meaning that changing your views would be ‘going back to the drawing board’, so you vigorously defend the route you’ve chosen (others here do that too).

It will be quite a while before I start implementing my model in software (apart from prototyping some things for testing certain assumptions), as I wont implement before I’m convinced I tackled all the important details of the model. I believe in the engineering approach, ‘refactoring’ is the devil’s spawn.

The video made me feel as if I were back at school.

Jan Bogaerts - Mar 12, 2011:

The video made me feel as if I were back at school.

The video was pretty intense wasn’t it ?  I think Mr. Hutter’s concepts are probably the best for reaching A.G.I. that I have seen so far !

He focuses on the bottom line, the results; what it does, and doesn’t waste time with nebulous terms.  The level of details in his explanations is awesome; no words like ‘concepts’ for example.  I also very much enjoyed this video !

Victor Shulist - Mar 12, 2011:

no words like ‘concepts’ for example.

... and yet we all know that humans perceive their reality as interlinked concepts

By the way, he might not have been using the word ‘concept’ but his talk mentioned many things that ARE actually concepts. Besides that, the view about ‘conceptual perception’ is not something I dreamed up, it is a pretty accepted model throughout the strong-AI research world.

From David J. Chalmers, Robert M. French, Douglas R. Hofstadter:

High-level perception, on the other hand, involves taking a more global view of this information, extracting meaning from the raw material by accessing concepts, and making sense of situations at a conceptual level. This ranges from the recognition of objects to the grasping of abstract relations, and on to understanding entire situations as coherent wholes.

http://consc.net/papers/highlevel.pdf

Hans Peter Willems - Mar 12, 2011:

However, your stance is perfectly understandable as you are in the implementation stage, meaning that changing your views would be ‘going back to the drawing board’, so you vigorously defend the route you’ve chosen (others here do that too).

Yes, and I and others are making progress, not spinning our wheels.    We don’t need to defend our approaches, I see the output generated by members on this site that prove their approaches for themselves   But of course, no matter it does, since it doesn’t use Hans Peter model of reality , you can safely ignore it.

Hans Peter Willems - Mar 12, 2011:

Victor Shulist - Mar 12, 2011:

no words like ‘concepts’ for example.

... and yet we all know that humans perceive their reality as interlinked concepts

By the way, he might not have been using the word ‘concept’ but his talk mentioned many things that ARE actually concepts. Besides that, the view about ‘conceptual perception’ is not something I dreamed up, it is a pretty accepted model throughout the strong-AI research world.

From David J. Chalmers, Robert M. French, Douglas R. Hofstadter:

High-level perception, on the other hand, involves taking a more global view of this information, extracting meaning from the raw material by accessing concepts, and making sense of situations at a conceptual level. This ranges from the recognition of objects to the grasping of abstract relations, and on to understanding entire situations as coherent wholes.

http://consc.net/papers/highlevel.pdf

Well of course, if you know how to place the word ‘concepts’ in context like that , one knows what you mean.  These people know how to write clear documentation.

What’s with all the aggression? Do we need an entire thread with the original post dedicated to slamming one member of the forum?? Victor, thank you for sharing the work of Marcus Hutter but this post could have been handled much better.

We all make strong statements about our ideas for the way a chatbot/AI project should function and I think that’s fine. If you disagree, you should be able to say so. (I certainly have.) But to start a whole thread with the theme of being contrary just sets up a fight.

Agreed.

No unpleasantness

Victor Shulist - Mar 12, 2011:

But of course, no matter it does, since it doesn’t use Hans Peter model of reality , you can safely ignore it.

Victor, I don’t understand what your problem is here. I’m posting my views, discussing my ideas and point to certain research if the discussion warrants it. I never said that I have the ultimate solution. I do however voice my opinion on things like NLP as the core for intelligence. I am free to question your approach, as you are free to question mine. I tend to reply to your inquiries with information. You start discussions seemingly only to slam my views, and while I keep coming back with reasonable replies in the vein of keeping up a civilized discussion, you keep posting things like above.

Anyone is free to voice their ideas and views on a forum. Those views might oppose yours; deal with it. Engage in a civilized discussion, bring arguments to the table to defend your stance, whatever, but refrain from acting like an a*hole.

And when I post some link to research information that validates my use of the idea of ‘concepts’, you come back with:

Victor Shulist - Mar 12, 2011:

Well of course, if you know how to place the word ‘concepts’ in context like that , one knows what you mean.  These people know how to write clear documentation.

No reaction to the information at all, or to the point about ‘concepts’, just more negativity.

All in all, I’m pretty much done with you at this point.

Gee, have a nice day.

There is a lot to say about reading, and learning before challenging you peers.
Thanks to everyone.

peetee le trickfox

Raymond! It’s good to see you again! I hope everything is well with you. I would love to hear your thoughts about this and the other discussions under way. Feel free to chime in, if you wish.

Thank you Dave
I’m in the hospital right now, recouping from open heart surgery and a couple of related after effects, but tomorrow I’ll try and catch-up and I’ll post a few thoughts in a day or so.
Like I said I’ve been reading a lot about what everyone else i saying, and I’ve been learning a whole lot from just listening to all of you.
You cannot imagine how much I appreciate this place.
It’s a source for a lot of valuable wisdom in AI programming.
Perhaps I could help but let me find my “nitch” first!

peetee le Trickfox
Aka
Raymond

Absolutely, Raymond! By all means, take your time, and regain your health/strength! I’m praying/thinking good thoughts for your full and fast recovery, and I’m glad that you find the forums here an enjoyable place to visit. Get well, my friend.

I’m in the hospital right now, recouping from open heart surgery and a couple of related after effects, but tomorrow I’ll try and catch-up and I’ll post a few thoughts in a day or so.

I hope you feel better soon.