Scientific American article on Computer Chess

[syzygy]

I think that by definition the human brain cannot fully understand how it works.

If the human brain can understand how it works then if you tell me to choose a number(1 or 2) than I can have an algorithm to calculate exactly if I choose 1 or 2.

Not at all. I can fully understand how a random generator based on atmospheric noise functions without being able to predict its result on any particular run.

Uri seems to be equating fully understanding to being able to predict the outcome of a random number generator. For instance, to explain the atmospheric noise RNGs, we should be able to memorize the state of the atmosphere, solve the navier-stokes etc.. all in our tiny brain. I believe fully understanding should not be reduced to this ridiculous level and that understanding the general mechanism (e.g. "choosing 1 or 2 in 50-50 manner with random noise added to it", should be enough of understanding, atleast for me.

Exactly, I just want to understand enough of the working of the human brain to be able to program a "mechanical" brain that I can feed all kinds of sensory inputs to make it learn stuff about the world. That would in no way enable me to predict whether Uri will choose 1 or 2, but that is not where my interest lies anyway.

[Frank Brenner]

It is easy to give a rough definition about what "AI" is:

"I" is something that regulary wins the Turing-Test against computer science students of department "AI" who already have their bacchelor
There is no time limit for the students for the talk (**)


There is an A before I when that entity is a computer.



(**) i think this are sufficient conditions

The most easy german Entity that can sometimes win the Turing Test is the following short piece of code

Begin
Output "Hey Alte, was geht ??"
End.


This is slang in germany from some years ago and means "Good Day Madame. How are you?" Many random people from the street (from street menas without preparation) would immediately stop the test and decide for "Human"

The Turing Test is worthless when it is played against random ppl from the street.


What happens for a chess engine or Mathematica ?

If the Human says "Hello How Are you?", the chess engine or mathematica answers with "parse Error".

So, either the Chess Engine nor mathematica is AI.

They are nothing more than something like math, or discrete math, or algebra or logic.



However, the above definition of "I" is only a rough definition. Dogs are in fact little intelligent, but the Turing Test does not allow dogs to win this test regulary.

[mclane]

Wrong or right moves are not found by a search tree building a main line.
That is how computers do it.
It it is wrong.

It is not AI. It is kind of brute force although it is highly intelligent.

It is AI, and it is not at all intelligent.

And what you call "forgetting" is in reality "relegating to your subconscious". Letting your neurons do it without your conscious thoughts "understanding" it. Driving a car or letting good moves "well up" in your brain is not done by forgetting but by training your neurons.

Yes. But that's how humans learn something. If computers want to learn something, they need to forget the unimportant things by creating consciousness and subconscious.

Without this, AI is not possible. Everything we learn is been done by forgetting about it. So as long as computers cannot put something into subconscious and filter out important and unimportant things, they cannot become intelligent.

The programs that came the closest to this were the very selective chess programs. Because these programs did not follow everything.
Mephisto III , sci sys mark V and later cxg Sphinx 40/50.

Another thing we should not forget is that we play chess by looking onto the board with eyes. We SEE patterns. Chess programs do not watch , they see The Whole board. Again this is the same mechanism. By NOT seeing the whole board but only some trajectories, geometrical patterns and chess words , we filter unimportant things out.

In everything we do, we have the same mechanism of filtering.
This is done by our 2 brain halves. Our left brain halve is trying to learn by heart , but this way you cannot really become a master. Learning by heart is the first step. Driving like a beginner . The important step is to sleep/forget about it by throwing all into the right brain halve. The right brain halve has a different logic. While the left brain follows aristotelic logic the right brain halve mainly uses associative logic.

As long as computers and chess programs cannot reproduce this, they are not AI but stupid. They play chess like beginners or drive cars like beginners. Nothing they do can reach perfection. They have no intuition. They do not forget. They do not filter unimportant things out.

What Komodo and stockfish are doing looks like chess, but it is how beginners play chess. It is not AI. It is learning by heart. Aristotelic logic.

[Frank Brenner]

It doesn't understand what it's reading.
.

Science today understands how the brain works at micro and macro levels.

Understanding how the brain works is not a big step forward, it is only something like "Hey i found a new kind of turing machine"

A big step forward would be to extract the software stored in a brain.

After this just compile it to any computer language you like and you have a piece of code that you can call "AI"

[Lyudmil Tsvetkov]

what could an electrical circuit understand, no matter how complicated?

Uh? The human brain is an electrical circuit. Conduction is by ions rather than electrons, but that just makes it a bit slower.

it is an electrical circuit immediately connected to the body, and able to assess events as positive or negative. able to feel pain and joy.
machines are not able to do that.

it is certainly not possible to separate egoism from pure logic, what concerns the human brain.

sorry, no time to go into detail now, but read some interesting input in the thread, so will push along a few thoughts:

@ Harm: computers can hold significantly more bits in their memory than the human brain. - no, it is the other way around, synapses are just a gross buidling element of the brain; there are finer, sub-material particles, some of which still not discovered by science, and others in the process of development in the human brain. future generations will have a lot of them.

@ Uri: humans can not understand how their brain works. - of course they can, everything, absoluetly everything in this world could be understood, no matter how long it takes. but, as perfect knowledge is circular in nature, once a very advanced psychological entity understands everything, attains perfect knowledge, it will lose its love. Because it will have to do one last big effort to attain the very last part of knowledge. Big efforts mean losing one's love. So, a human who achieves this will immediately want to go bacl to the non-perfect, real world with less knowledge, but still be in possession of precious love.

Perfect knowledge is attainable of course, but for this, one should be able to see the whole circle, and, as necessarily we are part of the circle at every given moment in time, seeing the whole circle is only possible when one leaves it entirely to take an objective wholistic out-of-the-world view. when you do this, you get separated from life, as life is possible only within the circle.

@ other people: sorry, I already forgot what you have said...

finally, to finish another useless thread post, there is certainly no AI at all currently, and not possible to create one ever in the future. no artificial creation will ever know what life/meaning/understanding is. there will always be part of a, even though very elaborated, human design, unable to independently take decisions, even slightly differing from the designer's goal. (unless you understand computer crashes as such)

but the use of ai as a afigure of speech, synonymous with advanced technical/artistic desing, is quite normal and fully acceptable.

[hgm]

it is an electrical circuit immediately connected to the body, and able to assess events as positive or negative. able to feel pain and joy.
machines are not able to do that.

Says who? The 'Tsvetkov Oracle'?

@ Harm: computers can hold significantly more bits in their memory than the human brain. - no, it is the other way around, synapses are just a gross buidling element of the brain; there are finer, sub-material particles, some of which still not discovered by science, and others in the process of development in the human brain. future generations will have a lot of them.

Not discovered yet, but you know about them...

[syzygy]

it is an electrical circuit immediately connected to the body, and able to assess events as positive or negative. able to feel pain and joy.
machines are not able to do that.

Says who? The 'Tsvetkov Oracle'?

Certainly the body is nothing special here. If we succeed in connecting a human brain to a mechanical body (or just enough to let the brain communicate with the world) then surely we still have "real intelligence".

And is there anything special about the brain that could not be simulated artificially? That seems extremely unlikely.

[Uri Blass]

Wrong or right moves are not found by a search tree building a main line.
That is how computers do it.
It it is wrong.

It is not AI. It is kind of brute force although it is highly intelligent.

It is AI, and it is not at all intelligent.

And what you call "forgetting" is in reality "relegating to your subconscious". Letting your neurons do it without your conscious thoughts "understanding" it. Driving a car or letting good moves "well up" in your brain is not done by forgetting but by training your neurons.

Yes. But that's how humans learn something. If computers want to learn something, they need to forget the unimportant things by creating consciousness and subconscious.

Without this, AI is not possible. Everything we learn is been done by forgetting about it. So as long as computers cannot put something into subconscious and filter out important and unimportant things, they cannot become intelligent.

The programs that came the closest to this were the very selective chess programs. Because these programs did not follow everything.
Mephisto III , sci sys mark V and later cxg Sphinx 40/50.

Another thing we should not forget is that we play chess by looking onto the board with eyes. We SEE patterns. Chess programs do not watch , they see The Whole board. Again this is the same mechanism. By NOT seeing the whole board but only some trajectories, geometrical patterns and chess words , we filter unimportant things out.

In everything we do, we have the same mechanism of filtering.
This is done by our 2 brain halves. Our left brain halve is trying to learn by heart , but this way you cannot really become a master. Learning by heart is the first step. Driving like a beginner . The important step is to sleep/forget about it by throwing all into the right brain halve. The right brain halve has a different logic. While the left brain follows aristotelic logic the right brain halve mainly uses associative logic.

As long as computers and chess programs cannot reproduce this, they are not AI but stupid. They play chess like beginners or drive cars like beginners. Nothing they do can reach perfection. They have no intuition. They do not forget. They do not filter unimportant things out.

What Komodo and stockfish are doing looks like chess, but it is how beginners play chess. It is not AI. It is learning by heart. Aristotelic logic.

I disagree.
From some experience of teaching children who are weak players
I can say that one of the differences between beginners and experienced chess players is clearly speed of ability to make a brute force search.

I can prove it by the following exercise that I invented.
put a knight in the board and ask a beginner and an experienced player to show me all the possible move of the knight.

Experienced players can show me immediately all the legal moves and the only limitation is the speed of their hand when
beginners who only learned the rules usually need to think where the knight can go.

Same is also for the queen and when I asked a weak player(11 years old) to put pawns at all the longest moves that the queen can make to all directions I can see that she did not see immediately the right square and when I put a queen at d4 she looked for one second at g8 and probably understood that it is wrong square because g8 is a different color than d4
and I do not think that her intelligence is below average.

A strong player does not need to think and immediately put a pawn at h8 and the same for the rest of the directions.


If you need to think to answer the question where your pieces can go that is practically doing 1 ply search then you miss a basic knowledge in chess and cannot be a good chess player.

One of the conditions to be a good chess player is this type of knowledge.
Of course you need more knowledge to become a strong chess player but basically the difference between a strong chess player and a beginner is knowledge and beginners often cannot even do one ply search by brute force and will simply miss part of the moves.

Not that experienced players do one ply search in games but at least they can do it with no mistakes if they want to do it.

[syzygy]

If you need to think to answer the question where your pieces can go that is practically doing 1 ply search then you miss a basic knowledge in chess and cannot be a good chess player.

Calling it "missing knowledge" seems prone to cause misunderstandings here.

Your pupil obviously KNEW the rule. Otherwise she would not even have figured it out by thinking about what the answer should be.

Throsten will say that she had not yet "forgotten" the rule. But forgetting is obviously not the right word here.

Your pupil's problem is that the sets of reachable squares do not (yet) come "automatically" to her. With a bit of practice that will improve. That is not about forgetting but about internalising the patterns.

In the case of reachable squares, a computer does not need to go through such a process: it is easy for a programmer to code it directly and efficiently.

[Lyudmil Tsvetkov]

it is an electrical circuit immediately connected to the body, and able to assess events as positive or negative. able to feel pain and joy.
machines are not able to do that.

Says who? The 'Tsvetkov Oracle'?

@ Harm: computers can hold significantly more bits in their memory than the human brain. - no, it is the other way around, synapses are just a gross buidling element of the brain; there are finer, sub-material particles, some of which still not discovered by science, and others in the process of development in the human brain. future generations will have a lot of them.

Not discovered yet, but you know about them...

someone should mediate from Sirius down to earth...