Disregarding the sensationalistic title of the BBC Earth video on youtube, I was very impressed by the work of scientists from The Primate Research Institute, Kyoto University, described here Chimp vs Human! | Memory Test | BBC Earth .
First thing that came to my mind was the question how did they manage to do it? Note that a chimpanzee does not have to actually understand the meaning of each particular symbol with respect to integer quantity it represents, in order to perform this task, but it has to understand exactly and precisely the order of symbols. If I were to teach a chimpanzee the order of arabic numerals, by using a computer with a touch screen and a reward deliverer interface, my program would first display the numerals in a fully ordered fashion that is easily followed even by humans, which are surprisingly inferior to chimpanzees in this exercise. For example, in a horizontal line from left to right in a strictly ascending order, if touched sequentially it would deliver them a banana. Then I would display it in a diagonal line from left bottom to right top of the screen to show that the slope of the line is irrelevant, finishing with vertical line from bottom to top. Then I would start to reverse the order from ascending to descending but still keeping it strict, to show that orientation of the line is irrelevant. Then, I would display it very slightly disordered, by changing the place of just 8 and 9, all that in order for them to reinforce the new knowledge that they acquired about what the actual order is, and not to lose the thread and motivation by not being able to accomplish the task of restoring that order on demand every time.
Next step, before scattering the numbers in a random fashion across the screen which is a target goal of the exercise, would be to lose linear alignment but still preserve some degree of order with respect to such alternative alignment. For example zig zag from left to right and mixing a bit more the order of numbers. That step might even follow immediately after the initiation, before making any changes in the ordering, just to show that linear alignment is not necessary to present the ordered sequence. Each step can be repeated multiple times before going on with more challenging tasks, that already require that they are certain about the order of numbers, not to mention that not all numbers should be introduced at once.
The point is that I can easily imagine the way how can such a task of training chimpanzees to understand the order of numbers and demonstrate that understanding on demand, be accomplished systematically by breaking it into simpler steps, that are feasible regardless of the communication barrier between humans and chimpanzees. Actually, if it is important to an explainer that certain information is passed from the explainer to the explainee, it is a sign of intelligence of the explainer to overcome that barrier. That applies both to training animals and programming computers to do things we want them to do. And describing such a complex relation between species with a single variable, such as communication barrier, and reproductive barrier, sounds very statistic.
After these initial contemplations, and considering the fact that chimpanzees showed superiority in comparison with humans both in speed and accuracy, when performing the task of remembering positions of numeric symbols (invented by humans) on a screen and demonstrating their memory, I started to wonder how would I teach them to play chess, applying the same logic of breaking the task down into feasible simpler steps, trying not to lose their motivation, offering sufficient incentive for their progress along the way, and why has noone ever succeeded such a thing? I believe strongly it is not because they are incapable of learning that game, I think it is all about overcoming communication barrier, hence it is actually more a task that tests human intelligence, rather than chimp’s.
I would start with a simplest possible KQK ending, by showing them that if I play white I would always win a banana, and if they play white and they silly allow me to capture their queen, there will be no banana for them. I believe they would grasp the concept of watching out for their queen where they place it fairly quickly, and soon they would be able to utilize their pieces to checkmate me. OK, that leap is maybe not that simple, but it can be simplified by starting the training with positions in which the black king is already at the edge of the board, constrained by the white queen that is placed on the penultimate line, and the white king is already near to support the checkmate blow, in a couple of right moves. That should teach them both not to allow their queen to be taken, and not to allow the enemy king to escape from the trap. Eventually, after they acquire the concept of checkmate, and how the queen and king move, and how the queen can be captured by the enemy king, and how to avoid stalemate, I would move on slightly, and introduce other pieces, for example in KRK endgame. The rook not only cannot move and check diagonally, it also can be captured by allowing the enemy king to approach it diagonally, which is again a new concept to learn in comparison with KQK endgame, where the queen can be captured only if placed (by the side that owns that piece) in contact with the enemy king, without being supported by her own king, so it requires a lot less attention not to lose a queen (when the other side owns only a king), in comparison with the rook.
So, I can envisage a lot of difficulties that need to be overcome in order to accomplish even the modest initial subgoals, let alone the most ambitious one, the main task of learning them how to play the whole game, but if they are enabled to play certain setup sufficient amount of times, and kept motivated in their progress, they can learn any game as well as human can, maybe even better, who knows?
I would certainly combine the possibilities of them playing one another, enlarging their community exposed to that game making it potentially part of their culture, playing against human, playing against computer, of course always strictly via touch screen and a computer program that does not allow them to make illegal moves, and automatically rewards the winning side, whatever seems to be the most effective way to motivate them to learn. Even making it the only way for them to get food, when their enthusiasm for the game drops a bit.
As for why noone succeeded this, and what are potential benefits of such accomplishments, I have two competing theories. One is that it requires much more patience and effort than that what is already invested in similar projects, without a clear plan on how to exploit the eventual successful result, not to mention the side effect of offending a lot of anthropocentric people, and those who are chess professionals, as a strong deterrent to such an endeavour. Proving that it is possible to program a computer to play chess and defeat the world champion was a huge blow to the ego of that champion, proving that it is possible to train an ape to do the same thing, might be even worse. I mean, exactly the same thing might be a bit a stretch of the imagination, but it would be interesting to know how far apes can go in a game that requires that much planning, focus, visualization, etc.
The other theory is that my lack of anthropocentrism leads me to a fantasy world that is simply not realistic, no matter how much is invested in its realization, that people already tried and failed miserably with much simpler strategic games, and are afraid to report their effort in order to avoid attacks by the mentioned anthropocentric people which may conclude that it was a waste of resources, that the problem is not primarily in an interspecies communication barrier, but in some deficiencies of chimpanzee’s minds, that does not allow the planet of the apes scenario to become true.
Only time can prove which theory is right, I hope not by escalation of the war in Ukraine. However, before it escalates to that extent that planet of the apes becomes a likely outcome, let us utilize the latest achievements of the existing human civilization, and let us see what Chat GPT thinks about it. I asked it the same questions I posed to my self, plus I will ask it to revise my essay as soon as I refresh it a bit so that it becomes available for me again, at the moment it gives me "An error occurred. If this issue persists please contact us through our help center at help.openai.com."
Since Chat GPT is capable only to process the existing knowledge found on the internet, although it does it well, it is no wonder it cannot give an unconditional yes-no answer to the question if chimpanzees can be trained to play chess, since noone actually knows this at the moment, and its answer is a reasonable everyone's guess (mine included). Insisting to determine the ethics of the revised experiment also does not lead anywhere, because, there is no definitive answer to that question, it is a matter of opinion, as well as the question if it is ethical to keep pets at home. It may be legal, but if it is ethical or not, it is up to anyone’s decision. I think it is not, therefore I do not keep any:
Comments
Post a Comment