It seems as though the game of checkers is no longer a mystery. Computer scientists in Canada have announced that the their computer program, Chinook, can do no worse than draw against any competitor. The article from iTWire reports that the researchers have haled this as a “milestone in the history of artificial intelligence research”. A major claim considering the algorithm was basically tasked to review all possible moves in the game regardless of the starting position.
Is this really then artificial intelligence, or more a case of computing power over the power of the human mind to retain multiple scenarios. I mean, the computer isn’t actually thinking for itself here; it is effectively just replaying game after game that it has already stored in its memory. A simple Yes/No logic tree is what this is, no thinking or reasoning involved.
Where’s the challenge in that? I understand the effort involved from the researchers, and wholeheartedly congratulate them for the achievement, but I fail to see how this advances the free-thinking ability of machines over man.
{ 4 } Comments
Bill said…
I mean, the computer isn’t actually thinking for itself here;
Don’t get too semantic about it. You have to understand the definition or cognitive process of Learning and Thinking before you leaped into a misguided comment.
it is effectively just replaying game after game that it has already stored in its memory.
Yes, but that is exactly how human thinks, don’t you think? The program also learns and adapts to new situation, that was not foreseen in advance and again this is characteristic of how human thinks. The misconception about AI is that it can solve anything that a human does. No, it can’t, but it definitely solves things much faster than even a human would take to achieve in a reasonable amount of time. A sub-branch of AI, known as Machine Learning is a study of how to mimic human thinking by implementing learning algorithms that computers could execute. Yes , they do learn, if you understand the definition of learning. Human basically, mainly learns cognitively by 2 ways. Inductively (learning from experience) and Deductively (learning by thought process), and now there are tons of algorithms that could do either learning by Induction & Deduction or both.
A simple Yes/No logic tree is what this is, no thinking or reasoning involved.
Again, this is exactly how human thinks, but the thinking or reasoning is done via inductive learning or deductive learning (see above).
I myself don’t use the word AI or avoiding using it at all, which is one of my specialist area. I prefer to use the word Computational Intelligence or Machine Intelligence. Most AI Researchers (where I am not one - just a practitioner) don’t or hardly use the term AI at all. When I attend an AI workshop or conference, you talked with researchers at coffee break, there is never mention of the term AI. It is always the subject of Computational Intelligence or Machine Intelligence that is being discussed. AI is a preferred term, that journalists and media have liked to report news about Computational Intelligence or Machine Intelligence since it sounds convincing (draw the crowd to buy papers) and also it is easy to comprehend. If the media use either Computational Intelligence or Machine Intelligence in their report, then definitely, no one would know what they mean.
So, just understand that these computer programs learn and how far off their learning capability in comparison to a human, is another question, but they do definitely learn according to the proper cognitive definition of learning.
Hi Falafulu,
Thanks for your comments. Isn’t semantics what this is all about? Calling this AI (as the article did - not me) is stretching the common meaning.
Now, I didn’t mention anything about learning in my post, which would definitely be a feat for a machine to master. You’re obviously more passionate about the whole AI (Comp Int, Mach Int) field than I am; but ‘thinking’ (the product of experience) is what you commonly expect an AI to do.
I just don’t see how this achievement adds to the whole Machine Intelligence field. Once the program has reached all of the possible scenarios it needs to win or draw a game, where’s it’s purpose now?
Bill said…
where’s it’s purpose now?
The knowledge of solving of such algorithm leads to application in other areas. For example, a program to diagnose medical conditions of patients (which are currently available at clinics & hospitals), run a complex control process of a nuclear power plant (currently in use) or do check up of the launch processes of a Shuttle aircraft before take off. So, it isn’t because they the game developer has reached the ultimate, it is the research about the algorithm that will benefit other applications in other domains, and that is exactly what research is all about. Not about what it solves at present but about the knowledge that was gained from it and also its potential to apply in other area.
Take computers for example. They were built solely for nuclear physicists & scientists to use in simulations in the early days. They (pioneers) never knew that after over half a millenium, computers would be found in everything necessary to run a modern society.
My meaning was - what is the purpose of that particular application? I fully understand the implications for use in other areas of science, and for that I congratulate those responsible
Post a Comment