Alpha Zero the AI defeating the chess world champion, we have already covered the human-AI chess tournament, which took place about 20 years ago. It’s this ending. We bring you a conversation between Lex Friedman, a young AI professor from MIT, who believes AI will be developed as part of humanity. Not just a business tool And another one is Gary Kasparov, world champion chess. People who compete in chess against AI.
Kasparov said the incident was not only the first AI victory but also ended his long-standing unbeaten record. And when having time to consciously for 20 years and look back at the past This has given many ideas.
Alpha Zero ,The AI advantage
Many people tend to understand that chess is a representative of intelligence. And always thought that the smarter would win. Which this is not true Because indeed “Even the best chess programs in the world today, such as Stockfish and Komodo, are not doing the best math. It has to calculate 10 to the 46 possibilities that even the world’s fastest computers today cannot do at this level. “
“But what AI chess does is Calculations for walking with the least mistakes. So it can be said that chess is a Loser Game – who misses more than loses. Especially competing on the pressure of the time limit. Even the World Champion couldn’t deny this feeling. It is this pressure that humans make the wrong move. Also AI is able to keep taking this small advantage over and over and eventually lead to victory, ”Kasparov said.
Winning in chess is not a smart situation.
Kasparov also said that AI is able to overcome the game of chess. Considered to work on a fixed factor or Closed System, which is whether the rules are different, there is no variation. And without any unexpected factors Therefore, it is difficult to use this game to gauge the true intelligence of humans. Because the real world is the Open System that is full of changes. No certainty, no rules
What AI can do is still not comparable to human capabilities. Especially when it comes to interpreting and understanding language. For example, we often encounter new language research that often claims AI has similar capabilities to humans. But now AI can only work in context. In short, give an example of the best AI in 2019 about reading a long article and be able to condense it. Or answer questions Is still considered to be very limited
AI will only answer direct questions such as who, what, where, but if it is an analytical question, it cannot be answered. Or may not answer as well as they should, for example, if reading a murder novel And we hope that AI will analyze who is the killer, AI will not be able to.
In addition, Kasparov considers that current machine learning is a direct learning process, which cannot be prioritized. And still unable to “Ask the right question” or analyze questions and things that should be learned by yourself.
In short, almost all chess AIs are just Type-A systems. Meaning models that try to consider every route and make the least mistakes. But there is only one system where Kasparov is likely to be the starting point for the true future of AI is Deep Mind’s Alpha Zero.
Alpha Zero is an AI that approaches intelligence.
Alpha Zero is able to learn new patterns that go beyond computer calculations. Which was caused by repeated trials of millions of boards Which of these things is truly an approach to wisdom.
Kasparov calls it “Machine produces knowledge,” and Lex says the ability is like a human “intuition” or “intuition”.
However, Alpha Zero still has some point of improvement, and Kasparov was invited to discuss and analyze Alpha Zero’s weaknesses with the research team. They found that there were some pieces that Alpha Zero made mistakes and that Alpha Zero could learn this mistake. Must pass a losing number on this piece of a hundred thousand boards Which is too much for learning to mistake.
And all of this is the story of a 20-year retrospective a historic event. The AI defeated the chess world champion. It can be seen that Kasparov is also capable of chess. He also has knowledge and understanding of Computer Science as well. And from that day hold today. When the outcome is ignored or won. We will find that today AI is constantly evolving and has penetrated into the way of life of people in all dimensions. Although not perfect yet, AI is a technology that has helped to advance humanity’s new history. And we must admit that the digital world cannot be true without the AI that fills the innovation gap.