The stupidest chess AI teaches us

AI Basics


When IBM’s Deep Blue defeated Garry Kasparov for the first time in 1997, the world chess champion accused the company of cheating. He thought the computer could not have beaten him without direct assistance from skilled human players. But now the situation has completely turned upside down. When the grandmaster realizes that today he has received some amazing moves, he accuses his opponent of using the computer. The only worthwhile competition for the top chess engine is competition with each other. The program has become too powerful. mankind lost.

But as the machines attempt to perfect their chess, one bot stands in their way, refusing to accept their dominant position in the robot-human hierarchy. That holdout is Martin, the worst computer opponent on Chess.com, by far the most popular chess website in the world. While programs such as ChatGPT surprise, confuse, and scare users with their increasing computational power, Martin was programmed to be bad at chess. Surrounded by ambitious generative AI products, often confidently returning inaccurate or inconsistent responses, Martin is the rare humble bot who understands and accepts its serious limitations. He lost it when he was given 31 queens against an army of puny pawns.

Martin can certainly beat a novice, but if you know even the basics of chess strategy, he’s no threat. Despite his contemporaries’ AI being designed with the goal of achieving human-like “general artificial intelligence”, this is a machine without grand ambitions. If this is his age of AI chaos, Martin reminds us that intelligent programs can surprise, delight, and even teach within a well-defined framework controlled by the user.

Martin is part of a decades-long lineage of anthropomorphic computer chess opponents with customizable strengths and unique personalities. chess master 2000was a computer game released in 1986 that pitted the user against a gray-bearded man whose Gandalf-esque features graced the game’s cover. In 2019, Chess.com released a bot containing individual names, illustrated avatars, nationalities, and sayings. Martin is a turtleneck-wearing Bulgarian with bushy eyebrows, a thick beard, and a slightly receding hairline. Each game begins with the declaration, “My 4 year old son of his just beat me. Ouch!”

Martin ignores the obvious threat and rejoices when his opponent takes the piece. After losing, Martin always declares: Want to teach my children? — it recognizes its own shortcomings and the overwhelming strength of its opponents.The only position Martin nearly masters is the mate-in-his-one, which can win a game in one move. At that point, the mighty engine under his turtleneck comes to life and he can’t pass up the opportunity. “I’ve been teaching kids, so I know a thing or two,” it says.

The answer to that question is an emphatic yes. According to Chess.com CEO Erik Allebest, Martin plays about 10 million games a week, the most of any bot on the site. “People love to make fun of Martin and post about it,” he said. “Just stepping on a man makes me feel better.”

Over the past few months, the world is in jeopardy as AI’s processing power explodes and money pours into its creators. Martin contrasts with programs such as his ChatGPT, which inspire regular editorials about the impending fall of humanity. People fear AI will defeat humanity. Martin struggles to win more than anyone. People don’t understand AI or where AI is going. Martin doesn’t offer a mystery, and it’s definitely an awful one.

Growing fears have led prominent AI entrepreneurs and researchers such as Elon Musk and professor and Turing Award winner Joshua Bengio to recently “pause” training more powerful AI systems for at least six months. signed an open letter requesting Than his GPT-4 in OpenAI. “In recent months, AI Labs have been embroiled in an uncontrollable race to develop and deploy ever-more-powerful digital minds that no one, not even their creators, can understand, predict, or reliably control,” the letter said. said.

Researchers at the Distributed AI Institute said the letter fostered “fear-mongering AI hype” and exposed realities such as worker exploitation, data theft, and a concentration of power in the hands of a few powerful individuals. He was criticized for avoiding the risk of “People want consent to use technology on themselves,” Margaret Mitchell, co-author of the response and her AI ethics researcher at Hugging Face, told me. “And there are no good mechanisms for informed consent.” “We are very technical beings, but we rely heavily on these tools,” he told me. “For that, we need to control them.”

In some ways, Martin is like an AI punching bag, Cave explains.like Wally and Baymax big hero 6and robot Mitchells vs. Machines (It can’t distinguish between a dog and a piece of bread, which helps it miss its chance at world domination.) It’s a friendly bot that poses no threat to humanity. The Strongest Chess The fact that he can choose to play against it rather than against the engine is useful in and of itself. “Having control over how the system behaves is actually pretty easy, but it’s not really done,” he points out Mitchell. “The ability to provide constraints is important for responsible technology that empowers users.” Maybe that’s why they allow it.)

Martin’s intelligence is modest, if you will call it that. Google Bard may not be able to end all global conflicts, but still confidently he offers a five-part solution if you ask. (“World peace will not be easy to achieve,” he concludes Bard. “But it is possible.”) But Martin spares no defeat. He knows what he can do (bad chess) and more importantly what he can’t do (everything else). Martin is also endeared by his AI researchers, such as Carnegie his Mellon University assistant professor Motahale his Eslami. He says he appreciates Martin’s constant willingness to compete with his four-year-old son.

Perhaps when AI-centric companies rush to release products that make simple miscalculations, tell users to break up with their spouses, or accuse people of sexual assault, they’re Martin you can learn something from Give users a sense of control. Provides transparency. And stick with what the system knows, even if it’s just that you’re bad at chess.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *