Breaking News

New artificial intelligence makes mistakes – on purpose


It took about 50 years for computers eviscerate humans in the venerable game of chess. A standard smartphone can now play the kind of moves that make a grandmaster’s head spin. But one artificial intelligence program takes a few steps back, to appreciate how average humans play – the blunders and all.

The AI ​​chess program, known as Maia, uses the kind of cutting edge AI behind the best superhuman chess programs. But instead of learning how to destroy an opponent on the board, Maia focuses on predicting human movements, including the mistakes they make.

Jon kleinberg, a Cornell University professor who led the development of Maia, says this is a first step towards developing an AI that better understands human fallibility. The hope is therefore that it can be better at interacting with humans, teaching or helping them, for example, or even negotiating with them.

One possible use, Kleinberg says, is health. A system that anticipates errors can be used to train doctors to read medical images or to help them find errors. “One way to do this is to take into account the issues in which human doctors make diagnoses based on medical images and look for images where the system predicts a high level of disagreement between them,” he says.

Kleinberg says he chose to focus on chess because it’s one of the first areas where artificial intelligence triumphed over humans. “It’s this kind of ideal system to try algorithms, “he says.” Kind of a model of AI dominance. “

Plus, he says, chess has been studied intensely, making it something similar to the fruit fly, or Drosophila, in biology. “Chess has the distinction of having been called the Drosophila of psychology by Herb Simon and Drosophila from AI by John mccarthyKleinberg said, referring to two giants in their respective fields.

Maia was developed using code adapted from Leela Zero, an open source clone of Alpha zero, a revolutionary AI program created by the Alphabet subsidiary company DeepMind.

Alpha Zero broke with conventional AI chess programs by teaching computers, independent of human instruction, how to play the game expertly. Within the program, a simulation neural network contains virtual neurons that can be tuned to respond to input. For chess, Alpha Zero is powered by board positions and movements generated in practice games, and it adjusts the fire of its neurons to favor winning shots, an approach known as reinforcement learning. Alpha Zero can use the same approach to learn to play other board games such as Checkers or Go with minimal modification.

The Cornell team modified Leela Zero’s code to create a program that learned by promoting accurate predictions of human movements. Other AI chess players, including Deep Blue, the IBM machine that then defeated the world champion Garry Kasparov in 1997, may attempt to look ahead in a game by exploring possible moves. But Maia is unusual in the way she focuses on finding the most likely move a human will play.

The WIRED Guide to Artificial Intelligence

Supersmart algorithms won’t take every job, but they are learning faster than ever, ranging from medical diagnoses to serving ads.

Maia learned using LiChess data, a popular online chess server. The result is a chess program capable of playing in a more human way. Multiple versions of Maia, suitable for different game strengths, can now be challenged at LiChess.

More sophisticated forms of AI can eventually overtake human intelligence in all kinds of fields, from mathematics to literature and beyond. But Kleinberg says “there will be a long transition period where AI and humans will work together, and there will be some communication between them.”

Long before that, an AI capable of predicting and mimicking human behavior could have immediate applications in chess and other games. “It’s a great idea,” says Matthew Sadler, British chess grandmaster and author of Game changer, a book about the chess skills of Alpha Zero. “Club players have a huge need to have engines that speak their language.”



Source link

Leave a Reply