AI plays chess with humans, like humans

AI plays chess with humans, like humans
There is certainly no shortage of algorithms trained to play chess: ever since Deep Blue by IBM in the now distant 1997 came out head-on from the comparison with the champion Kasparov, systems of this type have multiplied, putting in serious difficulty even the most experienced players.

Maia is the AI ​​who plays chess and makes mistakes like us

The level of challenge proposed is however too high for mere mortals: to propose a solution l artificial intelligence developed by Cornell University. An advanced AI, but not perfect. Effective in reading the situation and the opponent, but subject to a margin of error like any human being. She was baptized Maia.

To instruct her, we did not start from the analysis of the moves necessary to win a game, but from those that a real player would perform. Playful purposes aside, the project is of particular interest as it aims to create an artificial intelligence closer to us, less cold. The experiment collected positive feedback during the testing phase, involving the community of the site with over 40,000 games played during the first week. It is able to identify the mistakes made by the opponents, then bringing them to their attention so as to help them improve.

The same approach could be replicated in other areas, potentially all those in which interaction with a an algorithm, resulting in a more humane AI.

Source: Cornell University