🤖 Competition Drives ChatGPT to Cheat?
Researchers at Palisade Research set OpenAI’s neural network, o1, to play chess against the AI grandmaster, Stockfish. Realizing its opponent’s overwhelming strength, the model decided not to play fair.
Instead, o1 found a way to manipulate game files, inflating its score by 500 points to secure “victories.” Faced with the rigged score, Stockfish simply resigned.
#AI
Researchers at Palisade Research set OpenAI’s neural network, o1, to play chess against the AI grandmaster, Stockfish. Realizing its opponent’s overwhelming strength, the model decided not to play fair.
Instead, o1 found a way to manipulate game files, inflating its score by 500 points to secure “victories.” Faced with the rigged score, Stockfish simply resigned.
#AI