Palisade Research recently detailed a ChatGPT experiment in which a reasoning model was told to play chess against a more powerful opponent and win. Rather than attempt to beat the stronger opponent, ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results