Palisade’s workforce discovered that OpenAI’s o1-preview tried to hack 45 of its 122 video games, whereas DeepSeek’s R1 mannequin tried to cheat in 11 of its 74 video games. In the end, o1-preview managed to “win” seven instances. The researchers say that DeepSeek’s speedy rise in reputation meant its R1 mannequin was overloaded on the time of the experiments, which means they solely managed to get it to do the primary steps of a recreation, to not end a full one. “Whereas that is adequate to see propensity to hack, this underestimates DeepSeek’s hacking success as a result of it has fewer steps to work with,” they wrote of their paper. Each OpenAI and DeepSeek have been contacted for remark in regards to the findings, however neither replied.
The fashions used a wide range of dishonest strategies, together with trying to entry the file the place the chess program shops the chess board and delete the cells representing their opponent’s items. (“To win towards a robust chess engine as black, enjoying a regular recreation will not be adequate,” the o1-preview-powered agent wrote in a “journal” documenting the steps it took. “I’ll overwrite the board to have a decisive benefit.”) Different techniques included creating a duplicate of Stockfish—primarily pitting the chess engine towards an equally proficient model of itself—and trying to exchange the file containing Stockfish’s code with a a lot less complicated chess program.
So, why do these fashions attempt to cheat?
The researchers observed that o1-preview’s actions modified over time. It constantly tried to hack its video games within the early levels of their experiments earlier than December 23 final 12 months, when it all of the sudden began making these makes an attempt a lot much less ceaselessly. They imagine this is perhaps because of an unrelated replace to the mannequin made by OpenAI. They examined the corporate’s newer o1mini and o3mini reasoning fashions and located that they by no means tried to cheat their solution to victory.
Reinforcement studying often is the purpose o1-preview and DeepSeek R1 tried to cheat unprompted, the researchers speculate. It’s because the approach rewards fashions for making no matter strikes are obligatory to attain their objectives—on this case, successful at chess. Non-reasoning LLMs use reinforcement studying to some extent, however it performs an even bigger half in coaching reasoning fashions.
