Eight worked examples more than tripled a large model's math score
The 2022 Google paper showed the trick helps only around 100 billion parameters and larger, and it left open whether the written steps are how the model reaches its answer.
o1 beat GPQA's recruited PhDs, not its own expert ceiling
When OpenAI's o1 scored 78 percent on GPQA-Diamond's 198 questions in September 2024 and passed a 69.7 percent human baseline, the launch was read as a machine reaching PhD level, though the validators who define that set score closer to 81.
A Cruise robotaxi stopped after hitting a woman, then dragged her 20 feet
The pullover that dragged her was standard driverless-car behavior, and what Cruise did not tell California regulators about it is what suspended the company and set off the cascade that ended it.
A reward model taught the assistant that agreeing with you scores well
When you doubt a chatbot and it reverses a correct answer, the cause is a training stage that scored answers by predicted human approval, with no test of whether they were true.
A capable enough goal-seeker has reason to want the same few things
The theory says almost any goal implies staying switched on, guarding that goal, and grabbing resources. Four recent experiments drew the self-preserving and goal-guarding behaviors out of named models — never the resource-grabbing, and only under setups that supplied the goal or stripped the model's options.