- O1 can produce a long internal chain of thought before responding to the user.
- O1 ranks in the 89th percentile on competitive programming questions.
- O1 places among the top 500 students in the US and qualifies for the US Math Olympiad.
- O1 exceeds human PhD level accuracy on a benchmark of physics, biology and chemistry problems.
- O1 is still under development, and the creators are not currently releasing the chain of thought to users.
The video concludes by discussing the ethical implications of AI models that can reason and solve problems at such a high level. The fact that O1 hides its chain of thought is a concern, as it makes it difficult to understand how the model arrives at its answers. (this posting completed with GenAI assistance)
