OpenAI’s new o1 model can solve 83% of International Mathematics Olympiad problems https://www.hindustantimes.com/business/openais-new-o1-model-can-solve-83-of-international-mathematics-olympiad-problems-101726302432340.html
World’s largest ethanol-to-jet fuel plant finalized, 250mn gallon yearly output | The 60-acre facility will revolutionize the global aviation industry by providing a scalable supply of low-carbon jet fuel.
MetaKnowing on September 15, 2024 5:15 pm OpenAI’s previous model GPT-4o in comparison could only solve 13% of problems correctly vs 83% now. The new model uses a “chain of thought” process, which mimics human cognition by breaking down problems into logical, sequential steps. The model achieved gold-level performance at the International Olympiad for Informatics, which some have described as the “Olympics of coding” It also answered questions on GPQA (GPQA: A Graduate-Level Google-Proof Q&A Benchmark) above PhD level. Appears to be quite a leap forward, but I guess time will tell as more people use it.
ftgyhujikolp on September 15, 2024 6:24 pm It can’t tell you how many Rs are in “strawberry” correctly. (Unless they patched it today) The only way it is solving complex math problems is if it has seen the answers before.
2 Comments
OpenAI’s previous model GPT-4o in comparison could only solve 13% of problems correctly vs 83% now.
The new model uses a “chain of thought” process, which mimics human cognition by breaking down problems into logical, sequential steps.
The model achieved gold-level performance at the International Olympiad for Informatics, which some have described as the “Olympics of coding”
It also answered questions on GPQA (GPQA: A Graduate-Level Google-Proof Q&A Benchmark) above PhD level.
Appears to be quite a leap forward, but I guess time will tell as more people use it.
It can’t tell you how many Rs are in “strawberry” correctly.
(Unless they patched it today)
The only way it is solving complex math problems is if it has seen the answers before.