Which, if any, GPT-n will outperform AlphaGeometry merely via prompting, by 2030?
3
46
Ṁ79Ṁ235
2030
1D
1W
1M
ALL
7%
GPT-4
19%
GPT-5
21%
GPT-6
12%
GPT-7
12%
GPT-8
12%
GPT-9
16%
None
Resolves to the lowest numbered GPT that scores higher than "25" on the benchmark test set of 30 Olympiad geometry problems, as used in the AlphaGeometry paper: https://twitter.com/GoogleDeepMind/status/1747651826730610696
Both GPT-n or a derivative fine-tuned version of GPT-n count. It also cannot use any special scaffolding: it must take in the problem description in its prompt, and output the geometry problem solution in the first outputted answer (potentially after some chian of thought).
In case the architecture changes significantly such that question is no longer applicable, I will resolve as N.A..
Get Ṁ600 play money
Related questions
What will be true about GPT-5?
Will mechanistic interpretability be essentially solved for GPT-2 before 2030?
23% chance
Will an open source model beat GPT-4 in 2024?
62% chance
Who will release a GPT-4o comparable model in 2024?
Will any GPT beat Stockfish in a fair fight before 2030?
36% chance
Will any Google model exceed chatGPT interest? (by 2025)
29% chance
Will there be a GPT 5.5?
37% chance
Will we have an open-source model better than GPT-4-Turbo before 2025?
62% chance
Will a GPT-3 quality model be trained for under $10.000 by 2030?
81% chance
Will the GPT architecture be replaced by another, more efficient architecture by the end of 2024?
31% chance