OptionProbability
2027 or later
2026
2024
2025
42
24
17
GPT-4.5
Grok 3
91
9
by 2040 (Age ≤45)
by 2039 (Age ≤44)
by 2035 (Age ≤40)
by 2036 (Age ≤41)
by 2037 (Age ≤42)
by 2038 (Age ≤43)
by 2034 (Age ≤39)
by 2032 (Age ≤37)
by 2033 (Age ≤38)
by 2031 (Age ≤36)
by 2028 (Age ≤33)
by 2025 (Age ≤30)
66
59
50
41
34
10
No, but transformers are still key (the breakthoughs augment upon transformers)
No, a paradigm shift is needed (e.g. JEPA, etc.
Yes: transformers scale the best
No, but the new technology is still a “large foundation model” (non-transformer e.g. Mamba, Diffusion)
46
35
11
7
