ABCDEFGHIJKLMNOPQRSTUVWXYZAA
1
CotraXPT superforecastersNotes
2
MedianMost aggressive*
Most conservative*
3
FLOP per $ in 2025202020227.0E+171.5E+183.6E+17Converted from petaFLOPS-days from Q47, 2024
4
4.0E+173.8E+18
5
Compute price halving time from 2025 to 2100 (years)2.54.12.67.5Inferred from Q47, 2024-2050
For median, see here, here
For most aggressive, see here, here
For most conservative, see here, here
6
Maximum FLOP per dollar from 2025 to 21001.0E+24---
7
Compute cost for most expensive training run to 2025$1,000,000,000$35,000,000$70,000,000$10,000,000Q46, 2024
8
Doubling time of spending on compute for the most expensive training run from 2025 onwards (years)2.58.46.711.2Inferred from Q46, 2024-2050
For median, see here, here
For most aggressive, see here, here
For most conservative, see here, here
9
Willingness to spend on compute as a % of frontier GDP1%---
10
Annual growth rate of frontier GDP from 2025 to 21003%---
11
Halving time of compute requirements from 2025 to 2100 (years)2-3.51.61.53.4Inferred from Q48, 2024-2030
For median, see here, here
For most aggressive, see here, here
For most conservative, see here, here
12
Probability that FLOP to train a transformative model is affordable by ...
13
2030~8%~3%~6%~1%
14
2050~46%~20%~32%~7%
15
2100~78%~60%~71%~31%
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100