In 2020, Joe Carlsmith estimated that 10^15 FLOPS is "enough" to perform human functionality. The next post estimating this number will believe it is >=10^15
Plus
12
Ṁ256May 25
32%
chance
1D
1W
1M
ALL
In OpenPhil's 2020 report https://www.openphilanthropy.org/brain-computation-report , Joe Carlsmith writes
> Overall, I think it more likely than not that 10^15 FLOP/s is enough to perform tasks as well as the human brain (given the right software, which may be very hard to create). And I think it unlikely (<10%) that more than 10^21 FLOP/s is required. But I’m not a neuroscientist, and there’s no consensus in neuroscience (or elsewhere).
This question will resolve when someone posts an estimate >1.5k words with >10 citations. If no such estimate is posted by May 24th 2024, I will resolve YES because `>=10^15` will still be the standing estimate.
See also https://aiimpacts.org/brain-performance-in-flops
This question is managed and resolved by Manifold.
Get
1,000and
3.00
1,0003.00Related questions
Related questions
When will a US government AI run overtake private AI compute by FLOP?
Will a machine learning training run exceed 10^25 FLOP in China before 2025?
72% chance
At least one of the most powerful neural nets at end of 2026 will be trained using 10^27 FLOPs
81% chance
Will a machine learning training run exceed 10^27 FLOP in China before 2028?
44% chance
Will any supercomputer reach 1 zettaFLOP before 2035?
82% chance
Assuming SB 1047 is passed, will the "compute threshold" of 10^26 flop be raised before 2030?
64% chance
Will a machine learning training run exceed 10^26 FLOP in China before 2027?
53% chance
Will a machine learning training run exceed 10^27 FLOP in China before 2030?
66% chance
Will a machine learning training run exceed 10^26 FLOP in China before 2026?
52% chance
Will a machine learning training run exceed 10^26 FLOP in China before 2028?
64% chance