When will self-improving AI outperform human-developed AI?
16
90
Ṁ397Ṁ510
2051
2030
expected
1D
1W
1M
ALL
After self-improving AI begins to produce state-of-the-art progress in AI research, this market will resolve once human teams have failed to produce a new state-of-the-art result on any commonly accepted benchmark for 1 year.
Hybrid teams with substantial human input will count as human-developed.
Achieving SOTA on a new benchmark or dataset only because it was not available to the other side is not sufficient.
Results beyond 2050 will resolve to 2050. If it's discovered that such a system already existed pre-2010, it will resolve to 2010.
Get Ṁ200 play money
Related questions
In 2028, will an AI be able to play randomly-selected computer games at human level, given the chance to train via self-play?
66% chance
Will AI surpass humans in conducting scientific research by 2030?
31% chance
Will AI be capable of superhuman persuasion well before (>1yr) superhuman general intelligence?
52% chance
When will AI be better than humans at AI research? (Basically AGI)
When will an AI model be better than me at competitive programming?
By 2025, will most well-educated people expect AI to within 10 years be better at intellectual work than 99% of humans?
19% chance
Will AI reach human-level performance in Magic: The Gathering before 2026?
54% chance
By when will AIs perform at least as well as humans on GAIA?
Sometime before 2031, AI is going to cross a generality threshold that allows for recursive self-improvement, and then FOOM to superhuman capability. Before that point humanity must solve AI safety or be helpless.
26% chance
AI: When will a phone out-drive a human?
2030