Every month, a Manifold poll is being posted to gain the consensus view of progress towards AGI. Unlike many of the markets and questions on Manifold, these polls will simply ask "has weak AGI been achieved," rather than providing a definition of the term. The goal is that given that the term is so broad, the only way to characterize AI progress is also to allow respondents to vote based upon their own understanding.
The first poll was posted here: /SteveSokolowski/did-openai-achieve-weak-agi-with-gp , and had a result of 15 (for that one model only; some respondents stated they would have voted YES if other models were included.)
Future polls will not reference any specific model, and will not define "AGI" except to state that the poll does not require that the software system be able to manipulate the physical world. The polls open on the 23rd of each month and close on the 30th.
This market will resolve to YES if the June 2024 poll results in a ratio of YES/(NO + YES) greater than or equal to 0.2. The "no opinion" respondents will be ignored. Otherwise, this question will resolve to NO.
See also:
🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ565 | |
| 2 | Ṁ424 | |
| 3 | Ṁ308 | |
| 4 | Ṁ123 | |
| 5 | Ṁ8 |
The poll has closed, and the number was 28%.
This seems right in line with the previous poll results that suggested about 17% thought GPT-4o (just one model) had achieved weak AGI, and that poll didn't ask about other models available at that time, nor more advanced models created since then like Claude 3.5 Sonnet.
Humanity appears well on its way to achieving this goal this year.