Metaculus (Dec 2024)

[**578** comments](https://www.metaculus.com/questions/3479/date-weakly-general-ai-is-publicly-known/#comments)

**1.7k** forecasters

# When will the first weakly general AI system be devised, tested, and publicly announced?

Current estimate

20 Feb 2028

202020202020202020212021202220232024202720302036204220552074209621412200

Share

Predict

Top Key Factors

View all (5)

↑ reliable >50-step agent chains with published evals

Impact

later

Strength

2 votes

China starts a war with the land of Taiwan BEFORE said weakly general AI

Impact

Earlier

Strength

28 votes

↓ grid/permit delays & export controls on HBM/nodes

Impact

later

Strength

1 vote

↑ multi-year compute/colo contracts confirmed via filings

Impact

later

Strength

1 vote

AI become student of arts university in Vienna

Impact

later

Strength

23 votes

CommentsTimelineKey FactorsQuestion Info

Timeline

1d1w2mall

09 Oct 202503 Sep 202717 Mar 2030May 2033Jul 203609 Oct 202503 Sep 202717 Mar 2030May 2033Jul 2036

Dec 27Dec 29Dec 31Jan 02Jan 04Jan 06Jan 08Jan 10Jan 12Jan 14Jan 16Jan 18Jan 20Jan 22Jan 24Jan 26Jan 28Jan 30Feb 01Feb 03Feb 05Feb 07Feb 09Feb 11Feb 13Feb 15Feb 17Feb 19Feb 21Feb 23Feb 2520 Feb 2028

Resolution Criteria

For these purposes we will thus define "AI system" as a single unified software system that can satisfy the following criteria, all easily completable by a typical college-educated human.

- Able to reliably pass a Turing test of the type that would win the [Loebner Silver Prize](https://www.metaculus.com/questions/73/will-the-silver-turing-test-be-passed-by-2026/).
- Able to score 90% or more on a robust version of the [Winograd Schema Challenge](https://www.metaculus.com/questions/644/what-will-be-the-best-score-in-the-20192020-winograd-schema-ai-challenge/), e.g. the ["Winogrande" challenge](https://arxiv.org/abs/1907.10641) or comparable data set for which human performance is at 90+%
- Be able to score 75th percentile (as compared to the corresponding year's human students; this was a score of 600 in 2016) on all the full mathematics section of a circa-2015-2020 standard SAT exam, using just images of the exam pages.
- Be able to learn the classic Atari game "Montezuma's revenge" (based on just visual inputs and standard controls) and explore all 24 rooms based on the equivalent of less than 100 hours of real-time play (see [closely-related question](https://www.metaculus.com/questions/486/when-will-an-ai-achieve-competency-in-the-atari-classic-montezumas-revenge/).)

By "unified" we mean that the system is integrated enough that it can, for example, explain its reasoning on an SAT problem or Winograd schema question, or verbally report its progress and identify objects during videogame play. (This is not really meant to be an additional capability of "introspection" so much as a provision that the system _not_ simply be cobbled together as a set of sub-systems specialized to tasks like the above, but rather a single system applicable to many

... (truncated, 24 KB total)