It is widely agreed amongst experts that Artificial General Intelligence — an AI that can flexibly and effectively perform a very wide range of cognitive tasks at least as well as humans — would represent a dramatic advance in power and capability for their developer. This has led to both speculation and concern that, recognizing this, countries or companies might launch "Manhattan project" style efforts to develop it. While shortening the timeline, these could easily lead to a race condition that compromises safety or alignment, or leads to adversarial dynamics during development.

A 2020 paper analyzes this issue in some detail, focusing on the "lead up" to a Manhattan/Apollo project: what understanding must be in place be "on the runway"?

As an assessment of that question, we ask here:

When will the first (of any) Manhattan/Apollo project toward AGI be launched?

We'll define this fairly simply: the project in question will be counted if:

1. It has a stated aim to develop broad-purpose artificial intelligence system(s); and

2. It is constituted so as to generate a single "flagship" system or small number of related systems (rather than a diverse portfolio of more narrowly-purposed ones); and

3. There is a research and engineering budget under the direct control of a single (executive or board) project lead with a budget exceeding the equivalent of $10B in 2020 USD.

Some fine print:

• Question will remain open until one year after the chosen resolve date, or 2050-01-01, whichever is sooner. It is open at both ends, so as to include, for example, any potential secret joint Estonia-New Zealand AGI project.

• The project must be a single, unified one, under one directorship, not e.g. a general R&D funding program distributed across many groups.

• Funding will be in 2020 USD using US CPI index inflation, and PPP-adjusted.

• Note that the budgets of the Apollo and Manhattan projects were of order USD $80B.

• There is no condition on total funding, but the $10B must be an in annual budget and allocated to, in principle, be spent in a year.

• Resolution time will be, as best as discernible, the date at which the budget allocation to the project is made official by whatever entity is allocating the funding.

• In the event that Metaculus should become highly influential and trusted, it is conceivable that this question itself could lead to adverse race dynamics. The author reserves the right to hide the community prediction, make this question private, delete it, etc., at his discretion.