Skip to main content
MANIFOLD
What will be true about AGI and longevity in 2040?
46
Ṁ1kṀ4.4k
2040
27%
AGI is not developed
25%
AGI is developed and significant advances are made in longevity
20%
AGI is developed and turns out unfriendly/unaligned
9%
AGI is developed, but is not used in longevity research
75%
AGI is used in longevity research, but doesn't make significant advances
13%
AGI is developed, significant advances in longevity are made, but they require pre-natal intervention
5%
AGI is developed, but stays at human-like level of capabilities

AGI is currently generally expected to appear some time in 2030s: /ManifoldAI/agi-when-resolves-to-the-year-in-wh-d5c5ad8e4708

I would like to explore whether AGI will have any significant effect on longevity research and if no, then why.

Here are some definitions of things mentioned in the options:

  • AGI — and AI system, capable of autonomously doing any tasks involved in scientific research (since this question is about research)

  • Significant advances in longevity — increasing the life-expectancy at birth by at least 2x, to at least 160 years.

  • Whether AGI is used in longevity research — yes if some notable papers in the field are fully written by AI or some procedures are fully developed by AI. Domain specific AIs used for e.g. protein folding are not enough.

  • Human-like level of capabilities — the most complex high-level intellectual tasks done by AI can also be done by humans in no more than 10x amount of time it takes AI to do it.

  • Unfriendly/unaligned AGI — the actions of out-of-control AI lead to at least the state of national emergency in at least one country and result in significant slowdown in AI development either due to the regulations introduced as a result, or due to all or many AI researchers being dead.

The options in the question are not mutually exclusive.

I do not bet on my own questions.

  • Update 2026-10-02 (PST) (AI summary of creator comment): - The definition of significant advances in longevity is updated: this will resolve to YES if there is a consensus in 2040 that a baby born in 2040 has at least a 50% chance to live to 160, given the 2040 level of medical care throughout their life.

Technology
Science
Longevity
AGI
Long-term Markets
Get
Ṁ1,000
to start trading!

People are also trading

Sort by:

A measurement point for the "significant advances" option (life expectancy at birth of at least 160 by 2040): period life expectancy is computed from one year's death rates at every age, so it only moves as deaths actually fall. The US figure for 2024 is 79.0 years (CDC, January 2026) and no national figure is above 90. Doubling by 2040 would need mortality at every age to collapse within 14 years, not just a discovery. Even if an AGI found the right drug in 2030, the human evidence would lag: a cohort enrolled at 65 reaches median survival about 21 years later. The fastest readout in a mammal that shares our environment is the dog, about 4.6 years from age 8, which is why the lifespan trials of anti-aging drugs now running are in dogs. The arithmetic with sources (mine): https://w0lph.github.io/k9/why-dogs.html?v=2

Why the path to human longevity runs through dogs
Computed from the cited life tables and trials: a lifespan answer takes 4.6 years in dogs and 21 in people; dogs share our environment and diseases, lifespan has already been extended in dogs, and a regulator has accepted lifespan extension as an indication for dog drugs.

@w0lph The way life expectance is calculated is a bit hairy, so I'll accept as "significant advance" a situation in which there is a consensus that a baby born in 2040 has 50% chances to live to 160, given the 2040 level of medical care throughout their life.

https://www.bbc.com/news/articles/cz7rxy21lxwo

opened a Ṁ50 NO at 40% order

Manifold is surprisingly bad at making probabilities sum up to 1.

bought Ṁ70 NO

Am I correct in thinking their is some bias in the predictive power of the "Will AI be unaligned" question. The incentive structure encourages users to vote NO on that, as they would not be alive to cash out on any YES rewards? Or am I mistaken. Thanks.

bought Ṁ10 NO

Will only one of these resolve Yes? Shouldn’t this be a linked multiple choice market?

bought Ṁ5 NO

@ahalekelly The options in the question are not mutually exclusive.

@ahalekelly It is a multiple-choice market, though I tried to make different options mostly complementary to each other.

People are also trading