The Immortality Index

The current verdict

Measured global life expectancy at birth reached about 73.2 years in 2023, roughly double the pre-1900 scholarly estimates. That climb came from public health and medicine applied at population scale, and it has already reversed once recently, during the COVID-19 pandemic.

The cited UN medium projection expects slow further gains, to about 82 years by 2100. TaskNorth's scenario adds a stated uplift on top of that projection, then encodes three assumed step-changes with stated years: AGI-level research automation around 2045, mature quantum simulation of biology around 2080, and molecular-scale repair therapies around 2150. The third jump carries the curve past the longest verified human life, saturating at about 232 years by 3000 under a structural 250-year cap. Each milestone is an editorial assumption, not a prediction of when it happens. It is a reviewed editorial judgment, not a forecast, and no point on it implies a 1,000-year life is close or inevitable.

What moved the curve

Every marker on the chart is listed here in full: what happened, the strength of the evidence, the editorial judgment it carries, and the primary source behind it. Newest reviewed items first.

What AI is actually accelerating

Research speed is not the same thing as added years of life. These are the areas where acceleration is documented; none of them, on its own, has extended a human lifespan.

Aging biomarkers

Models that estimate biological aging signals from clinical data. Useful research instruments; their validation for predicting individual outcomes is still an open scientific question.

Target discovery

Machine learning that proposes which genes and pathways to drug. It shortens the candidate list; it does not prove any target extends human life.

Molecular and protein modelling

Structure prediction such as AlphaFold compresses months of laboratory work into hours for many research questions.

Candidate selection

AI-assisted screening ranks compounds before expensive synthesis and testing, raising the throughput of early drug discovery.

Trial analysis

Better modelling of trial data and real-world evidence can make human studies faster to read out. The trials themselves still take years.

What would change the curve next

The scenario moves only through a written revision, and these are the kinds of evidence that would justify one:

• A therapy shows replicated human outcome data for slowing a measured process of aging itself, not a single disease.
• A major demographic authority materially revises its long-run life expectancy projection, in either direction.
• An AI-discovered drug reaches approval with outcome data in an aging-related indication, shortening the discovery-to-approval path on the record.
• Negative trials, failed replications, safety signals, or worsening population outcomes accumulate; these move the scenario farther away.
• Policy or access changes measurably alter how fast approved therapies reach populations.

Methodology and sources

Metric. Global life expectancy at birth, in years. Life expectancy at birth is the average for newborns across a whole population and period. It is not maximum lifespan, healthy lifespan, biological age, or a prediction for any individual reader.

Observed series. UN World Population Prospects 2024 world estimates, 1950 to 2023, sampled every five years. Pre-1950 points are Riley (2005) scholarly estimates, shown as a separately labelled segment because they are reconstructions with wide uncertainty.

Projection. The UN medium-variant projection to 2100, cited below.

Editorial scenario. Explicit reviewed points, not a forecast and not sourced data. Through 2040 the points add a stated uplift to the UN medium projection. Beyond that, the shape encodes three assumed step-changes, each an editorial assumption with a stated year: AGI-level research automation around 2045, mature quantum simulation of biology around 2080, and molecular-scale repair therapies built on both around 2150. On owner direction (revision 1.3) the third jump breaks the former 150-year ceiling: the curve passes the longest verified human life and saturates toward a raised structural cap of 250, ending at 232 by 3000. No point implies a 1,000-year life. A milestone that fails to arrive moves its jump later or removes it.

Association is not causation. Event markers sit on the time axis as annotations. Their placement does not claim that one invention caused a precise number of life-years on the global series.

Revision history (method version 1.3)

• v1.3 · 2026-08-15 · Raised the structural scenario cap from 150 to 250 years on owner direction. The post-2150 molecular-repair jump now passes the former ceiling and the curve saturates at 232 by 3000. Sourced series and their 120-year cap unchanged.
• v1.2 · 2026-08-14 · Reshaped the post-2040 scenario from one smooth saturating path into three assumed step-changes with stated years: AGI-level research automation (~2045), mature quantum simulation of biology (~2080), and molecular-scale repair therapies (~2150). Cap and sourced series unchanged.
• v1.1 · 2026-08-14 · Extended the editorial scenario horizon from 2100 to 3000 using a stated saturating rule toward the structural 150-year cap (300-year time constant). Sourced UN projection unchanged; it still ends at 2100.
• v1.0 · 2026-08-12 · Initial dataset: UN WPP 2024 world estimates and medium projection, Riley (2005) pre-1950 estimates as a separately labelled segment, and first anchored scenario points with a stated uplift method.

Sources · last checked 2026-08

• World Population Prospects 2024, demographic indicators (world estimates) · United Nations, DESA, Population Division · Life expectancy at birth (LEx), both sexes, location World. Estimates cover 1950 to 2023. CC BY 3.0 IGO.
• World Population Prospects 2024: Summary of Results (medium variant) · United Nations, DESA, Population Division · Medium-variant projection, 2024 to 2100. Values read from the same demographic-indicators dataset. CC BY 3.0 IGO.
• Life expectancy, long-run world series (pre-1950 points from Riley 2005) · Our World in Data · World points before 1950 are Riley (2005) estimates, processed by Our World in Data (CC BY 4.0).
• Estimates of Regional and Global Life Expectancy, 1800-2001 · Population and Development Review · Riley, J.C. (2005), 31(3): 537-543. The underlying compilation behind the pre-1950 estimates.
• Oldest person ever (world record) · Guinness World Records · Jeanne Louise Calment, 122 years 164 days, fully authenticated.
• About Edward Jenner · The Jenner Institute, University of Oxford
• Investigation of the Broad Street pump outbreak · UCLA Fielding School of Public Health
• World Tuberculosis Day · World Health Organization · Marks 24 March 1882, when Robert Koch announced the discovery of the TB bacterium.
• Robert Koch, biographical · NobelPrize.org
• Leonard Thompson, the discovery of insulin at the University of Toronto · University of Toronto Libraries
• Alexander Fleming's discovery and development of penicillin · American Chemical Society
• 50th anniversary of the first effective polio vaccine, April 12, 2005 (MMWR) · Centers for Disease Control and Prevention · Live page blocks automated access; content verified against an Internet Archive capture of this CDC page.
• Randomised trial of cholesterol lowering in 4444 patients: the 4S study · The Lancet, via PubMed · Lancet 1994; 344(8934): 1383-1389.
• International consortium completes Human Genome Project · National Human Genome Research Institute
• Highly accurate protein structure prediction with AlphaFold · Nature, via PubMed · Jumper et al., Nature 596: 583-589 (2021). DOI 10.1038/s41586-021-03819-2.
• FDA approves first gene therapies to treat patients with sickle cell disease · U.S. Food and Drug Administration · Live page blocks automated access; content verified against an Internet Archive capture of this FDA page.
• Semaglutide and cardiovascular outcomes in obesity without diabetes (SELECT) · NEJM, via PubMed · Lincoff et al., N Engl J Med 2023; 389(24): 2221-2232. DOI 10.1056/NEJMoa2307563.
• Safety overview: GPT-6 Astra · OpenAI · OpenAI's own safety overview, recording the first Critical cyber capability classification under its Preparedness Framework.

Immortality Index FAQ

Will humans live 1,000 years?

Nothing observed supports that. The longest verified human life is 122 years. Global life expectancy at birth is about 73.2 years as of 2023, and the cited UN projection stays near 82 by 2100. The 1,000-year figure exists on this page only as a labelled reference far off the chart's scale.

What is the Immortality Index?

A sourced editorial model of longevity. It shows measured global life expectancy, a cited demographic projection, and TaskNorth's clearly labelled scenario for how AI could accelerate longevity research. The curve and its reviewed revisions are the index; there is no invented score.

What is longevity escape velocity?

The idea that medicine could someday add more than a year of expected life per calendar year, so death from aging keeps receding. No population data shows this happening. This page's status line reads: not reached.

Can AI cure aging?

No approved therapy reverses human aging, with or without AI. What AI demonstrably accelerates today is research: protein structures, target discovery, candidate screening, and trial analysis. Whether that converts into approved treatments with replicated human outcomes is exactly what this page tracks.

Is this medical advice?

No. The page is an editorial model of population-level research progress. It does not predict any individual's lifespan and does not recommend any treatment, supplement, or intervention. Talk to a clinician about personal health decisions.

How does the curve get updated?

New events enter through a reviewed dataset change with a primary source, an evidence level, and a written rationale. Scenario movements are separate, versioned revisions listed in the methodology section. Setbacks move the scenario farther away; nothing updates silently.