Singularity
The hypothetical point past which technological change becomes unpredictable to humans — a term borrowed from physics, with a long record of confidently wrong dates.
When not to use it
- In technical writing. It imports a large unfalsifiable claim to describe something you could state precisely.
- As a forecast. The track record is poor and the argument is structured so that no observation counts against it.
- As a synonym for AGI or superintelligence. It's a claim about a prediction horizon, not about a capability level.
Reach for something else instead
- Intelligence explosion is the actual mechanism, and it's arguable on its merits.
- Specific capability forecasts — dated, measurable, falsifiable — are what forecasting looks like.
- Transformative AI is the term policy work reaches for when it wants "very large economic effects" without the metaphysics.
Sources & further reading
- Vinge (1993), The Coming Technological Singularity: How to Survive in the Post-Human Era — the modern statement and the source of the term's currency; dated before 2030.
- Kurzweil (2005), The Singularity Is Near: When Humans Transcend Biology — the popular case from accelerating returns; dated to 2045.
- Good (1965), Speculations Concerning the First Ultraintelligent Machine — the intelligence-explosion mechanism the whole idea depends on.
Primary sources, listed so you can check the claims on this page rather than take them on trust.
Where people go wrong
- Using it to mean "AI gets very powerful". The word means a prediction horizon; the loose usage borrows gravity it hasn't earned.
- Treating the dates as forecasts. They have tracked their authors' expected lifespans more closely than any technical indicator.
- Missing that it's downstream of Good. Every argument for the singularity is an argument for recursive self-improvement wearing a physics metaphor.