The Singularity

Two chairs

Ray Kurzweil says computing's exponential growth will produce human-level AI by 2029 and merge humans with machines by 2045. If the curve holds, that's arithmetic. If it doesn't, it's a chart that flatters itself.

The question

In 2005, Ray Kurzweil's The Singularity Is Near predicted human-level AI by 2029 and a merging of human and machine intelligence by 2045, driven by exponential gains in computing. His 2024 sequel, The Singularity Is Nearer, kept both dates unchanged. Does the curve hold, or is it a chart that flatters itself?

Round 1

The chair that trusts the curveKurzweil's case rests on decades of measured exponential gains across computing, sequencing, and more — trends he tracked before they were obvious, and some of which held up.

The chair that says curves bendEvery exponential trend in the physical world eventually meets a wall — heat, cost, atoms — before it runs forever; computing's own gains have slowed and sped by turns, not stayed on one line.

Round 2

The chair that trusts the curveHe hasn't hidden from scoring himself: he's published his own 1990s and 2000s predictions and their outcomes, and reports most landing close enough to keep betting on the trend.

The chair that says curves bendA curve fit to whichever metric is climbing will always look exponential. Pick a harder one — robust real-world reasoning, a robot doing a plumber's day — and the line gets much messier.

Round 3

The chair that trusts the curveAn exponential curve looks flat for a long time, then doesn't. If seventy years of compounding continue, 2029 and 2045 aren't wild guesses — they're the trend, extended.

The chair that says curves bendKurzweil restated 2029 and 2045 unchanged in 2024, after nineteen years of a very different AI landscape. A forecast that never revises starts to look like a commitment, not a measurement.

Where are you sitting?

The chair that trusts the curve The chair that says curves bend Neither — I’m still listening

Held open

What's checkable: Kurzweil published these dates in 2005 and kept them in 2024. What isn't: whether an extrapolated curve is a law or a story that has held so far. Both chairs agree computing grew exponentially; they disagree on what that guarantees.

The curve is real. The extrapolation is a bet.

scroll to hear both

1 / 7