2026-01-20

The Singularity Countdown: AGI by 2029, Humans Merge with AI, Intelligence 1000x | Ray Kurzweil

Ray Kurzweil sits down with the Moonshots hosts and holds to his 1999 call: human-level AGI by 2029 and the Singularity (humans about 1000x more intelligent after merging with AI) by 2045. He defines AGI strictly as matching top experts in thousands of fields, expects longevity escape velocity around 2032, avatars of himself this year, and UBI-like systems in the 2030s. He argues the debate has shifted from 'will it happen' to 'is it good for humanity', says jobs will decouple from income, and rates himself a 10 out of 10 on optimism. Side threads cover consciousness, cryonics (his 'plan D'), computronium, slower low-power chips, and robotics arriving in 2026-27.

▶ Watch on YouTube

Topics

AGI by 2029 and strict definition of AGI AI ▶ 6:45
Kurzweil restates his 1999 call for human-level AI by 2029 and defines AGI as matching the best human in thousands of fields and combining insights across them. He expects a 3-4 year band of growing agreement ending by 2029.
Singularity 2045 and merging with AI AI ▶ 7:14
Humans will merge with AI via nanotech so biological and computational thoughts are indistinguishable, making us about a thousand times smarter by 2045. He frames the next 10 years as 100 years of progress.
Turing test, goalposts and exponential thinking AI ▶ 12:27
Kurzweil says his accuracy comes from thinking exponentially while most people think linearly. He dismisses the Loebner prize as a poor Turing test and says AGI is a stricter bar.
AI consciousness and personhood AI ▶ 22:28
Consciousness is subjective with no scientific test, Kurzweil says, but AIs that keep acting conscious will be accepted within a few years of 2029. Alex pushes for self-awareness benchmarks and AI personhood.
Avatars and digital resurrection AI ▶ 33:09
Kurzweil is building an avatar of himself from his books and articles, due this year, and has made one of his father. Full reconstruction of a person is 10-15 years away.
Longevity escape velocity and cryonics Longevity ▶ 36:38
Kurzweil predicts longevity escape velocity in 2032, driven by AI biological simulation rather than nanotech, and advises staying healthy until then. He calls cryonics 'plan D' and says he has signed up.
Future of work, UBI and society Economy ▶ 31:03
Employment will decouple from financial security; everyone will be wealthy by today's standards but jobs will be fewer. He holds to his TED prediction of UBI-like systems by the 2030s and sees this as liberating.
BCI, wearables and the end of the phone Other ▶ 1:04:28
Phones are already an extension of the mind; they will be replaced by always-on VR/AR interfaces and later neural interfaces. Kurzweil declines to predict how BCI will feel.
Computronium, compute and energy Compute ▶ 1:20:50
Computronium by 2045 would let one liter match 10 billion humans. He proposes slower, parallel, 3D molecular chips and reversible computing to fix the power problem, and is lukewarm on nuclear.
Robotics and advice for a 25-year-old Robotics ▶ 1:29:44
Robotics begins to matter in 2026-27, including robots that clear dishes. For young people, the opportunity to build AI-applied businesses is higher than ever.
Kurzweil's track record Other ▶ 0:29
Claims 120-odd predictions over 30 years with three wrong (self-driving cars, nanotech clothing, DNA computers raised by hosts); Wikipedia lists about 86% accuracy.

Predictions made

open Ray Kurzweil: AI reaches human-level intelligence (first published 1989, restated 1999).
· due: 2029 · ▶ watch
“we're going to reach human level AI by uh by 2029”
Your call:
open Ray Kurzweil: AGI defined as expert in every area (thousands of fields) with cross-field insight will be reached.
· due: 2029 · ▶ watch
“I define it as being an expert in every area”
Your call:
open Ray Kurzweil: General agreement that AGI exists will arrive after a 3-4 year band of growing acceptance.
· due: 2029 · ▶ watch
“that's a band of maybe three four years and I think it will end in 2029”
Your call:
open Ray Kurzweil: Despite Musk saying 2026, people will not be convinced of AGI in 2026; maybe 2027-2028, and certainly everyone accepts by 2029.
· due: 2029 · ▶ watch
“by 2029, I think everyone will accept that”
Your call:
open Ray Kurzweil: The Singularity: humans merged with AI become about a thousand times more intelligent.
· due: 2045 · ▶ watch
“we're going to become a thousand times smarter by 2045”
Your call:
open Ray Kurzweil: The next 10 years reach his singularity definition, at least a thousand times more intelligent.
· due: 2035-2036 · ▶ watch
“the next 10 years will get us to my definition of singularity”
Your call:
open Ray Kurzweil: 2025 to 2035 delivers as much change as the previous 100 years, with 2035 well past AGI and humans merged with machines.
· due: 2035 · ▶ watch
“2035 will be way past AGI”
Your call:
open Ray Kurzweil: Longevity escape velocity is reached: each year of medical progress gives back at least a year of life.
· due: 2032 · ▶ watch
“2032”
Your call:
open Ray Kurzweil: AI-driven biological simulation (millions of tests in a weekend) matures within about five years, plus a couple more years to cover most medical problems.
· due: 2031-2033 · ▶ watch
“five years is like an outside limit”
Your call:
open Ray Kurzweil: Most conservative modeling view: full biological simulation to predict how chemicals act is about 5 years away.
· due: 2031 · ▶ watch
“it will take about 5 years from now”
Your call:
open Ray Kurzweil: Every human organ could be replaced, making the body last more or less forever.
· due: within a decade or two · ▶ watch
“Certainly within a decade or two, we can replace all of our organs”
Your call:
open Ray Kurzweil: Interactive AI avatar of Kurzweil, available in any language, will be offered to interviewers.
· due: 2026 · ▶ watch
“that'll be this year”
Your call:
open Ray Kurzweil: AI can reconstruct everything about a person, including body state, eventually.
· due: 2036-2041 · ▶ watch
“probably another, you know, 10 or 15 years away”
Your call:
open Peter Diamandis: Kurzweil's standing prediction of high-bandwidth brain-computer interface (stated by Peter; Kurzweil declined to give details on feel).
· due: mid-2030s · ▶ watch
“we'll hit, you know, high bandwidth BCI uh in the mid 2030s”
Your call:
open Ray Kurzweil: A universal basic income or equivalent will be developed, as he predicted at TED.
· due: 2030s · ▶ watch
“we would develop UBI by the 2030s and I think that's still true”
Your call:
open Ray Kurzweil: Computronium (matter arranged for maximum computation, a liter-scale cube exceeding 10 billion people combined) exists.
· due: 2045 · ▶ watch
“we're going to have things like computronium certainly by 2045”
Your call:
open Ray Kurzweil: The capability to run the ultimate software on computronium exists, with intelligence beyond imagination.
· due: 2086 · ▶ watch
“running the ultimate software certainly by 2086”
Your call:
open Ray Kurzweil: Three-dimensional molecular-level circuits where one liter of computing matches 10 billion human beings.
· due: unspecified · ▶ watch
“one liter of computing will match uh 10 billion human beings”
Your call:
open Ray Kurzweil: Practical robotics (e.g. clearing dishes after a meal like a human) begins to affect daily life.
· due: 2026-2027 · ▶ watch
“I think that's going to begin to take place in 2026, 2027”
Your call:
open Ray Kurzweil: A few years after AIs act conscious (outer bound 2029), society will accept AI as conscious.
· due: 2030-2032 · ▶ watch
“when we're a few years into AI entities acting conscious, we will accept it”
Your call:
open Ray Kurzweil: The phone is replaced by always-on VR/AR interfaces, and later interfaces inside the nervous system.
· due: unspecified · ▶ watch
“it'll actually go inside our nervous system”
Your call:
open Ray Kurzweil: His autobiography is published.
· due: 2027 · ▶ watch
“It should be out within a year”
Your call:
open Ray Kurzweil: Longevity technologies unveil themselves to people within roughly 5 to 8 years (host framing, Kurzweil agreed).
· due: 2031-2034 · ▶ watch
“Yeah, absolutely.”
Your call:

Numbers that matter

Worth digging into

🕳️ Kurzweil's prediction scorecard
He is the most-tracked forecaster here and claims 3 misses of 120; the 86% Wikipedia figure disagrees with that.
🕳️ Longevity escape velocity 2032 and AI biological simulation
The claim rests on simulating millions of drug tests in a weekend within about 5 years, which is checkable.
🕳️ Computronium and Seth Lloyd's ultimate computer limits
The 10-billion-humans-per-liter claim anchors the 2045 singularity math.
🕳️ AI self-awareness benchmarks and personhood
Alex disputes that there are no tests; residual-stream activation detection is an emerging measure.
🕳️ Slower, parallel, low-power chips and reversible computing
A possible fix for the AI energy bottleneck that the hosts see as a bigger limit than intelligence.
🕳️ Kurzweil avatar and father avatar
A live test of digital-twin interviews planned for 2026 and an avatar-hosted Moonshots episode idea.