Elon Musk

Sources#

The full-length interview with Elon Musk

Summary#

Elon Musk enters this corpus through a single long-form source — an 85-minute Economist interview published 2026-07-29 — and is worth an entity page for one reason beyond his companies: he is the wiki's best-documented instance of a public reversal on AI risk by someone with the capacity to act on either position. The interview is unusual in that the interviewer walks him through his own prior statements and he does not dispute the record, only the conclusion he now draws from it.

He is a co-founder and original funder of OpenAI, founder of xAI (Grok), Tesla and SpaceX, and — by his own framing in this interview — a builder of the physical half of AI: humanoid robots as the "end effectors" digital intelligence lacks (see Post-Scarcity Macroeconomics).

The reversal, as the interview establishes it#

Four dated positions, each raised by the interviewer and confirmed by Musk:

When	Position
~2016 (Neil deGrasse Tyson podcast)	What terrified him most was rapid recursive self-improvement; humans would be "the pet labradors at best"
2023	Signed the six-month pause letter — which he now downplays: "I signed on as one of many, like 500 people… I was just one of many signatures on that thing"
~2025 (Ted Cruz podcast)	10–20% chance of "killer robots wiping out humanity"
2026 (this interview)	Risk "is not zero," but: "my philosophical conclusion is to look on the bright side… even if there was a stop button we probably shouldn't press it, because the most likely outcome is incredible abundance for all"

He does not retract the 10–20% figure when it is put back to him — his answer is that the probability is unchanged and the response to it has changed. Asked whether he would board a rocket with a 10–20% chance of killing him, he says yes and immediately supplies the reasoning that carries the whole position: "let's say there's [something] you can't do anything about it… I can't see any way to really stop this incredible momentum."

The interviewer's summary — "you seem to have decided it's inevitable, let's hope for the best and go for the ride" — draws "Yeah. I mean, yes, pretty much." Inevitability, not a lowered risk estimate, is what moved him. This is a distinct position from both governance poles the wiki records: not coordinated slowdown and not distribution-as-safety, but acceleration with a conscience clause — Cross-Lab Pre-Release Review is the clause.

The acceleration-regret account of OpenAI#

His stated reason for believing stopping is impossible is autobiographical, and it is the most interesting causal claim he makes:

"For the longest time I declined to participate in AI, or created OpenAI as essentially a counterweight to Google, because at the time they had more or less a monopoly on AI. And then Anthropic spun out of OpenAI… if anything it seems like what my actions have really accelerated AI. I wasn't really aiming to accelerate AI, but in creating OpenAI and then having Anthropic spin out of that… these actions have actually resulted in knock-on effects that accelerated AI, which wasn't really my intention. So it just seems like all roads lead to acceleration of AI."

The generalization from n=1 is doing heavy lifting — an intervention that backfired is read as evidence that all interventions backfire — but it is a genuine and specific mechanism (a safety-motivated counterweight becoming a competitive spur, twice), and it is the load-bearing support for his current stance. See OpenAI for the corpus's other accounts of that founding.

His grievance against Sam Altman is stated plainly and separately: a nonprofit "meant to be an open-source AI company… owned by the world" becoming "an $800 billion for-profit company with closed source." He reports that Anthropic's founders left OpenAI "because they didn't trust Sam Altman — otherwise Anthropic wouldn't exist," describes Dario Amodei as "a very principled person," and says of Anthropic staff that "no one set off my evil detector."

Reading his forecasts#

Musk makes an explicit calibration claim, unprompted, near the end: "I'm very good at predicting the future. I'm not perfect, but my batting average is very high for a human… this does have somewhat of a Cassandra effect where people don't believe that what I'm saying is likely to be true… maybe not exactly in the timing that I thought, but it will come to pass."

Two things to hold onto when using this source:

The timing carve-out is unfalsifiable as stated — a prediction that will come true but perhaps not on schedule cannot be graded. Where his claims here are dated (AI exceeding the sum of human intelligence in ~5 years; the AI-robot singularity ~10 years; deflation rather than inflation; civil war in Britain ~20 years), the wiki records the date with the claim so that grading stays possible.
He does his own ranking of them, which is the most useful thing he says about his own confidence: asked whether he is more confident about his AI forecasts or his political ones, he answers AI, and volunteers that the AI-robot singularity "will happen sooner than civil war in Britain… civil war in Britain is probably 20 years away, and the AI robot singularity is 10 years away," which "probably renders this other stuff that we've been talking about less important and less relevant."

Positions recorded elsewhere in the wiki#

Abundance and the end of money — Post-Scarcity Macroeconomics: quasi-infinite economy, deflation-not-inflation, universal high income, work-as-gardening, Stockfish-level software.
Cross-lab governance — Cross-Lab Pre-Release Review: bi-weekly frontier-lab calls, 1–2 weeks of competitor API access before release, government as escalation, Chinese labs included.
The compute geography — Effective Compute Scaling: power and cooling bind outside China, chips bind inside it, orbital datacenters as the way past the first.

Outside the wiki's domains, roughly 40% of the interview is a contested exchange on European politics, immigration, and his posting on X, plus Starlink's role in Ukraine and his account of DOGE. Those are not compiled — they carry no AI-technical content — with one exception noted above: his own ranking of political against AI predictions, which is calibration evidence about how he weights his own forecasts.

Corporate control, and why he says he wants it#

Asked whether it is sensible for one person to hold ~80% of voting shares, his answer is entirely about time horizon, and it is the same incentive analysis the wiki records elsewhere from the other side: quarterly-earnings pressure makes 5-to-10-year investment (a Mars base, in his example, "literally listed" in the S-1) punishable, and it traces to portfolio managers whose own compensation is evaluated over a year or less — "once you understand the incentive structure, the behavior naturally follows." On key-man risk he concedes the shape of the loss precisely: post-Jobs Apple makes "amazing phones, great computers" but no breakthrough products, and he expects his companies would "do very well for several years" on the existing roadmap.

Connections#

Cross-Lab Pre-Release Review — his governance proposal, and the piece that makes his acceleration position something other than pure fatalism
Post-Scarcity Macroeconomics — his economic forecast, where the abundance claim that justifies not pressing the stop button is set out in full
OpenAI — co-founded it as a counterweight to Google; his account of its transformation and of Anthropic's spin-out is the source of his "all roads lead to acceleration" conclusion
Recursive Self-Improvement — the specific mechanism he named in ~2016 as what terrified him most, and the one he has stopped citing
Frontier Pause Verification — the position he has moved away from: he signed the 2023 pause letter and now argues a stop button shouldn't be pressed even if it existed
Andrej Karpathy — founding OpenAI team member; the corpus's other voice from that lineage, on the engineering rather than the governance side
Frontier AI Standards Body — the provenance of his own governance proposal: he says he "talked to [Demis Hassabis] for a few hours before he published his piece," which appeared 2026-07-14, fifteen days before this interview. His three-tier proposal and Hassabis's Standards Body are two outputs of one conversation rather than independent convergence

Open Questions#

His dated forecasts are gradable and the wiki now holds them with dates attached: AI exceeding the sum of human intelligence ~2031, AI-robot singularity ~2036, deflation as the macro problem. Grade at each trigger rather than accepting the "right but mistimed" carve-out.
Is the acceleration-regret generalization sound — does the OpenAI case actually support "all roads lead to acceleration," or is it one intervention with an identifiable design flaw (a nonprofit with no mechanism to stay one)?

Sources#

The full-length interview with Elon Musk — The Economist, full-length interview, 2026-07-29 (prediction, ~1:25 runtime, auto-caption transcript, interviewer unnamed in the captions). Sole source for this page. Speaker labels are absent from the auto-captions, so attribution is by content; the prior-statement dates in the reversal table are as put to Musk by the interviewer and confirmed by him, not independently verified against the original podcasts.