Who said what about superintelligence: dated quotes, believers and sceptics

Superintelligence quotes get passed around without their dates, their venues or their verbs, so this page keeps them in one place. Below is a short, dated database of who said what about superintelligence, with the believers and the sceptics side by side. Every quotation here was checked word for word against a page we opened in October 2026, and where we could only confirm a paraphrase we say so and do not use quotation marks.

The point is not to crown a winner. The useful habit is to sort each line into one of three bins: an aspiration (what the speaker wants to build), a forecast (a dated, checkable claim about when), or a critique (a claim about the claims). Most of the confusion in this debate comes from treating the first as the second.

The builders: aspiration and product vision

Ilya Sutskever, co-founder, Safe Superintelligence Inc. Launch page, announced June 19, 2024 per Techmeme: "Building safe superintelligence (SSI) is the most important technical problem of our time." Context: a mission statement. It gives no date for arrival.

Sam Altman, CEO, OpenAI. Personal blog post "Reflections," January 2025: "We are beginning to turn our aim beyond that, to superintelligence in the true sense of the word." The sentence just before it reads: "We are now confident we know how to build AGI as we have traditionally understood it." Context: a company claim and a stated aim, with no timetable attached to the superintelligence line.

Mark Zuckerberg, CEO, Meta. Memo "Personal Superintelligence," July 30, 2025: "Personal superintelligence that knows us deeply, understands our goals, and can help us achieve them will be by far the most useful." Context: a product vision. Here "superintelligence" names a direction for personal assistants, which is a narrower thing than the term means in the safety debate.

The forecasters: dates you can check

Demis Hassabis, CEO, Google DeepMind. NBC News, March 17, 2025: "I think over the next five to 10 years, a lot of those capabilities will start coming to the fore and we'll start moving towards what we call artificial general intelligence." Context: a hedged range, and note the verbs: capabilities "start coming to the fore," systems "start moving towards" AGI. He does not say it arrives.

Elon Musk, CEO, Tesla and SpaceX, and founder of xAI. Davos conversation with Larry Fink, January 23, 2026, per a transcript: "AI will be smarter than all of humanity collectively." In the same answer he put this at "2030 or 2031" and said AI might be smarter than any human by the end of 2026. Context: the most specific public forecast on this page. The first deadline falls in about three months, so it will soon be testable. Our AGI timeline tracks it alongside the others.

Dario Amodei, CEO, Anthropic. Essay "The Adolescence of Technology," January 2026: "it cannot possibly be more than a few years before AI is better than humans at essentially everything." Context: a forecast conditioned on "if the exponential continues," which he himself calls "not certain."

The alarm raisers: forecast plus warning

Geoffrey Hinton, Nobel laureate and former Google researcher. Closed-door Capitol Hill briefing, September 16, 2026, reported by NBC News: "Now people are saying, a lot of the researchers are saying only a few years." Context: he is reporting what other researchers say about timing, then arguing Congress has "Maybe a year, but not much more than a year" to act. His earlier figures were "maybe 30 years, maybe 50 years." For how he and others put numbers on the risk, see P(doom) explained.

Yoshua Bengio, professor and founder of LawZero. Remarks to the UN Security Council, September 23, 2026, per BNN Bloomberg: countries have an interest "in not allowing the development of autonomous superintelligence systems until they can be controlled." Context: a policy ask rather than a date.

Statement on Superintelligence, Future of Life Institute, October 22, 2025: "We call for a prohibition on the development of superintelligence, not lifted before there is broad scientific consensus that it will be done safely and controllably, and strong public buy-in." Context: a conditional call that Hinton and many others signed. The page itself carries no date, and its signature count changes, so check the live figure.

The sceptics: a different kind of disagreement

Yann LeCun, AI scientist and Turing Award winner, now at AMI Labs. Synapse India Conclave, February 21, 2026, as reported by ThePrint (paraphrased here, because we could not confirm his exact wording on the raw page): he said superintelligence will arrive, but not through large language models. Context: this is a sceptic of a method, not of the destination. He expects superintelligence and disputes the route.

Timnit Gebru and Emily Bender, researchers. MIT Technology Review, September 22, 2026: "Claims of incipient, dangerous superintelligence are not based in good scientific or engineering practice." Context: this is a sceptic of the claim itself, and of the framing. They argue that talking about "rogue models" ascribes agency to products rather than to the companies building them. Bender's January 2026 encyclopedia entry on AI makes a related argument about the ideology behind artificial general intelligence.

One more voice sits at the edge of this page. Bill Gates told Ezra Klein on September 29, 2026, according to The Next Web's report, that AI crossed cyber and bio danger thresholds early this year and that relying on the industry to regulate itself is a mistake. We have not placed it in the table, because we could not open the primary transcript and will not put a secondary quotation in quotation marks.

How to read the table

Three patterns stand out. First, the builders mostly speak in aspirations, and the people with the most specific dates (Musk, Amodei) are also the ones whose deadlines are closest. Second, the "believers versus sceptics" split is too crude: LeCun and Hinton both expect superintelligence, and they differ on method and on urgency. Third, the warnings and the forecasts are often one sentence, which means a quote used to prove that superintelligence is near is frequently the same quote used to argue for regulation.

What we would do

Treat a quote as evidence only for what the speaker is positioned to know. A lab chief's date tells you about the lab's plans and fundraising stance. A researcher's warning tells you about their model of risk. A critic's claim tells you where the burden of proof sits. We will keep this page current: when a dated forecast comes due, we will mark whether it landed, and we will add quotes only after opening the primary page.