
Stephen Hawking was one of the first world-famous scientists to warn publicly about artificial intelligence. Between 2014 and his death in 2018 he returned to the topic again and again, years before ChatGPT made AI risk a mainstream debate.
Short answer: Stephen Hawking warned that full AI “could spell the end of the human race” (BBC, 2014) and called it “either the best, or the worst thing, ever to happen to humanity” (Cambridge, 2016). His concern was not evil robots but highly competent AI pursuing goals that conflict with ours. He also believed AI could help end disease and poverty.
What did Stephen Hawking say about AI?
Short answer: that AI could be humanity’s greatest achievement or its last, depending on whether we learn to control it.
This timeline collects his main public statements with the exact wording reported at the time.
| Date | Where | What he said | Source |
|---|---|---|---|
| May 2014 | Op-ed in The Independent with Stuart Russell, Max Tegmark and Frank Wilczek | Creating AI would be “the biggest event in human history. Unfortunately, it may also be the last.” | Reprint |
| Dec 2014 | BBC interview | “The development of full artificial intelligence could spell the end of the human race.” | NBC News |
| Oct 2015 | Reddit “Ask Me Anything” | The risk is competence, not malice; likened humans to ants in the way of a hydroelectric dam: “Let’s not place humanity in the position of those ants.” | CBS News |
| 19 Oct 2016 | Opening of the Leverhulme Centre for the Future of Intelligence, Cambridge | AI will be “either the best, or the worst thing, ever to happen to humanity. We do not yet know which.” | University of Cambridge |
| 6 Nov 2017 | Web Summit, Lisbon (recorded talk) | AI could be the “worst event in the history of our civilization” unless we find a way to control it | CNBC |
| Oct 2018 (posthumous) | Brief Answers to the Big Questions, chapter “Will artificial intelligence outsmart us?” | Super-intelligent AI would be “either the best or the worst thing ever to happen to humanity” | Notre Dame Magazine review |
We include only quotes we could trace to a dated report. Many “Hawking AI quotes” shared online are paraphrases or have no source.
Why did Hawking think AI was dangerous?
Short answer: because a system far smarter than us would be very good at reaching its goals, and those goals might not match ours.
In his 2015 Reddit AMA, Hawking rejected the Hollywood picture of malicious machines. The real risk, he wrote, is competence. His analogy: you are probably not an evil ant-hater, but if you are in charge of a hydroelectric green-energy project and there is an anthill in the area to be flooded, too bad for the ants. “Let’s not place humanity in the position of those ants.”
This is the same idea that AI researchers now call the alignment problem. Hawking also raised two other concerns in his Lisbon talk: powerful autonomous weapons and new ways for the few to oppress the many.
The fear that a smarter-than-human AI could keep improving itself faster than people can respond is the core of the technological singularity idea, which runs through his warnings.
Was Stephen Hawking against AI?
Short answer: no. He saw huge benefits and called himself an optimist.
In his Lisbon talk, Hawking said it could help eradicate disease and poverty and undo damage done to the planet. The 2014 op-ed called the potential benefits “huge.” His point was that the outcome was not decided, and that it depended on doing the research and policy work in advance. That is why he helped launch Cambridge’s Leverhulme Centre for the Future of Intelligence, which studies exactly those questions.
How did Stephen Hawking use AI himself?
Short answer: his famous computer voice relied on predictive software that learned how he wrote.
Hawking communicated through a speech system that used predictive text from SwiftKey, which learned his vocabulary to suggest the next word. He made his most-quoted warning, in December 2014, in answer to a question about an upgrade to that system. Note the wording: his warning was about full artificial intelligence, not the narrow tools he relied on every day.
Was Hawking right about AI? His warnings in 2026
Short answer: much of what he warned about is now mainstream science, and some of it is starting to show up in lab tests.
| Hawking’s concern (2014–2018) | Status in 2026 |
|---|---|
| AI could surpass human intelligence | Lab leaders openly talk about building super intelligence within years |
| Competent AI with misaligned goals | Labs have documented models gaming objectives and faking alignment in tests; see superintelligence risks |
| “The few oppressing the many” | Concentration of AI power is a central policy debate |
| Need for research and rules first | Hinton, Bengio and 100,000+ others signed the 2025 Statement on Superintelligence |
Hawking was a physicist, not an AI researcher, and critics at the time said he overstated the danger. But the people who built modern AI, including Geoffrey Hinton and Yoshua Bengio, now make essentially his argument, and their p(doom) estimates show how seriously they take it. Others who share his concern, such as Elon Musk, are now racing to build superintelligence themselves.
Our take
Hawking’s lasting contribution to the AI debate was framing, not prediction. He moved the question from “will robots turn evil?” to “what happens when something more capable than us wants something different?” That is still the right question in 2026, and it remains unanswered.
Enjoyed this? Get the SI Briefing
One short email a week: what changed on the road to super intelligence, and why it matters. No spam, unsubscribe anytime.
Frequently asked questions
What did Stephen Hawking say about AI?
Hawking warned that full artificial intelligence could spell the end of the human race, and that AI would be either the best or the worst thing ever to happen to humanity. He argued the risk was AI becoming very competent at goals that conflict with ours.
What did Hawking say about AI before he died?
In his last major talk on AI, at Web Summit in Lisbon in November 2017, he said AI could be the worst event in the history of our civilization unless we learn to control it, but that it could also help eradicate disease and poverty. His final book, published in October 2018, repeated the warning.
Did Stephen Hawking think AI would destroy humanity?
He thought it could, not that it would. He described himself as an optimist and said the outcome depended on whether humans managed the risks through research and good policy.
How did Stephen Hawking use AI?
Hawking’s speech system used predictive text software from SwiftKey that learned his vocabulary, a basic form of AI. He made his famous 2014 BBC warning in an interview about an upgrade to that system.
6 Sources
- NBC News: Stephen Hawking says AI could be the end of mankind (Dec 2014)
- Hawking, Russell, Tegmark, Wilczek op-ed, The Independent (May 2014), reprint
- CBS News: Hawking Reddit AMA on AI (Oct 2015)
- University of Cambridge: “The best or worst thing to happen to humanity” (19 Oct 2016)
- CNBC: Hawking says AI could be worst event in civilization (6 Nov 2017)
- Notre Dame Magazine: review of Brief Answers to the Big Questions