An AI system will help produce a Nobel prize-winning discovery within 12 months, Anthropic co-founder Jack Clark told an Oxford University audience this week.
Key Points:
- Clark predicts AI and human researchers will co-produce a Nobel-level discovery inside a year, with companies run entirely by AI generating millions in revenue within 18 months.
- He warned the technology still carries a non-zero chance of wiping out humanity, calling that risk one Anthropic refuses to downplay.
- By the end of 2028, Clark expects AI systems capable of designing their own successors.
Oxford Lecture Predictions
Clark laid out the timeline in a lecture at Oxford University on Wednesday, in remarks reported by The Guardian.
He described a "vertiginous sense of progress" across the field.
His forecasts stacked up quickly. He said bipedal robots would assist tradespeople within two years, and that companies operated solely by AI systems would generate millions of dollars in revenue within 18 months.
By the end of 2028, he expects AI to design its own successors.
Clark also kept the warnings front and center. He said plausible scenarios remained in which the technology held a non-zero chance of killing everyone on the planet, and stressed that the risk had not gone away. He pointed to Anthropic's restricted Claude Mythos model, which proved capable of exploiting cybersecurity weaknesses.
Also Read: XRP Loses Key Support, Now Eyes A Drop Toward $1.31
Why Clark's Warning Matters
Clark said he would prefer that humanity slow the technology down to buy more time, but he does not expect that to happen. He blamed a race between rival companies and countries where commercial and geopolitical rivalries drown out existential concerns.
Anthropic, founded by researchers who left OpenAI over safety disagreements, has been accused by the Trump White House and other accelerationists of fear-mongering to encourage favorable regulation. The company rejects that charge.
Prof Edward Harcourt, director of the Institute for Ethics in AI, which co-hosted the lecture, raised a separate concern. He warned that AIs doing more for humans risked "cognitive atrophy" that could weaken human judgment, and he argued for so-called Socratic AI models that ask people to do more of the thinking themselves.
The prediction is not the first time Anthropic leaders have set aggressive timelines. Chief executive Dario Amodei said in late 2024 that AI smarter than a Nobel laureate across most fields could arrive as early as 2026, and Clark himself has previously placed that milestone in late 2026 or early 2027.
Read Next: Ethereum Network Empties Out As Staking Locks A Record 32% Of Supply
