Steven Pinker criticizes doomsday rhetoric about AI, calls for practical safety engineering
In an open letter on Quillette, psychologist Steven Pinker rejects scenarios of human extinction caused by AI and instead calls for safety engineering with independent oversight, responsibility, and human control.
Harvard psychologist Steven Pinker published an open letter on Quillette addressed to psychiatrist and influential technology blogger Scott Alexander, who had previously challenged him to a public debate on the existential risks of AI. Pinker declined the debate, describing such events as a "spectator sport" (a spectacle for an audience). In the letter, he argues that catastrophic scenarios such as the "paperclip maximizer" (AI that sacrifices humanity to produce paperclips) confuse intelligence with a desire to dominate and assume that AI systems would obsessively fixate on a single goal.
Pinker acknowledges that he previously underestimated how capable large language models would become and did not foresee how carelessly AI companies would release their products to market — while, according to him, this rush is paradoxically led by people from the same community that warns about AI risks. He identifies bioterrorism, cyberattacks, and uncontrolled AI agents as truly serious risks. According to Pinker, doomsday rhetoric does not solve these problems; what works is sober safety engineering built on independent oversight, legal liability, and human control.
Scott Alexander described his position on his blog back in June, estimating the risk of humanity being wiped out because of AI at 20 percent and advocating both better safety research and international agreements that would slow AI development.
Why it matters
The dispute illustrates a divide within the expert community between camps focused on existential risks and camps advocating concrete, verifiable safety measures (oversight, responsibility, deployment control). For companies working with AI, the claim by Pinker that AI companies release products carelessly is particularly relevant — strengthening the case for internal safety processes and testing before deployment, regardless of whether one believes in existential risk.
Two audiences, two different impacts
What this means
For individuals
The debate shows that even respected experts differ in their estimates of AI risks (Pinker vs. the 20 % estimate from Alexander), so when forming your own judgment about AI safety, it is advisable to consider multiple schools of thought rather than a single authority.
More practical updates →For a business
Companies deploying AI products face public criticism for releasing features carelessly without sufficient safety oversight; the recommended approach includes independent oversight, clear responsibility, and maintaining human control over AI agents.
Risks and complianceCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.