TL;DR
Get school and study supplies delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
Scott Alexander has published an open letter responding to Steven Pinker’s earlier letter about AI, arguing that successive language models have shown increasing capability and that there is no clear basis for assuming progress will stop below human-level performance. The exchange is an argument between two writers, not a report of new technical findings; the supplied material does not establish whether Pinker will reply or take part in a debate.
Scott Alexander has published an open letter answering Steven Pinker’s arguments about artificial intelligence, disputing Pinker’s account of intelligence and saying that recent language-model progress offers little reason to assume future systems will stop short of human-level abilities. The exchange matters because it brings a prominent disagreement about AI progress and risk into public view, but it does not resolve the underlying technical questions.
Alexander’s letter follows an earlier open letter from Pinker to Alexander. Alexander says he had challenged Pinker to a public debate after becoming frustrated with Pinker’s social-media posts, but acknowledges that the debate proposal drew an excessive response from potential hosts and says he felt bad about the resulting pressure. He says Pinker replied in a reasonable and charitable manner and invited him to address the arguments directly.
Alexander takes up four arguments that Pinker set out, beginning with Pinker’s criticism of what he describes as an underdeveloped view of intelligence: the idea that intelligence is a general quantity that can be extended from animals and humans to AI and, eventually, superintelligence. Pinker’s argument, as Alexander presents it, is that intelligent systems are capable at some tasks but limited by what they can learn through observation and experimentation.
Alexander counters that, in ordinary usage, later generations of GPT models are more intelligent than earlier ones, and argues that the default expectation should be continued improvement unless there is a compelling reason to expect progress to stop at a particular point. He points to past model performance on arithmetic and other tasks, and cites an internal OpenAI model that he says solved the Navier-Stokes equations. The supplied material does not independently verify that claim or provide technical details about the model or result.
The Dispute Over AI Progress
The disagreement concerns more than how to label software. If AI systems continue to improve across a range of tasks, that could affect assessments of their future capabilities and the risks associated with deploying increasingly capable systems. Alexander’s central challenge is that critics of AI-risk scenarios should explain why progress is likely to stop before reaching human-level performance, rather than assume a ceiling without evidence.
That is Alexander’s argument, not a settled finding. Pinker’s position, as summarized by Alexander, emphasizes that problem-solving ability has limits tied to knowledge and experience. The exchange gives readers a clear example of how disagreement over the meaning and measurement of intelligence shapes broader arguments about AI risk, while leaving the practical and scientific questions open.
As an affiliate, we earn on qualifying purchases.
How the Open Letters Began
The letters followed a public disagreement that Alexander says began with Pinker’s social-media posts about AI. Alexander proposed a debate, but Pinker objected to in-person debates as confrontational and poorly suited to truth-seeking, according to Alexander’s account. Pinker instead published an open letter laying out arguments against some AI-danger claims and inviting a written response.
Alexander says he accepted that invitation and chose to answer the substance of Pinker’s points, though his letter also includes personal criticism and jokes about the debate proposal. The first part addresses Pinker’s view of intelligence and discusses earlier disputes over whether language models would overcome limitations seen in GPT-2 and GPT-3. Alexander says he predicted later systems would improve at tasks those earlier systems struggled with, including basic arithmetic.
“At some point, this intelligence will exceed human intelligence.”
— Scott Alexander, in his open letter
Claims Still Need Verification
The supplied text covers only part of Alexander’s letter and does not include Pinker’s full open letter, so readers cannot assess every point in the exchange from this material alone. It also does not provide independent evidence for Alexander’s account of model performance, including the reported Navier-Stokes result, or specify what the claimed result involved.
It remains unclear whether Pinker will answer Alexander’s latest letter, whether either writer will agree to a public discussion in another format, and how they define or measure intelligence across different systems. The exchange itself supplies arguments and recollections, not a peer-reviewed assessment of AI capabilities or a resolution of AI-risk questions.
Whether Pinker Responds
The immediate next development would be a further response from Pinker or a decision by the writers to continue the discussion in writing. Alexander says he will address Pinker’s arguments before returning to the question of whether to debate, duel, or seek a synthesis of their views. The supplied material does not confirm that any debate has been scheduled.
For readers following the dispute, the most useful next evidence would include Pinker’s full response, clear sourcing for the technical examples raised by Alexander, and a more precise account of what each writer means by intelligence and human-level performance. Until then, the exchange remains a public disagreement rather than a conclusion about the direction or limits of AI development.
Key Questions
What is the news?
Scott Alexander published an open letter responding to Steven Pinker’s arguments about AI progress and risk.
What is Alexander’s main argument?
Alexander argues that newer AI models have generally improved on earlier ones and says there is no compelling basis for assuming that improvement will stop below human-level ability. That is his argument, not an established forecast.
Why did Pinker decline the proposed debate?
According to Alexander, Pinker said that in-person debates can be confrontational and are not well suited to truth-seeking. Pinker instead published a letter setting out arguments for Alexander to answer.
Has Pinker responded to Alexander’s new letter?
The supplied material does not say whether Pinker has replied. It also does not confirm that a debate or another public discussion has been arranged.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
