TL;DR
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
Scott Alexander has published an open letter responding to Steven Pinker’s earlier letter about AI, after Alexander publicly challenged him to debate the subject. Alexander says he agrees that in-person debates can hinder truth-seeking, then argues that successive AI systems have shown increasing capabilities and that there is no clear basis for assuming progress will stop before reaching human-level performance. The supplied source excerpt does not include the full letter or Pinker’s complete response.
Scott Alexander has published an open letter to Steven Pinker addressing their disagreement over artificial intelligence and Pinker’s earlier response to Alexander’s public debate challenge. Alexander takes up arguments Pinker made about AI intelligence and the limits of extrapolating from current systems, while contending that successive GPT models have become more capable and that the point at which progress levels off is unknown.
Alexander, writing at Astral Codex Ten, says he challenged Pinker to a public debate after becoming frustrated with some of Pinker’s social media posts. Pinker replied with an open letter, according to Alexander, saying that in-person debates can be confrontational and unhelpful to truth-seeking. Alexander apologizes for any inadvertent offense but says he still considers Pinker’s public arguments about AI unsatisfactory.
Alexander says he will answer the substantive points Pinker raised, while making clear that his interest in a debate was not primarily to establish truth. He describes the exchange as a response to what he calls dishonorable writing, a characterization that is his own judgment rather than an independently established fact. The letter combines a critique of Pinker’s claims with personal criticism of how he has approached the public discussion.
The first argument Alexander addresses is Pinker’s account of intelligence. As Alexander summarizes it, Pinker objects to treating intelligence as a single quantity that can be extrapolated from animals to humans and then to superintelligent AI. Pinker’s position, in Alexander’s account, is that intelligent systems have particular strengths and limits, including limits imposed by what they can learn through observation and experimentation. Alexander counters that AI models have improved between generations and argues that future systems may continue to improve, though he acknowledges that progress could eventually level off.
The Dispute Over AI Progress
The exchange reflects a central disagreement in public debate about AI: whether improvements in specific systems support expectations of increasingly general capability, and what those expectations imply for potential risks. Alexander argues that improved performance from one model generation to the next makes it unreasonable to assume that capability gains will stop just before reaching human levels. That is his argument, not a demonstrated forecast of what future systems will achieve.
Pinker’s concerns, as presented by Alexander, caution against treating intelligence as a simple scale that can be projected across very different kinds of minds. The difference matters to readers because judgments about AI’s future depend partly on whether present-day advances are seen as evidence of broader, continuing capability or as gains that remain bounded by specific limitations. The letter lays out the disagreement but does not settle it.
The personal tone also shapes how the exchange is likely to be read. Alexander says he wants to answer Pinker’s substantive arguments while openly expressing anger about Pinker’s public remarks. That mix makes the letter both an argument about AI and a dispute over the standards of public intellectual debate.
Popular right now
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
De'Longhi Magnifica Evo Fully Automatic Espresso Machine with Manual Milk Frother, Silver
De'Longhi 15-Bar Pump Espresso Machine with Milk Frother & Temperature Control
As an affiliate, we earn on qualifying purchases.
How the Letters Began
Alexander says his debate challenge followed frustration with Pinker’s social media posts. He recounts that the challenge drew offers from podcasters, debating societies and other hosts before Pinker responded directly with an open letter. In that reply, as Alexander describes it, Pinker objected to in-person debates as a format that can be confrontational rather than conducive to inquiry, while denying Alexander’s accusations of personal attacks.
Alexander then turns to earlier disagreements about whether AI systems would overcome basic reasoning failures as they improved. He recalls a 2022 exchange involving Scott Aaronson and Gary Marcus, in which Marcus pointed to shortcomings in GPT-2 and GPT-3. Alexander says he argued that later systems would address some such failures, while Pinker was more skeptical. Alexander also cites a 2023 statement by Pinker expressing doubt that AI performance would improve exponentially. These historical claims and quotations are presented in Alexander’s letter; the supplied excerpt does not include the full earlier exchanges.
As evidence of changing capabilities, Alexander points to GPT-4 solving a simple addition problem, says GPT-5 handled harder problems, and refers to an internal OpenAI model beyond GPT-6 solving the Navier-Stokes equations. The excerpt offers no independent documentation, evaluation details or technical explanation for the internal-model claim, so those examples should be understood as claims made in the letter, not independently verified findings.
““Any intelligent system is a mechanism which is good at solving some problems but not others,” and is limited by knowledge it can gain through observation and experimentation.”
— Scott Alexander, paraphrasing Steven Pinker’s position
Limits of the Available Evidence
The supplied source is an excerpt of Alexander’s letter and ends during his discussion of research into cognitive abilities in animals. It does not include the rest of his argument, Pinker’s complete letter, or a reply from Pinker to Alexander’s open letter. The extent of their disagreement and whether they later exchanged further responses cannot be established from this material.
Several technical examples are asserted by Alexander without supporting documentation in the excerpt. In particular, it gives no details about the cited internal OpenAI model or how its reported Navier-Stokes result was assessed. The excerpt also does not establish that future GPT generations will continue improving at a particular rate, reach human-level intelligence, or pose a specific level of risk. Those are disputed forecasts, not confirmed outcomes.
Whether the Exchange Continues
Alexander says he plans to work through the arguments Pinker set out in his letter, beginning with the nature and measurement of intelligence. The supplied material does not state whether Pinker has responded to Alexander’s latest letter or agreed to a debate. For now, the next development depends on whether either author publishes a further reply; no debate date or host is confirmed in the excerpt.
Key Questions
What did Scott Alexander publish?
He published an open letter to Steven Pinker responding to Pinker’s earlier letter about AI and their disagreement over a proposed public debate.
Why did Alexander challenge Pinker to debate?
Alexander says he was frustrated by some of Pinker’s social media posts. He also writes that he did not propose the debate primarily as a way to seek truth, and criticizes Pinker’s public arguments in strong personal terms.
What is their main disagreement about AI?
As Alexander describes it, Pinker warns against treating intelligence as one scale that can be extrapolated from humans to AI. Alexander argues that AI systems have improved across generations and says there is no clear basis for assuming those gains will stop before reaching human-level performance.
Has Pinker agreed to a debate?
The supplied source does not say that Pinker accepted the challenge. Alexander reports that Pinker objected to in-person debates as confrontational and not conducive to truth-seeking.
Does the letter prove future AI will exceed human intelligence?
No. Alexander argues that future systems may continue improving, but the excerpt does not establish that they will exceed human intelligence or show when or how AI progress might level off.
Source: rss
