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 arguments that advanced AI may be less dangerous than some critics fear. Alexander disputes Pinker’s account of intelligence and points to past disagreements over whether successive GPT models would improve, but the supplied text does not establish that Pinker has replied to this latest letter or that a debate is planned.
Scott Alexander has published an open letter to Steven Pinker disputing Pinker’s arguments about artificial intelligence, including whether intelligence can be treated as a capacity that grows across increasingly capable AI systems. The exchange follows Alexander’s public challenge to debate Pinker; Alexander says he will address Pinker’s reasoning in writing, while arguing that successive GPT models have shown why expectations about AI capability should remain open.
Alexander says his debate challenge followed frustration with Pinker’s social media posts. He writes that Pinker answered with an open letter of his own, rejecting Alexander’s accusations of personal attacks and setting out arguments for why AI may be less dangerous than some critics believe. According to Alexander, Pinker also said in-person debates can feel confrontational and may not help people seek truth. The supplied source contains Alexander’s reply, not Pinker’s original letter in full.
Alexander’s first substantive response concerns the meaning and limits of intelligence. He argues that recognizing improvement between model generations does not require treating intelligence as a single, mysterious substance: in his view, it is enough to observe that later systems can solve problems earlier systems could not. He cites the progression from GPT-2 and GPT-3 to later models, saying that systems have improved on tasks that once exposed weaknesses, including elementary arithmetic and more difficult problems.
To support that argument, Alexander recalls earlier public exchanges in which Pinker, according to Alexander’s account, questioned whether language models would overcome particular reasoning failures. Alexander says Pinker predicted in 2019 that AI progress would not reach the capabilities some expected, and in 2023 expressed doubt that improvement would be exponential. Those characterizations and quotations are presented by Alexander; the excerpt does not independently establish the full wording or context of Pinker’s past statements.
The Dispute Over AI Progress
The exchange highlights a consequential disagreement in public discussion of AI: whether improvements in model performance justify expecting further advances, and whether that trajectory should influence judgments about future risks. Alexander’s argument is that observers should not assume progress will stop at current or near-human capabilities without a persuasive reason. That is his position, not evidence that future systems will reach any particular level.
Pinker’s reported counterargument, as summarized by Alexander, emphasizes that intelligence is problem-specific and constrained by what a system can learn through observation and experimentation. That distinction matters to readers because broad claims about AI becoming more capable can obscure uncertainty over which tasks improve, how reliably, and what limits remain. The letters make those assumptions visible, but the supplied material does not resolve them or provide a shared assessment of the risks.
The episode also shows how debate over AI safety can become personal even when participants say they value civil inquiry. Alexander apologizes for the pressure created by his debate challenge, while also criticizing Pinker’s public arguments in unusually combative terms. The written exchange offers a record of their disagreement; it does not show that either has accepted a debate or changed position.
As an affiliate, we earn on qualifying purchases.
How the Exchange Began
Alexander’s letter says he challenged Pinker to a public debate earlier in the month after taking issue with Pinker’s social media posts. He describes receiving offers from podcasters, debating groups and other organizers to host, then says he recognized that the attention could amount to unwanted pressure before Pinker had accepted. The letter’s humorous account of increasingly elaborate efforts to get Pinker’s attention is Alexander’s own description, not independently verified reporting.
Pinker later published an open letter addressed to Alexander, according to the response. Alexander says Pinker apologized for any offense caused by his posts, denied making personal attacks and laid out several arguments against claims that AI poses severe dangers. Alexander says he will answer those points in sequence. The supplied excerpt develops his response to the first argument and begins a discussion of studies examining whether cognitive abilities are correlated in animals.
Alexander invokes research on rats, primates and birds to question whether a general pattern of linked cognitive abilities is unique to humans. The excerpt ends before he completes that discussion or addresses the rest of Pinker’s arguments. It also does not provide publication dates for the earlier exchanges, model comparisons or cited studies, limiting what can be concluded about their precise chronology.
“I agree that in-person debates are confrontational and bad for truth-seeking.”
— Scott Alexander, writing to Steven Pinker
As an affiliate, we earn on qualifying purchases.
What the Letters Do Not Resolve
The available source is an excerpt of Alexander’s response. It does not include Pinker’s full open letter, the complete set of arguments Alexander says he will answer, or the remainder of Alexander’s discussion of animal cognition. Readers therefore cannot assess the full exchange from this material alone.
It is also unclear whether Pinker has responded to this latest letter, whether either writer intends to continue the exchange, or whether a public debate will take place. Alexander’s letter describes past predictions and model capabilities, but the excerpt supplies no independent benchmark data, detailed study citations or common measure for comparing the systems. Its claims about the pace and eventual limits of AI improvement remain disputed arguments.
As an affiliate, we earn on qualifying purchases.
Whether the Debate Continues
The immediate next step depends on whether Pinker responds and whether Alexander publishes the rest of his planned replies. Alexander says he intends to address Pinker’s arguments in order, but the supplied excerpt does not confirm a publication schedule or any agreement to debate.
For now, the public record described here is an exchange of open letters, with Alexander challenging Pinker’s account of AI progress and Pinker’s full reasoning available only through his separate letter. Further statements from either writer, or a transcript of a debate if one is arranged, would clarify how they address the evidence and the limits of their competing claims.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the open letter about?
Scott Alexander’s letter responds to arguments Steven Pinker made about AI, including how intelligence should be understood and whether future systems may become more capable. Alexander also revisits their earlier disagreements about language-model progress.
Did Steven Pinker agree to debate Alexander?
The supplied material does not say that Pinker accepted the debate challenge. Alexander says Pinker responded with an open letter and that he plans to answer its arguments in writing.
What does Alexander argue about AI intelligence?
Alexander argues that later generations of AI can improve on earlier ones without requiring intelligence to be treated as a single, mysterious quantity. He expects capability to continue advancing, but the letter’s prediction is not a confirmed outcome.
What does Pinker argue, according to Alexander?
As Alexander summarizes it, Pinker views intelligent systems as capable of solving some problems but not others, with their knowledge limited by what they can learn through observation and experimentation. The full wording and context are not included in the supplied excerpt.
What remains unknown about the exchange?
The excerpt does not establish whether Pinker will reply, whether a debate will be organized, or how Alexander addresses the rest of Pinker’s arguments. It also does not independently verify the historical claims and model comparisons Alexander discusses.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
