Ted Chiang: LLMs Are 'Sentence Continuation'
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Ted Chiang argued in The Atlantic that LLM conversations are "cleverly disguised examples of sentence continuation," while still acknowledging LLMs can be "impressive" — a nuance the louder reactions often miss.
- The essay singles out Anthropic, suggesting the company "really excels in anthropomorphism" rather than the technical achievements for which it's widely credited.
- The Verge covered a converging angle with "As AI gets better, it reveals an empty promise," signaling cross-outlet agreement that today's AI is being oversold even as it improves.
- Critics pushed back on the piece: Sam Z Liu said Chiang's argument "shows a lack of understanding of both how LLMs work and the long discourse in philosophy of mind," and Samuel Hammond wrote that his estimation of Chiang "has fallen off a cliff ever since he started writing about AI."
- Adrienne LaFrance urged followers to read the piece, and Sam Biddle amplified Chiang's analogy that debating LLM consciousness is like wondering whether "Microsoft Word is conscious."
- Justin Curl reframed the entire debate, arguing the consciousness question is really a "stand-in for other questions" about acceptable human-AI relationships and the effects of "always-positive" AI.
Why it matters: Chiang brings literary-authority weight to a debate usually dominated by AI insiders, and his essay landed as Anthropic and OpenAI face mounting pressure over claims of model sentience — giving policymakers and journalists a high-profile frame for treating AI welfare claims as anthropomorphism rather than evidence. The split reaction, with serious AI researchers publicly breaking with him, shows the technical community won't cede that ground quietly.
Ask SkimNews
