Cory Doctorow: Reverse Centaurs (locusmag.com)

Cory Doctorow argues that the real question about AI isn’t what the gadgets can do but who they do it for and to — the social arrangements around a technology determine whether it empowers people or exploits them. He contrasts two concrete cases: a centaur-style use of the open‑source Whisper model running locally on a laptop to transcribe 30–40 hours of podcasts (speeding legitimate research and preserving user control), versus a “reverse‑centaur” journalism fiasco in which a Hearst/Chicago Sun‑Times summer guide listed 15 books, ten of which didn’t exist after a freelancer (Marco Buscaglia) apparently relied on AI and lacked adequate fact‑checking. The latter turned the human into an “accountability sink” or “moral crumple zone,” forced to supervise machine output at the machine’s pace with insufficient resources. The piece is significant because it reframes AI debates: technical capabilities (e.g., offline transcription, generative text) are neutral but business models, staffing, and verification practices determine outcomes like hallucinations, job displacements, or genuine augmentation. Key technical takeaways are the value of local/open models (privacy, durability), the persistent risk of generative “hallucinations” without rigorous fact‑checking, and the policy/labor implications—how design, editorial workflows, and regulation can enable centaur collaboration instead of exploitative reverse‑centaur arrangements.