Hallucinations cannot be solved, argues Burkov: a solvable model would be untrainable
burkov · x · 2026-10-10
Responding to whether the hallucination problem has been solved, Andriy Burkov argues no—and it can't be. His reasoning: if hallucinations were solvable by the model alone (without external data like web search at inference), the model would become impossible to train. Probabilistic outputs can't guarantee factuality within the training objective itself; the real fix lies in external verification sources.
More from Models
- Liquid AI's decision model d1 lands on Vercel AI Gateway with vision support — maximelabonne · 2026-10-10
- Open TTS Leaderboard adds Paradee-8M, a Kokoro-82M distill matching WER at 1/10 params — realmrfakename · 2026-10-10
- Kimi gateway latency test ranks GitHub first, Neon second, ngrok third — mariorod1 · 2026-10-10
- Grok in group chats is 'really nice', but users note messages are no longer private — Angaisb_ · 2026-10-10
- Grok bot buys on Amazon first try while shopping agent Muse fails twice — jamesperkins · 2026-10-10
- Why long AI chats get expensive: full-history resends and brittle prefix caching — ClickOk5811 · 2026-10-10