InfoNewsLLM-specific
Perplexity and Intelligence
- Published
- Record updated
Summary
OpenAI argues that perplexity, the average next-token prediction error on a held-out dataset, can serve as a shared metric for comparing pretraining progress across AI developers and informing industry safety standards. The post publishes OpenAI's internal perplexity measurements, which it presents as a first step toward shared pacing efforts. It links the log-perplexity thresholds of successive GPT models to emergent capabilities, including fluent conversation, chain-of-thought reasoning, and transformative capabilities in cybersecurity and mathematics below 0.8.
Related items
- InfoBuild vs buy for AI coding agent security: own the plumbing, buy the coherent pictureSame vendor · Adversa AI Blog
- InfoCan Anthropic afford to put safety first?Same vendor · The Verge (AI)
- Info‘I was shaking, crying, having trouble breathing’: the unchartered trauma of breaking up with AI companionsSame vendor · The Guardian Technology
- InfoCan AI really help spot Black Friday deals? We put it to the testSame vendor · BBC Technology
- InfoIn a year of AI solving maths problems, Anthropic may have just solved the biggestSame vendor · BBC Technology