Companiesslug: openai-safety-and-alignment

OpenAI Safety and Alignment

OpenAI 安全与对齐团队

OpenAI's internal team for model safety evaluation and alignment research.

Curated Intelligence

1 items · Page 1/1
OpenAI 官方博客T1·Models
78 pts
OpenAI Chief Scientist Proposes Perplexity as Shared Safety Metric

OpenAI Chief Scientist Jakub Pachocki argues that progress in modern AI is fundamentally progress in training better compressors, i.e. predictors, and proposes perplexity (cross-entropy loss) as a shared, cross-developer metric to inform common safety standards. As a first step, OpenAI discloses more transparency into its pretraining progress, and notes that recent models already show meaningful design contributions from AI itself.