Stanford report gives failing transparency grades to major AI models

Stanford researchers grade major AI models low on transparency, with top score only 54% on their index.

Report examines models from OpenAI, Google, Meta, etc. on criteria like training data disclosure.

Calls for more openness as foundation models are incorporated into apps and services.

Ranks Meta's Llama 2 highest at 54/100, Amazon's Titan lowest at 12/100, OpenAI's GPT-4 at 48/100.

Aims to spur companies and policymakers to address AI model transparency and accountability issues.