"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis cover image

Nathan on The 80,000 Hours Podcast: AI Scouting, OpenAI's Safety Record, and Redteaming Frontier Models

"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis

NOTE

Safety record of self-driving cars and AI's performance in medicine

Self-driving cars have demonstrated an impressive safety record with zero injuries out of over 3 million miles driven, compared to an expectation of over 3 injuries for humans and under 25% property damage ratio for the Waymo system. The AI system GPT 4-V has outperformed human respondents in medical image evaluation, except in radiology, demonstrating promise in multimodality. It utilizes text to improve accuracy on image challenges but may miss obvious diagnoses. The system's potential to provide second opinion diagnoses and facilitate discussions with human doctors is seen as a valuable addition to medical care. Despite the need for more studies, the ethical and consequentialist viewpoint supports providing access to GPT 4-V for medical purposes. Additionally, the GPT 4-V team's involvement with OpenAI has led to insights into the system's capabilities and potential economic impact. The speaker emphasizes the need for human-AI collaboration studies and views the system as a valuable medical tool.

00:00
Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner