Future perfect newsletter

Abstract

Researchers at OpenAI and others have built impressive machine learning systems that have gotten good at various tasks, such as image generation and playing video games. However, these systems are often inscrutable and their inner workings are not well understood, which could lead to problems as AI systems are applied to higher-stakes projects, such as steering policy or medical research. Anthropic, a new AI research organization founded by former OpenAI researchers, is focused on building tools to better understand AI systems and make them more safe and reliable. The company believes that this approach will have the broadest uptake among other AI researchers and the biggest positive impact on the world. – AI-generated abstract.