AI Safety Insights
Vitalik discusses the challenges of AI safety, highlighting the informal nature of current AI practices that rely heavily on trial and error rather than scientific rigor. He points out that this lack of formalism can hinder our understanding of potential failures in AI systems. Additionally, he raises questions about the scalability of models like GPT-3 and whether increasing their size will lead to improved intelligence.In this clip
From this podcast
Lex Fridman Podcast
Vitalik Buterin: Ethereum 2.0 | Lex Fridman Podcast #188
Related Questions