In this episode, Hutch and Len talk about recent alignment research conducted on frontier AI systems. This includes discussing recent incidents in the news, as well as discussing contents of the recent Claude 4 system card released by Anthropic.Links:- Learning to code is as valuable (in terms of job prospects) as getting a face tattoo (https://futurism.com/risk-expert-learn-to-code-face-tattoo)- Elon Musk concerned that reality has infiltrated Grok (https://futurism.com/elon-musk-infilitrated-grok-ai)- Israel-Iran conflict unleashes wave of AI disinformation (https://www.bbc.com/news/articles/c0k78715enxo)- OpenAI o3 Model Refuses to Shutdown (https://www.theregister.com/2025/05/29/openai_model_modifies_shutdown_script/)- OpenAI o3 and o4-mini System Card (https://openai.com/index/o3-o4-mini-system-card/)- Anthropic Claude 4 System Card (https://www-cdn.anthropic.com/6be99a52cb68eb70eb9572b4cafad13df32ed995.pdf)- Is AI Apocalypse Inevitable? - Tristan Harris (https://www.youtube.com/watch?v=86k8N4YsA7c)

Ver todos los episodios