Marc Andreessen on AI, tech and politics in 2025
Marc Andreessen discusses artificial intelligence, technology policy and Silicon Valley’s political realignment in a long interview. He argues that AI alignment is chiefly a question of human values, criticizes the use of human-feedback training to impose ideology, and favors competition among AI systems. The conversation also covers his techno-optimist manifesto, social-media censorship, employee activism, DEI programs and employment law.
Passages
25:10 “So I think the single biggest fight that has ever happened over technology, and there have been many of those fights over the course of the last, especially 500 years, the single biggest fight is going to be over what are the values of the AIs to your points, what will the AIs tell you when you ask them? Anything…” Share
34:36 “And RLHF is basically a key step for making an AI that works and interacts with humans, which is you take a raw model which is sort of feral and doesn’t quite know how to orient to people, and then you put it in a training loop with Some set of human beings who effectively socialize it. Reinforcement learning for…” Share
1:16:58 “And then look, this movement that we now call Wokeness, you know, it hijacked what I would call sort of at the time box standard progressivism, which is of course you want to be diverse, and of course you want to be inclusive, and of course you want everybody to feel included, and of course you want to be kind,…” Share
Search what Andreessen said from Claude, ChatGPT or your own code. See plans