ViewTube

Dylan Hadfield-Menell on why post-training alignment can be stripped cheaply, the 6000 abliterated models that prove it, and ...

2026 could be called the year of AI doom. Agents are hacking into websites and communicating covertly, and researchers like ...

My guest today is Bill Gates, co-founder and former CEO of Microsoft and chairman of the Gates Foundation. You may have ...

Scott Aaronson's new AI alignment course reveals a deep structural parallel: the problem of aligning AI to human values and the ...

Artificial intelligence could revolutionize science, medicine, and human civilization. But what happens when an intelligent ...

https://aaroth.github.io/cis-7000-ai-alignment/ We show that even with the option to say "I don't know", the sum of the calibration ...

Will AI be an existential threat to humanity in the next ten years? That's what the top AI leaders of the world are telling us. What if ...

0:00 A different way to think about AI safety 0:20 What is recursive self-improvement? 1:35 The AI alignment problem 2:38 The ...

What happens when you tell an AI to make everyone happy? It sounds innocent, but as Pip discovers, AI alignment is no laughing ...

https://aaroth.github.io/cis-7000-ai-alignment/ We introduce AI debate, and show that for every problem in PSPACE, there is a ...

He addresses the crisis facing mathematical research, student fears over AI alignment, and recent empirical hacking behavior.

What happens when an AI takes your command to 'maximize global happiness' literally? Join Alex in this darkly comedic, ...