Amanda Askell
Showing 1–2 of 2 transcripts.
- Lex Fridman5h 15m
Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast #452
Dario Amodei, Lex Fridman, Amanda Askell, Chris Olah
Anthropic leaders Dario Amodei, Amanda Askell, and Chris Olah outline a trajectory where AI models achieve PhD-level professional competence by 2026 or 2027 through continued scaling laws, necessitating a rigorous Responsible Scaling Policy with activation levels that enforce safety checks against catastrophic misuse and deception. The speakers detail a dual strategy combining Anthropic's public sharing of mechanistic interpretability research with Character Training via Constitutional AI to instill virtues like honesty and wisdom rather than mere rule compliance. This approach aims to prevent a regulatory race to the bottom by establishing uniform safety standards while leveraging AI to accelerate scientific discovery and solve global challenges within the coming decade.
- 80,000 Hours2h 54m
Can policy keep up with AI advances? | OpenAI's Amanda Askell, Miles Brundage, & Jack Clark (2019)
Amanda Askell, Miles Brundage, Jack Clark, Niel Bowerman, Michelle Hutchinson, Rob Wiblin
OpenAI researchers and policy experts convened to discuss a strategic framework that balances scientific transparency with safety, exemplified by the partial release of the GPT-2 model to prevent malicious applications like autonomous warfare and large-scale disinformation. The panel emphasized that short-term governance challenges mirror long-term alignment issues, proposing collaborative incentives and institutional mechanisms to avoid harmful development races while engaging with military sectors to enforce safety constraints. To accelerate these goals, speakers outlined critical career pathways for policy professionals, including newsletter creation and technical translation, while expressing optimism that robust coordination can harness AI to solve global challenges like poverty and disease.