He lands in favor of redistribution-like regulations that benefit newer startups and «those catching up,» while disadvantaging the more powerful frontier labs.
Open source models should also be vetted once they reach a certain level, as the White House has agreed to. Open source brings «specific risks,» he argues, although he doubts they will solve the distribution or concentration problem.
On AI’s generally negative public image lately, Amodei says this is fundamentally an issue of trust:
— I think that ordinary people don’t trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over, he writes.
Curing cancer is a cliche, but doable
He contends that there have been a lot of warnings from him and others, that also balance the risk with the benefits, but these talks get clipped and appear overly negative on social media and news channels.
This won’t be solved by «a glitzy marketing campaign,» he says, but rather by delivering on their much touted benefits.
— Saying that AI will cure cancer is more a cliche than it is inspiring, and most people think it is deceptive. The thing that will work is *actually curing cancer,* is Amodei’s conclusion.
In the text, he lays out a timeline of AI-enabled research curing most human disease in ~5-10 years, and has long asked for streamlining the FDA approval process for the AI process.
— Anthropic is ramping up its efforts very quickly in biology and medicine, and we hope to have incredible results in the coming years and some early glimmers in the coming months, he promises.
Read the full essay on X. Writeups on TechCrunch and Business Insider. Discussion on r/Singularity.