Anthropic co-founder Jack Clark and economist Peter McCrory discuss advanced AI extinction risks and shrimp rights on Odd Lots

Great interview! Two things jumped out to me: 1) It's clear to those in leadership within Anthropic, etc. - and probably the gov't - that it makes no sense for all capabilities of all models to be available to everyone. We can democratize potential & minimize risk (preferably without all the drama on all side), as we've done before in other realms (physics -> nuclear weapons), recombinant gene therapy, and financial markets. Security clearances, access to capabilities or facilities are concepts that can be adapted to limit risk. We don't have to be so binary. And, as we've learned with social media (touched on in the interview), know your customer will need to be broadened to know your user & empower their guardians with real controls - especially vis a vis kids/teens and "autonomous" agents. No parent wants to live through the harm that Insta & Snap did to teens' minds in the 2010s. 2) @jackclarkSF's analogy that AI diffusion & economic impact may follow adoption of electricity in the early 20th century rings true. Except, that change didn't only increase worker productivity, it enabled wholly new kinds of products made in wholly different ways. So measuring impact in terms of time savings or conventional productivity metrics may miss the bigger shift, which is the ability for solopreneurs or small teams within big firms to create products and services (or eliminate layers of complexity) that until now were uneconomic in terms of time or ROI. Just as 20 years ago, cloud computing lowered upfront capital to start a company (no need to own racks and hire sys admins), going forward startups will be able to build offerings and find product-market fit (or fail) with less capital invested. Where electricity enabled huge factories and long-distance supply chains, AI + agent-enabled design and interoperability may have far more impact at the bottom (i.e. shifting the nexus of growth from large to small companies) while large enterprises go through a wrenching transition in how they work. That may show up as a kind of smudge in overall productivity stats, hiding beneath the surface, profound shifts in how stuff gets made.