As the Frontier AI labs have a massive incentive in protecting their own infrastructure themselves, would they pay for (and get) early access of each others models as well?
Would google for example pay for a preview of Claude. And OpenAI pay for a preview of Gemini?
Would this possibly even speed up the flywheel of AI development as they would be sharing information at an earlier stage?
Thanks Philipp, interesting thoughts. I believe this will could become the future business model: to create a (super) intelligence, to declare not being able to control the tool, while selling it to the highest bidder. The correct path should have been this: Do not let a tool or a virus or a potential weapon leave the laboratory before you can control it.
Private or public, there are rules and guidelines for advanced technical research in place. Only AI firms seem to ignore them. Imagine a pharmaceutical company releasing a virus while selling a few others the antidote.
There’s a lot of debate about whether technology this powerful should be controlled by a private company or whether the “public,” meaning governments, should own it instead.
Anthropic has both an economic and reputational interest in making these advancements public. Governments (and their intelligence agencies) have a proven track record of using it in their own interest: postpone and exploit the technological advances before making it public and using them to spy on enemies, allies, and sometimes their own citizens.
So private ownership, combined with strong and independent regulatory oversight, might be the least bad solution.
As the Frontier AI labs have a massive incentive in protecting their own infrastructure themselves, would they pay for (and get) early access of each others models as well?
Would google for example pay for a preview of Claude. And OpenAI pay for a preview of Gemini?
Would this possibly even speed up the flywheel of AI development as they would be sharing information at an earlier stage?
Interesting thoughts!
I'd love to get the newsletter in English, and I'd prefer the Doppelgänger Podcast in English as well.
Thanks Philipp, interesting thoughts. I believe this will could become the future business model: to create a (super) intelligence, to declare not being able to control the tool, while selling it to the highest bidder. The correct path should have been this: Do not let a tool or a virus or a potential weapon leave the laboratory before you can control it.
Private or public, there are rules and guidelines for advanced technical research in place. Only AI firms seem to ignore them. Imagine a pharmaceutical company releasing a virus while selling a few others the antidote.
Very interesting!
There’s a lot of debate about whether technology this powerful should be controlled by a private company or whether the “public,” meaning governments, should own it instead.
Anthropic has both an economic and reputational interest in making these advancements public. Governments (and their intelligence agencies) have a proven track record of using it in their own interest: postpone and exploit the technological advances before making it public and using them to spy on enemies, allies, and sometimes their own citizens.
So private ownership, combined with strong and independent regulatory oversight, might be the least bad solution.