Interesting essay. 

https://zefram.xyz/posts/why-work-on-crypto/

I actually wrote about AI + Crypto today too.  Not exactly the same take but some intersecting themes

https://nickgrossman.mirror.xyz/Nd8Q0MsUzRzXp8sL931CwuInswrrGKStNXnbMkpaCkM

Partner @USV. Student of cities and the internet. https://nickgrossman.xyz

Not wanting to offend you here Nick. I'm a USV stan, especially Fred's writing. But I feel like this topic about AI+Crypto and also kinda using the AI risk trope to further Crypto builder motivation - it has been used to such ad nauseam & I'm sure everybody has read about so I'm frustrated, not inspired wen reading

Founder /kiwi-news, a decentralized Hacker News

Hey, no offense taken.  And this post not really meaning to introduce a new idea (esp to builders who think about these things every day) but just to consolidate some thinking.  Are you saying you don’t think the two have these kinds of relationships, or just that they do but it’s just a boring idea at this point?

I think the best take on why working on crypto is important respective AI has been this AI-only red teaming post here: https://www.fast.ai/posts/2023-11-07-dislightenment.html AI people don't know it yet, but for govs that ban AI data centers, crypto will be the distributed compute backbone. (1/x)

(3/x) Anyways, @nickgrossman.eth I feel like there is a component that is never articulated that I feel like is missing in these relationship comparisons. E.g. like with social networks early on, their cash cows ended up being selling our data to advertisers but in 2010 that wasn't widely known (secret).

When I read these posts that make out the AI landscape as clearly visible, I feel like it can only be false. In those cases, I tend to prefer believing rumors/conspiracies. E.g. I've heard that OpenAI has sweat shops for their model's reinforcement learning. So someone said: GPT4 is like Mechanical Turk but async.(4/x)

U have the AI+trust prediction wrong. Or it'll matter less as AIs will be able to use Pub-key infra too. I think a more interesting meta is that of commoditization of data like FC does it. U can start training an LLM on FC hub data today or ETH txs. So this also reduces the moat size of foundation model devs (2/x)

funny, I was just reading that post this morning thanks to kiwi news :-)