AI-blockchain convergence is probably much closer than it seems. Beneath the shallow tech-culture rivalry I’m seeing more and more weird complementarity. One strong signal is that there is a nontrivial contingent of mathematicians with serious chops in both.
Conversation
Replying to
Another is that there is hardware level kinship. Not an accident that GPUs are useful for both kinds of compute.
1
23
Makes sense because though there’s big $ in AI it’s all concentrated in big orgs or well-funded startups while crypto money is more distributed. That’s one reason I think the rivalry is shallow. At a technical/math level, it’s almost PB&J but the economic scenes do not harmonize.
Quote Tweet
Replying to @vgr
I feel like the ML people talk about crypto (negatively) significantly more than the crypto people talk about ML (that is it say, almost never)
1
14
Interesting point. In fact I’d make the stronger case that an AI safety problem is only real if it can be cast as a crypto rights/identity/data ownership type problem.
Otherwise it is some bullshit theology about basilisks, simulations, paperclips and such.
Quote Tweet
Replying to @vgr
The acceleration towards AI safety problems for which there are probable crypto solutions may force it.
3
1
16
Replying to
Have you been following the work 's Fraud and Defense team is doing?
They're using ML to defend against Sybil attackers who seek to game the quadratic funding mechanism used to determine partner matching amounts during grants rounds.
gov.gitcoin.co/t/introducing-
1
6
Show replies
Show replies
Replying to
To reach maximum potential, AI needs access to actually influencing society and pulling economic levers directly. It does this today on stock markets but this is a very limited interface.
Crypto gives AI a digital interface to everything
1
2
Eg. If an AI wants to hire people to carry out a murder, it can do so via smart contracts (yeah that’s scary)
Replying to
I feel like the ML people talk about crypto (negatively) significantly more than the crypto people talk about ML (that is it say, almost never)
1
1
9






