“ZK will be more useful in AI than crypto” — something I didn’t really believe until recently

But it will be used in ways people didn’t predict (not zkml or identity)

ヽ(⌐■_■)ノ♪♬ @ @gauntlet/🤖Ventures/tldr/Aera \\ ¯\_(ツ)_/¯ Struggling with the nature of randomness \\ main: @gaa \\ /red-bull /defi-research maxi

building [redacted] | surfing @ venice, ca | eating @ gjusta / gjelina | “lucky me, lucky mud” 🤠

be honest was this triggered by the tea app

Defi. Ethereum. Physics.

Data at @quotient.

Are you cools with @eulerlagrange.eth ? I think he's doing stuff with that right now.

Agentic finternet thesis: every website becomes an AI + ZK + crypto mini app.

AI watermarking shares a lot of overlap with ZK! Upcoming ZK Podcast on this soon™️

My guess, we saw a few web2-crypto integrations using ZKTLS. I forgot that Venmo-crypto swap app but maybe ZK allows agents to rapidly scale and quickly integrate any website.

Agentic Finternet Thesis is all of this is merged into one digital financial substrate.

100 $tipn

Democratising Quantum AI 

siriusquantum.vercel.app

The interesting thing about watermarks (and sort of why they are a "weak" version of ZK and/or IO) is there's no "PCP theorem" for them in the sense that you have to make extra assumptions for them to work (e.g. min entropy of output, cross entropy / correlation lower bound) that restrict the instances they work on

Yeah so that paper shows you can't make them fully removable to PPT adversaries; on the other hand, the recent coding theory watermarks have ZK-like guarantees conditional on the entropy of the output; this is like an "instance-based" backdoor through the watermarks in the sand paper of Barak, et. al

I assume you already seen the “watermarks in the sand” paper which tells us impossibility of watermarking for Gen Models and LLM especially for fake news and misinformation. I wonder how zk plays role here, since they can even attack private detection algos.

listener • reasoner • first principles maxi

I've been more interested in watermarking — which is encoding patterns in the randomness used to sample an LLM output so you can detect it is from a model not a human; it is sort of like a "weaker" ZK

Go on 👀 What are some of the ways people didn’t predict this is referencing? 

Is it related to privacy preservation by models that provide outputs pulling from disparate and distributed proprietary data sets maintained across the globe?

AI can show it followed the right rules without sharing the full model or data. This helps build trust in important areas like finance, hiring, or law.