Vitalik wrote a proposal to teach you how to sneak in on a big AI model
Original title:Vitalik wrote a proposal to teach you how to sneak in on a big AI modelIt's the only thing that's going on
Original language: Deep tide TechFlow
THE WORLD IS TALKING ABOUT AI, AND THE TIME LINE IS QUIET ABOUT ENCRYPTION。
At the same time, ETH was running around 2000 for almost two months, and Vitalik said what to do, and there wasn't much to care about。
BUT I RECENTLY TURNED OVER HIS X AND FOUND THAT IT WASN'T JUST US THAT WERE AFFECTED BY AI. A LOT OF WHAT HE SENT OVER THE PAST MONTH WAS RELATED TO AI, AND TO THE EXTENT OF TECHNOLOGY。
The most interesting of these is the proposal he sent jointly with Davide Cripis, Director of AI, Etherfield Foundation, on February 11th, entitled "ZK API Usage Credits"。
IN ONE SENTENCE, IT'S A ZERO-KNOWLEDGE DEMONSTRATION THAT GIVES YOU AN ANONYMOUS CALL TO THE BIG AI MODEL。
Now you can use ChatGPT, or Claude's API, and there's only one way to pay:
Registered accounts, mail boxes, credit cards。
every time you talk, every prompt, the platform knows you sent it. what did you ask, when, how many times, all tied to your true identity。
The Vitalik and Cripis proposals offer another way forward。
1. USERS CHARGE A SUM OF MONEY IN SMART CONTRACTS, SUCH AS 100 USDC。
ON A JOINT DATE, THE DEPOSIT WAS ENTERED IN A CHAIN-BASED ENCRYPTED LIST. EVERY TIME YOU SWITCH, YOU DON'T HAVE TO SHOW YOUR IDENTITY, YOU JUST HAVE TO PRODUCE A ZERO-KNOWLEDGE CERTIFICATE
3. It can prove to the service provider two things: you have enough on the list. But the proof itself does not reveal which one of you is on the list。
A service provider can receive money and prevent abuse, but never know who you are。
You can interpret this proposal as one thing, namely, Vitalik thinks that in the AI era, it's not the first time that we've seen itUSERS SHOULD NOT HAND OVER THEIR IDENTITY TO USE AN AI TOOL。
THE PROPOSAL IS STILL AT THE RESEARCH STAGE, STILL DISTANT FROM THE SITE, AND LARGE MODEL MANUFACTURERS MAY NOT AGREE WITH THIS APPROACH; AT THE SAME TIME, THE COMMENTARY ON THE PROPOSAL IS FULL OF REBUTTALS AND QUESTIONS THAT AI MODEL PLANT HAS ALWAYS FOUND A WAY TO KNOW YOUR TRUE IDENTITY。
However, in the view of the author, the significance of this proposal is not entirely due to its own landing。
Privacy is what Vitalik has done for 10 years. From early support to Tornado Cash to the promotion of zero knowledge as a core technology route for the Ether House, this line has never been broken. Only in the past few years, privacy has been in the context of the encryption industry, with a sufficiently big story to take on it。
AI MADE UP THIS STORY。When you say more to a big model every day than anyone, privacy is a real need。
Vitalik hug AI
From February to now, a significant portion of Vitalik's X-shots are related to AI, so dense that they don't talk。
Yesterday he sent a long post saying that he had recently gone to a meeting on cryptography, where people were concerned about privacy, about open sources, about resistance to censorship... but had no feelings for the chain。
Among those people, he did a thought experiment:
Forget "We're a Taipei community" and start from scratch, where the ETA is most useful。
His conclusion was that the lowest value of the Taifeng was to be used as a bulletin board. A place where anyone can write, anyone can read, no one can change, no one can delete。
In the context of AI, this may be the most important sentence Vitalik has said in these two years。
WE ARE ENTERING AN ERA OF INFINITY. WORDS, PICTURES, VIDEOS, IDENTITIES, AND AI CAN BE CREATED IN BULK. WHAT BECOMES SCARCE WHEN EVERYTHING CAN BE FORGED
They all end up in the same place: an open, durable and irreversible data layer. And a record that no one can alter is exactly what the Taifeng can do。
Over the past two years, the challenge facing the Taifeng has come down to a sentence: What are you going to be
Now look, Vitalik didn't answer the question head on。
However, it has done a few minor things over the past year: a 50-person privacy team, a nearly 50-person privacy research cluster, the Kohaku privacy framework, with a dedicated AI director, was launched, and institutional privacy and faster transaction recognition were given top priority in the 2026 Road Map。
LOOKING BACK AT HIS INTENSIVE OUTPUT THIS MONTH, HE BASICALLY DISCUSSED THE PRIVACY AND EFFICIENCY OF THE TAIFENG IN THE CONTEXT OF AI。
I think Vitalik is gambling on one thing:THE STRONGER AI, THE MORE RIGID THE NEED FOR PRIVACY AND CERTIFICATION INFRASTRUCTURE. IT'S ANOTHER MATTER WHETHER THE ETHER HOUSE CAN CATCH UP TO THIS DEMAND, BUT HE'S OBVIOUSLY CHOSEN THE TABLE。
ETH'S STILL AROUND 2000. MOST PEOPLE STILL DON'T CARE WHAT HE SAID RECENTLY。
But perhaps in a few years' time, it is time to look back。
