ZKML: the convergence of AI, Blockchain and Cryptography

Tatiana Revoredo
5 min readNov 13, 2023

--

The Diamond Age

There is a very good science fiction novel by Neil Stephenson, called The Diamond age, which talks about an AI tool known as The Illustrated primer — a type of artificially intelligent device that has the function of mentoring and teaching people throughout their lives.

Just as there are many situations today that seemed impossible fifty years ago and are now a reality, Neil Stephenson’s vision is also already coming true.

In the words of Isaac Asimov:

“Today’s science fiction is tomorrow’s science fact”.

A screen capture from the film 2001: A Space Odyssey

With this in mind, this article deals with the importance of the intersection between Blockchain, Artificial Intelligence and Cryptography.

How important is the intersection between these three technologies?

AI, Blockchain and Crypto as a natural counterbalance

AI is a technology that thrives on and enables “top-down” centralized control, while “public” blockchains and crypto are technologies that are all about “bottom-up” decentralized cooperation.

Blockchains and cryptography are the base technologies for building decentralized systems, which enable large-scale cooperation without the need for a centralized server.

Parallel to this is the boom in Artificial Intelligence affecting privacy.

In this sense, AI directly impacts all kinds of incentives that lead us to have less individual privacy.

In this sense, AI directly impacts all kinds of incentives that lead us to have less individual privacy.

This is why companies fiercely seek access to as much information as possible. Data is the fuel for training AI models.

If on the one hand, we have AI with the collective aggregation of all possible data for training these huge AI training models in order to make them as good as possible.

On the other hand, blockchain and crypto take us in the opposite direction — towards increased sovereignty, where users seek control over their own data in an attempt to preserve individual privacy.

That’s why AI and blockchain are each other’s counterweight.

AI is clearly a powerful technology, a creative tool that will lead us to an infinite abundance of media and creativity in many ways. Blockchain and cryptography, on the other hand, are a counterbalance to this, because they help us verify and distinguish between what is created by humans and what is created by AI.

Public blockchains and cryptography are essential for maintaining and preserving what is truly human in a world where millions of pieces of content are artificially generated.

However, although these three technologies are diametrically opposed in many respects, their convergence has many significant impacts, as we will see below.

The Age of Convergence and the intersection between technologies

Age of Convergence

Looking at the bigger picture, the techniques used in AI seem diametrically opposed to the techniques used in blockchains and cryptography.

After all, Blockchains are about cryptography, decentralization, consendo algorithms and tokenomics, while Artificial Intelligence is about statistics, machine learning mathematics and so on.

However, there are many points of intersection where one technology complements and helps to enhance the other — what is known as the Age of Convergence.

In this sense, there is already an emerging area of the application of cryptography to artificial intelligence, namely the use of Zero Knowledge proof [ZKp] in machine learning [ML], which has been dubbed “ZKML”.

The reason Zero Knowledge Machine Learning [ZKML] is interesting is that Zero Knowledge Proof [ZKP] cryptography techniques have improved dramatically because of their application in blockchains.

ZKML: Machine Learning and Privacy

Zero Knowledge Machine Learning

A decade ago, this idea of zero knowledge proofs and proof systems in general were considered very theoretical.

However, due to all its applications in blockchains, there was suddenly a big push to make uses of Zero Knowledge proof cryptography more practical and real and, as a result, there has been tremendous progress in this area. And now ZK is actually implemented and used to protect real systems. That’s what we’ll see next.

Examples where Zero Knowledge Proof is useful for Machine Learning

Imagine that you have spent a lot of time and resources training an Artificial Intelligence model, and now you want to verify that your AI model really works.

How would this verification be possible, while complying with the requests of a third party, without either the model’s secret or the third party’s sensitive data being revealed?

To put it another way, imagine that Catherine has paid for a certain AI model, and wants to be sure that Liam is really training that model.

Or maybe Catarina paid for the latest version of Chat-GPT and wants to be sure that Liam is really using the latest version.

ZKML: Preserving privacy in Machine Learning

In all of the above examples, Zero-Knowledge Proof [ZKp] cryptography would be very useful, as it makes it possible to validate a statement without revealing the underlying facts that make it true or false.

In this way, it would be possible to prove the efficiency of AI models publicly, without compromising their secrecy.

And whenever someone sent you data, you could run the AI model on that data and send the results back to the requester, along with proof that the model was evaluated correctly.

In the same way, Catherine would now have a guarantee that, in fact, the latest version of the AI model she paid for is the one she actually ran on her data.

Does all this sound far removed from your day-to-day life? Not really.

Have you ever thought about an AI model deciding whether you are entitled to a loan or a mortgage?

A financial institution might want to use an AI model like that and, of course, it would be important to prove that the model applied to all customers is the same as the one applied to you — without any kind of discrimination.

We are still in the early stages of using ZKp in Artificial Intelligence and Machine Learning [ML] models. But the important thing here is to understand how and in what situations the intersection between these technologies can lead us into the Age of Convergence.

Sharing knowledge generates prosperity! I wish you a great week! See you soon!

👉🏼 If you like my content, leave your 👍, share ♻️ with your friends.

🔔 Subscribe to my page if you don’t want to miss any posts/articles.

--

--

Tatiana Revoredo

LinkedIn Top Voice | Blockchain | Web3 | Technology & Innovation | Oxford Blockchain fdn • #2Top50 Cointelegraph Br