For this article, you don’t need to be a computer scientist or a geek. I’m sharing my small tech analysis at a global level without diving into technical details.
The adoption of artificial intelligence continues to grow significantly. Today, many countries around the world now have an artificial intelligence plan (a real plan? Not really, but everyone wants to be at the forefront). This February, a summit on artificial intelligence is being organized in Paris, shortly after Donald Trump, the new U.S. president, announced Stargate, a vast $500 billion program. In this article, I will talk about all of this, the social implications of AI, and much more. But for now, we’ll discuss how AI will be the fuel for the massive adoption of zero-knowledge proof (ZKP) technology and its extensions, such as FHE, etc.
Introduction: ZK
To begin, ZKP technology, or ZK for short, is a technology that allows proving something without revealing it. I know you have many questions about how this works, but for now, we won’t delve into the mathematics behind this technology. So, please, take my word for it without asking too many questions (we’ll likely revisit this in the future). Thanks to this technology, I can have an argument, without revealing it, ask a verifier to check it, and, with a high probability (over 95%), ensure that my argument is true or false.
Currently, this technology is used in the expansion (or growth) of blockchains or to ensure privacy on the blockchain. One of the most striking uses is the Starknet technology, for example, a blockchain that allows scaling the Ethereum blockchain. By 2025, more people will use Ethereum less for classic transactions (fund transfers) and prefer to use rollups, which are cheaper and more secure. For some, Ethereum will remain only a verification layer, as its network is quite secure.
Speaking of Layer 2 solutions like Starknet, there is currently a lot of research to also scale Bitcoin, the most robust blockchain. Yes, it is very urgent to do this now, not only to reduce transaction costs but especially because quantum computing is getting closer.
Crypto Has Been the Fuel for AI
At Token2049 in Singapore, a speaker said that “crypto has been the fuel for AI.” I didn’t understand it at the time, but after thinking about it, I realized that the mining race has brought significant funding into GPU research, and indeed, NVIDIA and OpenAI have emerged as winners. Often, we overlook that technological innovation is not a singular phenomenon but an assembly of several converging technologies. LLMs couldn’t function without the capacity for storage, transmission, and processing of massive amounts of data. Once the technological barrier is crossed, multiple players can create similar technologies (Peter Thiel talks about horizontal innovation in Zero to One).
Photo by Elias Gamez: https://www.pexels.com/photo/person-holding-a-graphics-card-10558582/
Why AI Will Be the Fuel for ZK
After the Singapore event, I was discussing with a friend the real use cases for crypto: of course, many don’t have any, but perhaps we don’t fully grasp them either. Don’t worry, there are predictions that 60% of the jobs in 2030 don’t exist yet, and why would I understand the use cases of all these technologies when it’s hard to predict the future? What’s certain is that I wouldn’t have predicted the massive adoption of TikTok, as for me, YouTube was the ideal video format.
Predicting the future is about looking at the past and the present: for now, generative AI is causing a lot of damage: copying, identity theft, etc. While many joke around by impersonating others’ appearances, voices, or writing, malicious actors can use these tools for harmful purposes. This can range from simple scams to triggering a nuclear war. What’s certain is that many people underestimate the real power of artificial intelligence. Yesterday, I followed the French president saying that AI won’t be able to approach human emotions during a phone call or whatnot. But no, it will, and even better.
Photo by Andrea Piacquadio: https://www.pexels.com/photo/woman-in-grey-notched-lapel-suit-825982/
When I dial my mother’s phone number: am I talking to her or to a tool using artificial intelligence mimicking her voice?
When the military receives an order to attack, is it the commander who gave the order or a tool created by a teenager living in Nairobi or Goma (well, before, they would have been in Palo Alto, St. Petersburg, or Mumbai)?
To protect against identity theft, I believe that in the coming world, all our communication tools will need to have the ability to create proofs and verify their authenticity. I’ve been thinking about many tools to create, a project I’m currently working on, but one tool that has emerged is the Orb, which allows verifying humanity. The Orb by World (formerly Worldcoin), which is close to OpenAI, uses ZKML (zero-knowledge machine learning, which I won’t discuss in detail) to verify humanity. This device has a powerful chip capable of reading the iris, generating a proof, and thus creating your account on the World network. So, you are a human. Since the iris is unique to each human, it’s almost impossible to have duplicates. During a conference with some Worldcoin developers, I asked if we could use DNA, but there are many challenges with DNA legislation and other details, DC Builder replied. I’ve already been verified on Worldcoin, and I encourage you to do the same. For developers, explore World’s tools (App Store, APIs, etc.), as it’s the future.
Mass Adoption of ZK Tools
I’m convinced that in a few years, we won’t be able to handle AI tools anymore: I hope we won’t verge on paranoia due to the impersonations that could occur in our lives. Everyone will need to verify that the call received, the message received, or the email received comes from the real source. Thus, there will be a need for proof generation for every action. Today, end-to-end encryption reassures us, for example, on WhatsApp or Telegram.
In the future, we will need chat applications proving that we’re talking to the person we’re supposed to talk to or the robot we intended to talk to.
Why aren’t these tools on the market yet?
Well, some of these tools are gradually emerging, but there are several limitations: processors are still slow 🙂 Those who have the latest iPhone, Samsung, or Google Pixel think they have the fastest phone and that no more speed is needed. No, to generate a proof on my M1 Pro using SP1 or Triton, it takes me about 4 minutes for small proofs of nothing (Fibonacci, etc.). For more intense proofs: voice recognition, facial recognition, etc., it will take much more. For proof generation for every action on the phone, we need phones or portable devices that are maybe 200 times more powerful than current phones on the market.
Is It Possible to Achieve This?
I’m convinced we’ll get there: apparently, nothing is impossible for humans, and multiplying power by 200 isn’t too hard for manufacturers in the coming years.
Note that with the $500 billion investment from the Trump administration on Stargate, a large portion will go toward hardware research. The same goes for France ($100 billion) and perhaps much more in China and the Persian Gulf, which I love so much.
Well, many say Apple isn’t too keen on artificial intelligence, but I believe they’re trying to enable the execution of AI models locally on your devices: MacBook or iPhone, for example. Current models run on huge computers with very powerful GPUs. But apparently, DeepSeek has made progress in optimizing these.
Last year, Microsoft announced the NPU, a chip that will allow executing part of the models locally on Microsoft laptops.
With these advancements in local execution of models for more privacy, I’m convinced we’ll achieve a convergence of AI, ZK, and FHE. These combined technologies will create a balance between AI and humans and will reassure the future.
What will happen when quantum computing arrives? We’ll come back to that…
