Okay guys, now we have another good friend, exquisite visionary. Please join me in welcoming Jack Hidary on stage. Jack, please join us.
Mayor Jack, I know. Tell me about it. And Jack always likes to make an entrance. Here we go. We'll put you center stage.
Thank you. So I want to first tell us I want to congratulate you and the whole Monaco team on a fantastic Monaco House this year. Last year we had a lot of great energy for Monaco, but we were squeezed into a small jewel box, and this year we have proper space and wonderful.
You know what happened there, Jack? We hadn't told them you were coming, so they only gave us a small space this season. I'm sure we said Jack is going to be with us. Listen guys, Jack runs and is the CEO of SandboxAQ. He's at the forefront of AI and quantum. He's a great author. By the way, for those of you who haven't read his book, AI or Die, read it or you will die. And it's interesting.
It's a very subtle title, very subtle.
Um, okay. Jack has two books. The first book he gave me, I think I dropped at page square root of 8. That was kind of it was very mathematical, but this book is all about how CEOs can actually use AI. And the fascinating thing, Jack, which I haven't told you, is you actually explain how it works, like the first part of how a neural network and how AI works and renders. I think is a must-read for all. So let's start with something, and today is going to be a very interesting discussion because we hear a lot of things. We hear LLMs, LQMs, we hear AI, everybody's doing that agents. But I want us to start by really examining and for you to explain the difference between LLM and an LQM because I think that is very important and very few people understand.
Well, Tasos, first of all, yeah, great to be with you, great to be here and to support Monaco House and the innovation that's happening in this house as well. So the AI that you've all been using, ChatGPT, Gemini, Anthropic, great models, and that is really AI for the digital world. You can make documents, text, images, really fun, fantastic. But when we're going to look back 10 years from now, we will look at this period of time as a very small sliver of what ultimately AI will be. And one of the biggest parts of AI is AI for the real world. So AI for the digital world, fantastic, but if we want to make a new medicine, we want to make a drug for cancer, we want to make a new battery chemistry to store energy or for a vehicle, we want to make a new material to send a rocket to space with a lighter rocket so they could bring more payload. All these are questions for AI for the real world. And that cannot be a model that's trained on cat pictures and Reddit is not the model you want to use to send someone to space. And so we need another kind of AI, and that's what's emerging right now. That's what we focus on. That's what others are focusing on. Now, when you hear Jensen talk about AI for the physical world, physical AI, others are working on it as well. This is the next big wave of AI is the real world impact.
And that means that we cannot train on the internet. We cannot download the web, Tasos, and say, 'Hey, now let's go train a model.' And the models themselves have to be different. And in point of fact, when we look at the data that we need for the real world, it must come from the real world. It cannot come from a textbook or downloaded from the web. We have to do real world experiments. We have to build molecules and test them in the real world. We have to look at if we're doing AI for financial services as an example, which again is another quantitative area, LQMs, large quantitative models, then we need to have data from the financial world. And if we're going to do data for the energy world, we need data from the material science and that world as well. So these are three big areas as examples where we need LQM, large quantitative models, and LLMs are not fit for purpose.
And Jack, this is great, but give us an example. And you're using LQM in pharma. What is a practical case study of how these are being used today already?
Yeah. Well, right now that's happening. We were on stage yesterday with Sanofi. We have other pharma partners and biotech partners, UCSF, University of California, San Francisco, many others. And what you find is that the pharma community did not have software and systems to really predict what's going to happen when you take a drug, manufacture it, and put it into a human body. And that's why you see the huge failure rate, Tasos, before this software came about. The failure rates in pharma, I don't know if people even realize what the failure rates are. It takes about 10-15 years to develop one drug, and then when it gets to clinical trial, we have significant failure rates. I'll just ask the audience, Tasos, if we could. What are the failure rates of drugs in a clinical trial? I'll start the bidding at 20% failure. Any other bids?
Any other bids? And John Buros of Alvan, keep quiet because you know you're in the industry. Anybody else?
Yeah. So it's about 85 to 90%. Right. Failure. Now if we were in the construction business in this room and we said we're going to build a skyscraper, Tasos, so 50 stories, and there's an 85-90% chance it's going to fall down in the first 30 days, that would not be a great construction business. That is the pharma business today. And the only reason why it works is that once in a while one building stands and it's worth multiple billions of dollars and it pays for all the failures. That is not a great business model, and it's also not great for patients. When we have failures in clinical trials, it's not just headlines in the business news. It's also headlines, unfortunately, for the patients. And we've got to do better. Alzheimer's, dementia, 40 years of research, Tasos, what do we have to show for it? Does anyone here, anybody aware of a fantastic drug for Alzheimer's? Anyone aware of a fantastic drug for dementia? No. Every one of our families in this room, unfortunately, will be hit, will be touched by these neurodegenerative diseases because as we grow older, that's when they hit. Cancer, the same thing. Heart disease here in Switzerland, across Europe, in the United States, in most countries, heart disease, cardiac disease is the number one killer. We have all the statins. We have all the drugs. We have all the diagnostics. And yet we have people dying. And that's the number one killer in the world. So we have a challenge in the pharma world. And we need better systems to model the body, model the drug, and understand what's going to happen before it goes into the first human. That's what LQM are about. That's what looking at training on biology, chemistry, and physics. We can model the molecules themselves before they hit the first human.
And Jack, we have a lot of people here from the financial sector, including the Minister of Finance of Monaco, Derek Joyce from XYB. I see Francesco Gley from CMBB, and I always speak and explain the risks and the dangers of quantum breaking all their encryption and systems. And I know in Sandbox it's something that you have worked for and you have solutions. How big is the problem, and what can people do about it today? And I say that because it's a unique gathering here of regulators and private enterprise, and I think the understanding is very important.
Yeah. Well, banking is very important for Monaco and it's very important for the world. And there are two key parts that we really want to focus on when it comes to financial services and banking. First of all is the quantum threat. You absolutely put your finger on it, Tasos, with Q-Day. So Q-Day is coming. I have to tell you, I have to share that with you. Q-Day is the day that quantum computers will scale and will break the key cryptography that we use every day to keep our banks and transactions safe, to keep our telcos working, to keep our governments working, confidential information inside companies. All this uses a set of protocols that were started in the 70s and 80s. One of them is called RSA after Rivest. One is called ECC, elliptic curve cryptography. These are the standards we've all been using. When you have your phone and you connect via an app to your bank, that is going and using RSA or ECC, and that will be broken by quantum computers. We don't have those quantum computers today that are scaled enough to run that algorithm, but they're coming soon. First of all, it's public information. The Chinese have made it public that they're working on it right now. They've empowered Pan Jianwei, a very noted physicist, to run this operation. And he's building, and his teams are building several quantum computers going to scale in the next coming years. The US, Europe, there are many, many quantum computers being built all across Europe, all across the US, and eventually we'll get to a scaled quantum computer that has error correction. That is, that if there's any fault in it, it will find it and correct itself. And once that happens, then it can crack any of those keys that are used for encryption. And of course, related to financial services and very popular at this conference as well is cryptocurrency. Cryptocurrency. What is the crypto in cryptocurrency? Cryptography. Cryptography that's used in cryptocurrency and Bitcoin and Ethereum and all blockchain is RSA and ECC. I see Steve Funk here. He's worried. He's checking his Bitcoin right now. And this is an issue. We must engage with the Bitcoin, with the Ethereum, with the blockchain community and engage this community so that we move over to a quantum-safe foundation for all blockchain, for all cryptocurrency. Otherwise, people can double-spend your Bitcoin once this kind of computer exists, and that will not be a good thing. It will have complete chaos and an implosion of the entire cryptocurrency. And Anthony, who I know well, was just talking about Bitcoin, and he's been very involved in that, and he and I have been talking, and others are talking as well, about how do we get the community to move over. So there's the traditional banks and the SWIFT system and the ACH system. We must protect that. We must protect all these transactions, and we must also move blockchain to the new era. Now, the second part, Tasos, of the financial service arena is portfolio construction, portfolio optimization, and risk management. Recently, a large insurance company, one of the top five in the world, came to us and said, 'We have a problem, Jack. We are trying to make risk models work. We have traditional risk models that go back 30 years. We only can run them once every quarter or so. We've got a problem. We want to be dynamic in our risk assessment. Every day we should be looking at our risk profile.' Having a hundred-plus billion dollars of policies in this insurance company, and those are not available in a large language model. A language model trained on words is not going to help you manage your risk as an insurance company, as a bank, as a lender. That is not the right tool. It's not fit for purpose. But a quantitative model trained on these areas is the thing that is fit for purpose. So in the financial world, we have both a threat that all the foundation is actually going to shake very soon. We must move over to the new standards, and the good news is there are new standards that we could all move to, adopted by Europe, adopted by the US and others. And we must also now start to use these tools to look at risk management and how we can construct and optimize portfolios going well beyond the old models of Monte Carlo. As much as we love Monte Carlo...
