Beyond LLMs: LQMs and the Next Phase of Enterprise AI | Jack Hidary on Bloomberg

追加: 2026-05-05

[00:00:00]- Financial firms are split on how to invest in quantum computing with Goldman and JP Morgan diverging on their strategies in recent years.

[00:00:08]It suggests a disagreement around the technology's potential benefits, at least in the near term.

[00:00:13]Now, our next guest runs a platform that helps accelerate AI and quantum research across industries such as biopharma, cybersecurity, and healthcare, using large quantitative models, a type of enterprise AI that uses physics to sharpen its outputs.

[00:00:28]Joining us live, I'm pleased to say we have Jack Hidary.

[00:00:30]He is the CEO of SandboxAQ.

[00:00:33]Jack, great to have you with us.

[00:00:35]Let's start on quantum computing because it's sort of pitched as the next big thing after AI and I feel like people are still trying to wrap their minds around it and understand it.

[00:00:46]And with that in mind, do you think it's realistic that that enterprise would even know what to do with it?

[00:00:53]- Well, it's a great question.

[00:00:55]We, we see AI and quantum actually as synergistic, working together, and that's the AQ, obviously, in SandboxAQ, AI and quantum.

[00:01:03]But it's a great question.

[00:01:04]There are quite a number of quantum companies right now on the hardware side, we're on the software side, but on the hardware side, there is about three dozen very serious efforts to build and scale quantum computers.

[00:01:16]They're making really good progress.

[00:01:18]If you look at Google IBM, Amazon, if you look at PsiQuantum, IonQ is a public company.

[00:01:24]Rigetti is a public company, D-Wave, a public company.

[00:01:27]So both public and private companies making good progress.

[00:01:29]In fact, recent announcements just in the last 60 days indicate that by 2029, we're likely to see a scaled quantum computer good enough to start cracking the cryptography that we use, the encryption that we use around the world for the banking system, for the government, and for private sector as well.

[00:01:47]So I would say that the companies are making really good headway to scale this.

[00:01:52]And we're seeing a number of public companies, we're expecting about five more companies to go public in the next three years as well in the quantum space.

[00:02:00]- And I wanna wrap it into, you know, what you see when it comes to quantitative AI and large quantitative models, of course, which you're very involved with when it comes to SandboxAQ because you position, basically, LQMs as where the real upside is in AI when you get beyond LLMs.

[00:02:19]- Katie, that's exactly correct.

[00:02:21]LLMs are having a major impact on productivity.

[00:02:24]Companies are saving money.

[00:02:25]You're seeing that roll through the economy right now and increase in productivity across the Fortune 500 will continue.

[00:02:32]But the question is how do you actually create new value, new products, new revenue, new EBITDA?

[00:02:38]That's really gonna come with the quantitative models, as you mentioned, not large language models, but large quantitative models, LQMs, these two different kinds of AI will work side by side.

[00:02:49]And I think, I'll make a prediction here that this year, we'll start seeing these two types of AI talk to each other in large enterprises.

[00:02:57]And so, you'll have a large scale company, let's say a pharma company, an energy company, a chemicals company, a big bank using LLM for some productivity, handling documents, handling summaries, emails, things like that, customer service.

[00:03:10]But they'll be speaking also then to a quantitative model that knows how to do risk modeling at the large bank that knows how to make new molecules for drugs that knows how to make new chemicals for large chemical and oil and gas companies.

[00:03:23]These two forms of AI are complimentary and I think we'll see their convergence this year.

[00:03:28]- Jack, I do wanna talk a little bit about security and security threats.

[00:03:32]Obviously, there was a lot of talk over the last couple of weeks about Anthropic's rollout of Mythos or maybe lack of a rollout there and some of the concerns behind that.

[00:03:40]Everyone's talking about agentic AI, Jack, and we know the potential benefits of some of that, but there are also a lot of pitfalls as well.

[00:03:47]And I am curious as to whether you even think corporate America or really any use case right now is prepared for that potential threat?

[00:03:56]- Well, it's a great question. There are three major factors in a successful use of a agentic AI.

[00:04:01]First and foremost of course is cybersecurity, as you mentioned. We've got to know what these agents are doing, not just in terms of hackers getting hold of them, 'cause if hackers get hold of them, it's more than just a break in.

[00:04:12]They now have the keys to the kingdom.

[00:04:14]These agents have their own logins, non-human identities, they're called NHIs.

[00:04:18]And so that's a very, very big threat there.

[00:04:21]There's many companies now across the board that are moving to secure the agents from a cyber point of view.

[00:04:26]But we also have to look at the ROI of these agents.

[00:04:29]These agents, they go through lots and lots of token usage, a lot of spend.

[00:04:34]They're like your kid sometimes when they're out there spending money on e-commerce and you don't know about it.

[00:04:40]That's what agents are doing.

[00:04:42]And so it's very important, not only do we secure the agents, but we also find out how much they're spending relative to their impact.

[00:04:49]And that's gonna be your ROI analysis that happens also with agents.

[00:04:52]And finally, interoperability.

[00:04:54]There's Google agents, there's Microsoft agents, Agent Force from Salesforce, all interesting platforms.

[00:05:00]How do you tie them together so they don't start conflicting with each other?

[00:05:04]- Yeah. - Those are the three major issues now in agents.

[00:05:07]- Yeah, absolutely. And you touch on a key point for this earning season, and that is of course, what the ROI looks like.

[00:05:14]That is the question that we're going to be asking over the next several days.

[00:05:17]But to bring it back to to LQMs, of course, large quantitative models, you posit that this is basically where we start to talk more about revenue generation, where it goes beyond just cost savings, productivity.

[00:05:31]And I wanna hear a little bit more about that.

[00:05:33]I mean, you talked about some of the R&D efforts that this can assist with when you think about healthcare, when you think about physics and other things.

[00:05:40]But I have to imagine, Jack, that we're talking about quite long timeframes here.

[00:05:45]- Well, actually what's happening now with LQMs, they're having impact right now.

[00:05:49]If you look at a broad range of companies, let's look at CPG companies, Proctor and Gamble, Unilever, the major CPG companies out there, Colgate and others, they have a strong need to issue new products as quickly as possible with good formulations.

[00:06:03]That's chemistry. That needs a quantitative model, not a language model.

[00:06:08]The same thing with pharma companies.

[00:06:10]Over the next 36 months, we are hitting the patent wall in the Pharma market, 200 plus billion dollars of annual revenue going away because of expiring patents.

[00:06:20]They need to have impact in the next few years, three, four years, not 10, 20 years.

[00:06:25]Again, that's a role for LQMs, large quantitative models, and the same in risk management and new products.

[00:06:32]If you think about BlackRock, State Street, Fidelity, they all need to create new financial products.

[00:06:38]And those financial products need their own particular quantitative model to make sure they're coherent and hit the sharp ratios they're going for.

[00:06:47]So all this about 80 plus percent of the market, - Yeah, - of the GDP is quantitative, not language-oriented.

[00:06:54]- All right, Jack, we gotta leave it there.

[00:06:55]Really appreciate you taking time.

[00:06:56]We gotta get you back - Great to see you both. - for a little bit more of a cohesive interview, Jack Hidary there.

[00:07:00]The CEO of SandboxAQ, which was of course, was spun out of Alphabet back in 2022.