00:03:43 · Pete Shadbolt
[00:03:43] S0: Yeah, it's a great question. There are basically very formal first principles reasons why quantum computers will always be necessary to solve the hardest computational problems and where AI will not solve those problems in general. And you're right, amazing progress is being made with applying AI to chemistry, to material science, to drug discovery, fuels, and many of the same kind of applications that we care about. They will encroach on that territory a little bit, but we also have extremely strong reasons to believe that they won't go and take all of that territory. One way that you can see this, I think if you look at some of the AI drug discovery companies, material science, chemistry, etcetera, Very often what they're doing is actually building wet labs. So they're literally building a robotic system to go and collect the data that they then use to train the model. And that's amazingly effective as we know. AI has been astonishingly effective at making approximations that are learned from data. But it's also brute force. It's also pretty intensive to do that data acquisition. And it tends to produce a model where you're really good in one narrow domain. Also, see that these models are really effective when we have rather simple molecules or non correlated electronic structure problems. But when we want really, really high precision and in the hardest instances of these chemistry problems, reaction chemistry, material science, etc, Those models still fall down. And that's what we hear from our customers. So when we work with people who do drug discovery, people who do lithium ion battery chemistry, etc, they're making huge use of AI for these problems. But there are also people in those teams who say, ultimately, we're going to need something better. Either we're going to need a much more powerful source of data to train these models, or we're just going to need to do these first principles calculations directly. And so, yeah, it's amazing to see what has been done with AI, but we also have very strong reasons to believe that there's a limit to where that can go. One way to think about it is that what AI has done is it's been really successful in making approximations that are learned from data.
00:06:14 · Pete Shadbolt, Adam Miller
What quantum computers do is fundamentally different and we think really complementary in that a quantum computer calculates exactly from first principles. So when I add two and two and get four, that's a first principles calculation and I have total confidence in the answer. I'm not learning that. And that's how quantum computers produce data about chemistry, about material science, etc. And obviously, that's a very universal capability when we turn that on. So we think that that kind of synthetic data that you can produce with a quantum computer with complete first principles confidence, that's going to be hugely complementary when we go to train these big next generation foundation models for the physical world. [00:07:04] S3: And I want to try and bring this down to what the average person might be able to understand because they see tens of billions of dollars invested in this space, the Australian government investing $1,000,000,000 taxpayers' money, and for the public, they don't get to see many results because this is such a long and drawn out process. And so a lot of people are questioning, why is the money in this space, the attention in that space? For the average person, why is this eventually going to matter?
00:07:31 · Pete Shadbolt
[00:07:31] S0: Yeah, it's a great question. So I think that the first thing to say is that quantum computing is not a mass adoption technology. You should not expect that millions of people are directly interfacing with the quantum computer. Quantum computing is going to be narrowly available to a very small number of acute, hard problems. And that is true of many of the most important industries that we have on our planet. So So if you think about the biggest, most consequential, most important companies of our time, TSMC, ASML, SpaceX, NVIDIA, what number of people directly interface with an ASML EUV tool? How many people have ever done a tape out of TSMC? How many people design a payload for a SpaceX satellite? How many people have ever actually written a kernel for a GPU cluster? It's a really small number of people. It's like hundreds of people who really do that work. And so the first order utility of the product is opaque to the average person. And that's fine. Like, that's a very natural, healthy property of our world. And it's really the downstream products. It's the second and third order effects of these pretty alien technologies, right? Like a semiconductor fab, 99.9 of people will never step foot in that building. And if they did, they'd seed stuff that's totally baffling, like boiling sulfuric acid and two nanometer transistors and so on. They don't care. What you care about is that you can watch YouTube on your cell phone at the end of the day or that you can buy a drug in the pharmacy or you fly on an aircraft where that wing wouldn't have been designed were it not for a leadership class supercomputer in a DOE national lab. So quantum computing has exactly the same property. The machine and the simulations and the calculations that that machine does, those are always going to be the domain of weird people in a building somewhere who, you know, talk in code, but it's the downstream effect. It's the drugs, the fuels, the catalysts, the fertilizers, the next level of mastery over chemistry, physics and math that will flow down into our economy, into our society and ultimately into our daily lives. And so I really hope that in the near future, people are using catalysts, fertilizers, drugs, fuels, whatever that wouldn't exist were it not for the computational capability of these machines. And honestly, they might not know.
00:10:17 · Pete Shadbolt
They might not know that that thing was designed on a quantum computer. And to me, that will be success. I'll be delighted that they don't know that that thing was brought into life partly by virtue of this technology.
00:10:33 · Adam Miller
[00:10:33] S3: I want to thank Notion for sponsoring this episode. Without companies like Notion, these episodes wouldn't happen. They're ambitious and they cost money, and Notion backed us early. But the other way that Notion helped us is that we've literally been using the product since day one. We've been using it for media, operations, product. Within media we organise all of our videographers, our sponsors, our episodes on Notion. It's an absolute lifesaver. It's basically where our whole company operates. But the thing that was a real game changer for us was Notion AI, because our company is bigger now. We've got 25 plus people, different divisions, so there's a lot of docs, systems and structure. That's helpful, but with Notion AI, I can simply prompt something in a chat bar, so I don't even have to go into docs anymore. It's a lifesaver. Think episodes like the one you're watching today about PsiQuantum and Pete, it's really reiterated this idea to me about focus. If you want to be a standout company and do extremely ambitious things, you've got to focus on the one thing that matters. And for PsiQuantum, it's building a 1,000,000 qubit quantum computer. It's not docs, it's not administration, it's not boring paperwork. And that's what Notion allows you to do. As a founder, focus on the things that only you can do, people, culture, your product, and then let Notion take care of all the boring stuff in the background. We've got a special offer at notion.com/sachanandatom. Give it a try. I promise you, I've never heard anybody that's regretted it.
00:11:59 · Sachin
[00:11:59] S1: I think there's a really interesting analogy there, particularly with TSMC, one of the most important companies in the world, not many people actually understand how it works, and they understand that NVIDIA uses a lot of their chips, etc. I think for the audience watching, can you really clearly articulate what this world may look like in ten years? Obviously, you have dedicated your career to this. What is the use case that gets you out of bed every morning going after this problem? And what do you think the world will actually look like in ten years?