What quantum computers may actually be good for
Quantum computers are most plausibly useful for material science simulation, cryptography via Shor's algorithm, and search via Grover's algorithm, but a key algorithm for material science, the Variational Quantum Eigensolver (VQE), has a significant limitation: it requires the user to guess the ground state wave function of a material first, then optimizes around that guess. If the initial guess is wrong, the algorithm fails, and the problem gets worse as systems grow larger. One speculative path forward is a hybrid approach where AI systems generate the initial guess, receive results from a quantum machine, and iterate, potentially making quantum hardware most useful as a backend for AI-driven discovery.
- Krishna contrasts VQE with Feynman's original dream of direct quantum simulation, pointing out that VQE is not a straightforward quantum version of density functional theory, which already knows what electron clouds look like.
- AlphaFold's success at protein structure prediction is raised as a reason to question whether quantum computers are still needed for biology-adjacent material science, though the hosts note AlphaFold is not yet perfect.
- Omar Yagi, described as a 2025 Nobel Prize winner and pioneer of reticular chemistry, is mentioned as someone pushing 'AiOMETRY,' a blend of AI and material science, as a potential bridge between classical AI and quantum algorithms.
Transcript
This chapter, from the episode video's captions · 1,045 words
2:02:26theory behind it and why we're building this thing in the first place. Okay. And I I really personally I think it's personally I think it's because of Shores and Grovers Fineman's dream of material science. I think um in the age of AI it's less so much a priority right because I like just look at um Google deep mind's alpha fold right it won the Nobel Prize because it's really that good at predicting protein structure maybe we don't need a quantum computer to predict protein structure I will make a small note yes it's not 100% efficient yes there's we've gotten a lot of comments
2:03:07about yes alphafold is interesting but when brought into later stages. So, I'm just qualifying it. Yes, we are aware >> it's not perfect. We're not saying that it's perfect, but from when we where we were before >> to it existing and being able to then iterate and build on top of that, >> it's going to get good and there's no reason to believe that it's not going to get good for like material science itself and like you know, physics in general. Um the other thing why I the other um reason is like I think the best algorithm that's out there to the one that's like most frequently used to figure out things like ground states of a material is something called the
2:03:49variational quantum solver. This is you know just like how Shor's algorithm is built for this purpose of factoring. Um this VQE algorithm is built for the purpose of understanding materials. But that algorithm requires you to guess an onsat in L you know how um Hans beta had this idea of guessing uh guessing an answer and then optimizing around what that answer is. Well, this variational quantum solver >> requires guessing at what you think the ground state wave function looks like and then optimizing around that wave function. Well, if your
2:04:29guess is like completely out of left field, you're not actually going to get anywhere, right? So, it's not like one of these. It's not the dream of Fineman of like, oh, just like simulate it, right? You still need a quantum algorithm to do the the computation. And it's not as simple as like, oh, we just like we just take density functional theory and create a quantum version of density functional theory, right? Where like, oh, I know exactly what the electron clouds look. And the whole point is that thing is going to tell me what the electron clouds look like, right? So I have to guess at first description and then it only really varies the parameters around my guess. If my guess is wrong, I could be in trouble. And as I get to larger and larger systems, the probability of my
2:05:09guess becoming wrong gets higher and higher. So there's still not a good enough, I think, quantum algorithm to get me the material science promise that Fineman was dreaming of. Maybe one person who will help push the envelope there is someone we just covered uh in our last episode, Omar Yagi. Yes. who's just uh moved out to Beijing, former Nobel Prize winner >> 2025 and is the pioneer of reticular chemistry and is pushing this idea of what's being dubbed uh aometry which is AI material science and chemistry
2:05:50>> and trying to basically blend these disciplines >> and potentially I know that the quantum piece in quantum algorithms are not necessarily the basis of the concept But seeing where this implementation of AI in speeding up nextg material science discovery and then if you add a fundamental quantum algorithm discovery in that context I think the acceleration becomes very interesting. So >> yeah yeah I mean it could be like we could live in a future where like um AI systems are making the guess right >> right and then they get an answer from the quantum machine and then they they iterate on the guess. I mean, who knows,
2:06:30right? Maybe in the future only AI uses quantum machines because they're the ones creating such create like amazing algorithms. Who knows? It's it's it's a ripe field, but um that's how we got here. >> Yeah. And and I think that this this is great foundation. Again, the whole point of this show is we have an expert and a layman. Myself being the layman. our resident PhD Krishna Chowdery being the expert trying to navigate these complex topics at both levels at an expert level. So for those who are super technical, shout out members of the HRL team that made it through two hours of this podcast uh and
2:07:12literally no one. >> If you did, drop it in the group chat. I want to I want to hear you guys. And then the regular everyday people like me that do have an interest and curiosity about some of these things but may not have the foundational tools to be able to dive in and get it. And I think that blend is why so many people love the pod because we're able to communicate to two different audiences. And so for those who keep asking in the comments, why does Lester just sit there staring the whole time? >> Yeah. First of all, >> actually no, this is a this is a familyfriendly show, so I'm not going to say what I think I want to say. I just the structure of this show is very much present in other forms of
2:07:54media. You have subject matter expert. You have an everyday person and you're trying to communicate for two different audiences listening. And so my job is to smile, listen, and look pretty and sound occasionally kind of smart. Occasionally be like, "Oh, that's good." >> That's what I fight for on this show. question of being surprised that I made the Oh, made that connection. That's pretty good. Um, so again, part two we're going to actually dive into and hopefully have a physical copy of Yes. >> that we can hang on the wall and start our gallery wall in the back of this cover story um quantum silicon
From How Quantum Computing Actually Works (Part 1)
Part I of our quantum computing deep dive traces the field from Bell and Feynman to Deutsch and Shor—and explains what quantum computers actually do differently from classical machines.