Google Lost The AI Lead. Can Quantum Put It Back on Top?

Google Lost The AI Lead. Can Quantum Put It Back on Top?

Show Video

At the Santa Barbara lab, Google's shot at regaining its edge. So we're building a quantum computer, and that quantum computer will be able to solve problems that are otherwise completely impossible. Quantum computing the next big leap in technology that no one seems to agree on. This is the first event in history where a company CEO invites all of the guests to explain why he was wrong. But Google is all in and it could be key to winning AI.

If it fails, it wouldn't be the first time Google lost its lead in a race that it helped start. I'm Deirdre Bosa with the TechCheck Take. Google's next big breakthrough or bust. One of the fiercest debates in quantum computing right now is timeline to real world impact. And Google is out with a new countdown that will excite the enthusiasts. So we think we're about five years out from a real breakout kind of practical application that you could only solve on a quantum computer.

It's a controversial statement, given the wildly divergent time frames that have been voiced by some of the other players in the space, including from the Godfather of Computing himself, Nvidia CEO Jensen Huang. If you picked 20, I think, I think a whole bunch of us would believe it. Most people still think that's like a decade plus out.

So which is it? Another tech shift that's overhyped or are we just too early? Those same questions were asked of artificial intelligence companies pre 2022, or even about computers before they became mainstream. When computers were first starting to get serious, applications were primarily scientific and that was the incentive for developing the technology. But then as the technology continued to advance, many applications were discovered that hadn't been anticipated by those pioneers.

It could unlock new markets, solve previously unsolvable problems, and deliver massive returns for early movers. Where all that's going to go in the long run. Well, I think we're hampered by our limited imaginations to even guess. Google's Willow chip announcement in late 2024. It was a breakthrough moment for quantum.

It solved a problem not just faster than a classical computer could, but unimaginably faster. It takes about five m inutes to handle on a benchmark problem what would take ten septillion years on one of the world's fastest supercomputers. So you say septillion and I'm nodding my head. I don't actually even know like what septillion looks like. Yes, ten septillion. So this is a one followed by 25 zeros.

That is longer, way longer than the age of the universe. It wasn't just a theoretical capability. Google went further by showing a real world problem that a quantum chip actually solved.

A proof point, not just a promise giving the technology tangible weight. The other breakthrough from Willow was a dramatic reduction in error rates, which is crucial for scaling quantum computing toward real world applications. They make lots of mistakes and with the best technology we currently have under the best conditions, every time you do about a thousand elementary steps in a computation on a quantum computer, there's an error. We'd like to be able to run

computations with many more steps than that. Google's big breakthroughs showed that there is a reliable way to reduce those errors. Adding more quantum bits, or qubits, through its Willow chip. And that is a critical step toward getting to the real world applications that can actually solve diseases, create new materials, answer questions about the universe.

And that's kind of a milestone for the field. We've been wanting to see that for quite a while, and so it's a significant development. But look at what we're up against. Chemistry, searching for new pharmaceuticals, finding new materials with exotic and useful properties. We might require computations with billions

or even trillions of operations. So we have this enormous chasm to cross. We recently visited Google's Quantum Lab in Santa Barbara, where I was actually the first reporter to ever hold the Willow chip.

It's kind of crazy to hold it right now. Like, this is more powerful than a supercomputer. It's so c razy. And we even got up close and personal with Google's dilution fridge, one of the coldest places in the universe. This is what the inside of a dilution refrigerator looks like. You can see all the different wires and the different temperature stages.

All of this is because we have a control system at room temperature in the racks. We have our chip that's at the bottom of the cryostat, and we're just making connections between the two for like inputs and outputs. Investors too excited by the promise of Willow.

When it was announced in late 2024, it sent shares of Google soaring. Alphabet having its best day since April, after the company announced a major breakthrough in quantum computing. And created billions of dollars in value for smaller, less proven quantum stocks.

They're really focusing attention o n quantum computing stocks. We've seen those names really soaring in the past week. I think people were just reaching for anything with the name quantum in it. Quantum this quantum that.

Google kicked off a wave of quantum hype with its Willow breakthrough and other mega-cap rivals. They followed with their own quantum announcements. But the market and investors they've been less willing to buy in, making clear that Google has something here that others don't. Let's dig in. Microsoft. Microsoft announcing what it calls a quantum computer breakthrough with its new Majorana one chip. And Amazon.

Amazon jumping into the quantum computing race with its first chip. Just the latest tech giant to make a big play in this space. Bringing into focus the different layers of quantum hype. We talked about the macro hype,

the big, bold idea that quantum computing will someday revolutionize everything from drug discovery to climate modeling. But then there's the company-level hype, where Big Tech players are staking their claim in the race. But not all announcements are created equal. Google, Microsoft and Amazon's updates, they made noise, but they didn't move markets. They were more roadmaps than proof. In fact, Microsoft's splashy announcement around their Majorana chip is now facing serious scrutiny with experts in the scientific community.

But so far, we haven't seen very convincing evidence of the performance of the qubits in the Majorana one device. We really would like to see a published article that is available to the scientific community so we can see the evidence. And so far, that article hasn't appeared. Microsoft has worked for decades on this approach to creating qubits, trying to create a new state of matter called Majorana modes. The theory behind it is beautiful, and the theory is convincing that if you can build a device with certain specifications, it will give you a very good qubit. So far, it's been very challenging because the devices that they built are a little too dirty.

They need to be cleaner. They need to be more nearly perfect than they've been able to demonstrate so far. Essentially, Microsoft is making a long term, high risk bet on a beautiful theory, but one that has yet to be proven. Meanwhile, it's still struggling to demonstrate that it's successfully built even a single working qubit. Part of this is that there's a troubled history here. Microsoft had claimed in 2018 to make what are called Majorana zero modes, which are one of the basic building blocks of a topological qubit, and then that claim later had to be retracted.

I say just wait and see. I say that Microsoft ought to be given, you know, a few more months to publish the data, to have it peer reviewed. And one way or the other, you know, we're going to find out. There's nuance in Amazon's approach to experts say it's also in the very early stages, but that it could be a viable alternative to Google's.

Amazon also demonstrated as they increase the size of a quantum code, they get slightly better performance. Of course, that's just, you know, the first step along what is likely to be a long road. But it's an alternative to the Google approach, which might be beneficial in the long run. It's important to pursue these alternative paths, because we really don't know what's going to be the best way in the long run to make a large scale quantum computer work well.

The consensus is that Google, along with IBM, are working on a similar approach is far ahead in the race. But staying ahead is the real challenge. And if history is any guide, Google knows how quickly momentum can shift. For AI. It all changed with ChatGPT, which finally turned belief into adoption and revenue. Its parent, OpenAI, reaped the credit and the users it became the face of the AI boom, but it was actually Google that quietly laid the groundwork years earlier.

In 2017, Google researchers they published the now famous attention is all you Need paper introducing the transformer architecture that powers ChatGPT and nearly every large language model today. Given that they did come up with AI now, they may not have commercialized it, but they were there ahead of everybody else. Google had the foundational moment, but it went largely unnoticed by the public until OpenAI, a competitor, effectively deployed it with a product that was simple, accessible and instantly useful. Now, Google's Willow chip may represent a second chance to lead the next big era of technology, a shot to turn deep research into a defining commercial moment.

So what? We're going to just decide that all the problems with Google and Gemini and not knowing where they are in the real world should be trumped by this quantum computing. It's possible. Some believe quantum computing could also supercharge AI and boost Google's position in that race. When it comes to AI.

I think it's likely That future AI systems will continue to make heavy use of conventional computing, but we'll see more hybrid systems with a quantum processor incorporated into the system. One avenue would be the actual storing of data. In principle, you can store a lot of classical data very succinctly in a quantum computer. That process tends to be very slow. And so if we try to apply a quantum computer to problems that involve big data, like we do with conventional AI, now, the slow loading and unloading of information from the quantum computer may nullify any advantage in the speed of computation.

Another potential breakthrough the creation of data itself. AI's progress is starting to hit a wall because it's running out of fresh, high quality data to train on. As models grow their appetite for data grows even faster, but we've already scraped most of the internet. This is where quantum computing could come in. By simulating complex physical systems, they could generate entirely new synthetic data sets that classical computers they just can't produce. Feeding new breakthroughs in AI.

One of the potential applications that you can think of for a quantum computer is generating new and novel data. I think quantum computers will provide a lot of training data, which can then be used to solve quantum problems by conventional AI, much better than we've been able to do in the past. Not all experts are sold on the idea. From my perspective as a scientist who works on this, I think that the claims of quantum computing revolutionizing AI based on what we currently know, I'd say, are at least 90% hot air. And even Google itself says they're unsure. And there may be new methods that only kind of quantum information processing might be able to use to break down some of the fundamental mental barriers that may exist, but that's a little bit speculative, to be honest.

Still, Google is focusing on real world applications. Our focus really has been keeping our eye on when do we actually get to applications that can be solved on a quantum computer better than a supercomputer? And so what you see in our scientific results is we're we're typically making that comparison. And we think that the most important thing is to do that first and then have the product side can be involved. Once we have like the core demonstration of

the technology. Google's got its eye on scientific results and product key to turning a breakthrough into a business. And critical if Google is to win not just the quantum race, but the next big platform shift. Science is just the beginning.

Bringing it to market is the real victory for this piece. As you saw, we went to Google Santa Barbara Quantum Lab, and we sat down with Julian Kelly for an extended interview. He's been at Google's quantum unit for over a decade, and he helps us sort hype from reality. Here's the full conversation. What do you think interested Google about this space? I think what gets people so excited about quantum computing is that there are like fundamental theoretical foundations that show that quantum computers can solve problems that nothing else can solve. And I think it's that vision and that rigor. It gives companies like Google the confidence to make long term bets in these technologies, because we have a lot more certainty than a lot of other emerging technologies that when we make it, it's going to be useful, right? It feels like sort of all the mega-caps are now doing quantum in some way or the other.

But is it fair to say that, like Google was one of the earliest Big Tech companies in this space? Yeah, I think when Google got involved back in the day, a lot of people started taking the industry a lot more seriously. They're like, oh my gosh, Google's involved. Like, this looks like a real thing. We should we should pay attention. And I think we've been able to demonstrate the track record that that's been a that's been a good investment in terms of the scientific results coming out of our beyond classical demonstration 2019, showing that we can build a quantum computer that can out compute the world's largest supercomputer or something. And then our most recent breakthroughs with Willow. How do you make sure that you ship when you need to? Right. You move from the research and development

phase to make sure that you're taking advantage of commercialization whenever that happens, or breakthroughs when it needs to. That's it. That's a great question. And our focus really has been keeping our eye on when do we actually get to applications that can be solved on a quantum computer better than a supercomputer. And so what you see in our scientific results is we're we're typically making that comparison. And we think that the most important thing is to do that first and then have the product side.

Can be involved once we have like the core demonstration of the technology. Speaking of those core demonstrations, why was Willow such a breakthrough for the industry? Yeah. So Willow did two really important things. So one, we used Willow to revisit this standard benchmark where we compare it to the world's largest supercomputer, because that's basically what our competition really is. At the end of the day, we want to beat classical computers, and we run this benchmark problem that we can run on Willow in just a couple of minutes.

That would take ten septillion years. Ten to the 25 years. Way longer than the age of the universe. And what that means is that that problem, that problem would be just impossible for a classical computer to solve.

So that's one. And then number two is basically opening the door to the future. We show that we can be what is called below threshold for quantum error correction. What that means is Willow has shown that we can scale our system up, make very large computers in the future that will be able to solve generic and useful applications. How is that different than some of the releases we've seen from other Mega-caps over the last few months, especially for someone who, you know, doesn't really understand the error rate, or even ten septillion years. Like how can you can you put it in perspective? I think what's really important for people to understand, who don't have as much familiarity in spaces. At the end of the day, we're making a

computer. Sometimes people hear quantum computer and they kind of throw out everything they know. But it's a computer and computers have specifications. They have things like how fast you can run it, how large it is, what types of benchmark problems you can solve.

And this is how we grade ourselves. And this is a good metric to figure out how to understand the progression of technology in the field. So for example, with our Willow announcement, we released a spec sheet that goes with it that tells you all the different pieces that we think are important and what the numerical values are. And you can use that to compare with other technologies to see where they're at. And we also run these benchmark problems, like I told you, this ten septillion years one, that's another example where you kind of put the whole system together, see how well it runs, and you can compare it with other quantum computers and you can compare it with the world's largest supercomputer. Is that standardized? If you have a breakthrough, is it expected that you also release a spec sheet, for example, you know, did Microsoft have a spec sheet that accompanied its announcement? Did Amazon? I would say that not um, not all, uh, A programs are releasing spec sheets that go with their announcements. Sometimes you can find it

in, say, the scientific papers, or sometimes you can find pieces of it. But the thing that really want to we want to drive towards is kind of capturing all of those metrics, not just some, but all of them at the same time, to go with our announcements so people understand where the where the state of the art and the technology is at. Okay. Let's talk about how quantum computing relates to artificial intelligence is I guess when we spoke previously, there was this idea that in the future, quantum computing can supercharge AI. Is that right? Yeah. We think that quantum computing and AI can really

complement each other. Walk me through how quantum computing can complement AI. So there's a number of things that quantum computers can do. But in this particular case,

for example AI and AlphaFold, it trains on data that's provided, that's informed by quantum mechanics, but that's actually not that common. So a thing that, for example, a quantum computer could do is generate data that then I could be trained on. In order to give it a little bit more information about how quantum mechanics works. Got it. You talked about having your eye on the

applications. What are the applications? How far away are we from that? Yeah. Good question. So we think we're about five years out from a real breakout kind of practical application that you could only solve on a quantum computer. But there's a whole host of kind of interesting example application areas.

Again, I mentioned that quantum computers are so interesting because there's some kind of theoretical guarantees that they can simulate quantum mechanics that's relevant in many different areas. But of course, you can pick your favorite. Like there's a fun story around fertilizer. I don't know if you happen to know this one. So it turns out that a very large amount of the world's energy usage goes towards making fertilizer using something called a haber-bosch process. And it uses like high temperature and high pressure. And this is actually like a revolution

in industrial chemistry, like 100 years ago. Because before that, the way that we used to get fertilizer, we would actually send ships out into the Pacific Ocean. Chip bird poop off of these islands, out there in the middle of nowhere and then bring them back. So this process is way better than that.

But at the same time, if you just look in like the soil and plants, there's bacteria that will do this at like standard temperatures and pressures way more efficiently. And we'd love to just copy them. Right. They're doing a great job, but we don't know how. But if we had a quantum computer, we could see how they work.

Use that to solve, for example, this fertilizer production issue and then save humanity a whole bunch of energy. Interesting. Is that one of the solutions or applications we could see in five years? I think that one's a little bit further out, but it's one of those things that, again, is like a very concrete example of how a quantum computer could be applied. Right. What's something we could see in five years. I think.

And how will we know if how will we know if you've reached that benchmark? Like how do you measure. So that's that's a good question. Um, the first applications we think are going to be in places that are like really kind of like just kind of cutting edge physics areas where you've got some system that's sort of just out of reach of what a classical computer to do. But a quantum computer can kind of just put you past it to give you important scientific insights. Something that's really important and that we

take very seriously at Google is making sure that we kind of red team the problems that we're working on. Like, we want to make sure that if we're making a claim that something is better on a quantum computer, we really rigorously test it against the classical side and have people look at the best classical algorithms so that people can be confident that when we say that we have a breakthrough, we actually have it. And that's that's something very important for us.

When Jensen Huang of Nvidia, you know, like the godfather of computing, said, we're at least 15 years away. It felt like he was pouring some cold water on the space. How does that differ from what you're saying? What's he looking at? Um, I would say that, you know, there's a lot of nuance in quantum computing, and it depends on kind of what you're saying. Like when the revolution has happened, I think there will be, you know, early signs that, again, we're running things that can't be solved anywhere. And then there'll

be sort of a late retrospective of like, oh, clearly quantum computing is super important. And, you know, we're confident that we're going to get there because we know that these things can do things that nothing else can do. So I think it's just a matter of where maybe an individual would call the revolution has happened. When do you think enterprises start to adopt it? It felt like, you know, ChatGPT was the moment that made it clear to consumers and enterprises how they can use AI in their own businesses or personal use. When does that happen for quantum computing? When does it go past the scientific research phase into consumer and enterprise? Yeah, I mean, I think, again, the most important thing to look for in the next couple of years is this practical application where you can do something that has physical relevance. You can kind of

imagine how that transforms a real world problem, and it can only be done on a quantum computer. I think at that point, it behooves everyone to pay attention to the field. Right. You said that AI is one of these applications

that may be further away. What would we have to see to know that we're getting closer to that moment? The way that we see it is at Google where we have something called a hardware roadmap. So there's sort of a hardware track, and then there's an applications track.

And where we want to get to on the hardware track is to build sort of a scale model of this error corrected quantum computer. So this is the the thing that Willow opened the door to. It says this in fact is possible with exactly what you're building right now. You got to scale it up and make it bigger. Make it better. We think by the time we have that machine end to end, we'll have a very good we'll have completed this kind of 0 to 1 moment for making error corrected. Quantum computer will be very you

can understand it, benchmark it, look at it. And from the hardware side, that's really the thing that we're marching towards as fast as possible. Then you take that machine and you match it with an application that can then fit inside that computational capacity. And that's when you kind of have like the killer mix in quantum. What is maybe the most unappreciated thing about the space right now. Is t he timeline, to me, five years is not something that I hear on a regular basis, but you are very confident in that projection.

I think, um, you know, what's really interesting in, in the, um, in the field is, is that like when I started a number of years ago, people actually thought that building a quantum computer was impossible. They thought like, oh, there's no way the physics isn't going to work out. Quantum error correction is going to fall over, but there's just growing more and more confidence that a quantum computer is possible to build and will be built. And we spend more of our time thinking about what types of things will we do with it, how is it going to work to benefit society? Rather than kind of agonizing over if the fundamental physics is going to play out.

Do you think about competitors? Who would they be? . You know, it's a good question. Um, I would say that truly, we think our competition here is classical computing in nature. We want to harness nature and show that we can use quantum mechanics to actually solve problems that can't be solved in any other way.

And, um, you know, our competition is classical computing, which is it's a high bar. You know, humanity has put a huge amount of time and effort into making computers work extremely well, so we need to get over that hump, but we're confident that we will. Google was underappreciated in artificial intelligence for a long time. Do you think it's being underappreciated right now in quantum computing? I will say that I think, um, not the average layperson doesn't understand how important quantum computing will be and the types of problems that it can solve.

And it can be tricky, again, because there's a lot of noise in the space. It's hard to understand from one headline or another what's really happening. And so again, a thing that I want to get across here is like it is really it's a physics but also an engineering discipline. You can go look at the numbers. You can really make that tangible.

When you say physics and engineering, I understand like these are enormous feats, but it's hard for the average person to grasp onto. Right. And like we saw that with ChatGPT, there was so much interesting happening in AI. But until like, you could put an interface in front of someone and they could play around with it themselves. And the average person is not a physicist or an engineer. Like, what is that point? When do we get to that point with quantum? Do we ever? Is it going to be packaged into something else? So I think, um, by the time, uh, we build like a scaled error correction computer, then error correction applies this interface to it. That makes it much easier to access than the computers that we have right now. It makes it look much

more like a normal computer and how you program it. Whereas right now, if you go interface with the computers in the lab, it can feel a little bit more like a physics experiment than kind of fine tuning knobs and things of that nature. So I think when we have that sort of interface layer placed on top of it, it will open the doors to many other people being able to interact with the technology. And I think that's going to be a very transformational moment right now. We're at a moment last few months with AI, where we talk a lot about this data wall and how, you know, the x axis has kind of stopped and we're just going up the y axis now. Right? Because there's only so much data in the world.

What does quantum computing do to that chart and to models and frontier models. Yeah. Again, I think, you know, one of the potential applications that you can think of for a quantum computer is generating new and novel data. Um, and there may be new methods that only kind of quantum information processing might be able to use to break down some of the fundamental barriers that may exist, but that's a little bit speculative, to be honest.

Why is the promise of quantum computing doing that different than AI, creating its own synthetic data, which has sort of been not a very effective way to reach new frontiers? Yeah. And again, it comes back to this, um, the AI models that people use are built on classical computers, and they just they fundamentally can't speak the language of quantum mechanics or a quantum computer can. And it really goes down to that kind of foundational level of like, quantum computers speak quantum mechanics. They can access the way the universe works at the most fundamental level.

And there's a lot of potential in, you know, solving entirely new problems by being able to build a machine that works like that. Julian, thank you so much. That's great. Great.

2025-04-20 13:15

Show Video

Other news

Revolutionizing Construction: 13 Cutting-Edge Tools & Technologies Shaping the Future! 2025-05-06 09:36
Getting Data Off a Failed Pre-Built NAS 2025-05-04 02:20
Радиометр РК-01 2025-04-30 23:55