Microsoft's big quantum bet

0:00This BBC podcast is supported by ads outside the UK.

0:30for all the great deals happening now. Save big money at Menards. Every story is a technology story in one way or another. And on the Interface podcast, we decode the tech that's rewiring your week and your world. On this week's episode, we'll look at whether or not Google is about to destroy the internet. Can the Pope save us from AI? And is Roblox finally facing a reckoning? Listen on bbc.com or wherever you get your podcasts.

1:27Kleinman brings us the latest. And a little later, Shona McCallum laces up her boots to tell us about tech at the FIFA World Cup. Obviously, if you bounce the ball, you'll see a much bigger touch. And this just gives you an idea of the sensitivity of the ball. And as an alternative to all the hype, we catch up with one of the best known scrutineers of generative AI.

2:09We begin today with some major news about quantum computing and a high-risk bet made by one of the world's biggest tech companies, Microsoft. Quantum computers exploit the weird properties of physical particles, think photons, electrons and so on, as described by the science of quantum mechanics, hence their name. It's mind-boggling stuff, but they promise to be able to solve problems that conventional computers just can't, delivering valuable discoveries in fields like chemistry and medicine. Big tech firms like Google and IBM are investing huge amounts to be the first to develop one that can deliver practical, commercially valuable solutions to problems, as is Microsoft.

2:51And last Last year, the BBC's tech editor, Zoe Kleinman, managed to gain access to their quantum laboratory. Not many people have ever been inside this lab, but in here is one of Microsoft's quantum computers. It's shaped like a chandelier. It kind of hangs down. Let me show you what's right at the bottom here. This bit of kit, which is super heavy, is called the gold finger. It's gold-plated because the temperatures down there have to be super cold. We're talking colder than outer space. And right at the bottom here, right down here at the tip, is where Microsoft's new chip will sit. Well, I'm pleased to say they let Zoe out of the lab and she joined us to reveal some more news about Microsoft's quantum efforts.

3:34Well, another year and another quantum chip. This is the Mayorana 2, and it is very significant for the company, which is in this massive global race, if you like, to try and develop powerful quantum computers as quickly as possible. I suppose at this point we should just pause and tell the audience what quantum computers are. I mean, essentially, these are machines that will work in a fundamentally different way from what are known as classical computers. The computers that are in our iPhones, that are our laptops, that are even the big computers in data centers, the conventional supercomputers.

4:11These will work in a really fundamentally different way. And everybody's very excited because they promise solving problems that today's computers can't. They work differently to the computers we have now because they can do a lot of calculations at once. So a classical computer can, you know, even though it does it very quickly, it still works through a calculation in a linear way. A quantum computer promises to be able to do the whole thing simultaneously, which gives you the answer a lot more quickly. As Zoe said, there's a race to develop a quantum computer that can solve commercially valuable problems.

4:50But Microsoft has chosen what it calls a high-risk, high-reward strategy to take on its rivals. Its approach required harnessing the properties of a quasi-particle called a Majorana. An Italian physicist had theorised that Majoranas existed back in the 1930s, but no one had seen one. It was a big bet to use this as the basis of their quantum work, and when they unveiled their first quantum chip, there was a great deal of scepticism from scientists. scepticism about whether they had shown it could work and even whether they'd proved these particles existed. Microsoft stuck to its guns, pointing, among other things, to its involvement in a US Defence Advanced Research Project Agency, or DARPA, programme as corroboration of its efforts.

5:39Still, a lot is riding on this new chip, the Majorana 2. Minutes before we spoke to Zoe, she was able to record an interview with Jason Zander, who leads Microsoft's discovery and quantum program. The exciting thing in what we're announcing today is our Majorana 2 chip. Now, this particular chip is designed to be a breakthrough in reliability. And with this chip, one of the things that's super important is how long can a qubit last while it's doing compute. With this chip, we have a mean time of 20 seconds and up to a minute that's measured. And just to put that in perspective, most of the quantum computing modalities are out there.

6:17the other hardware types are lucky to get into the millisecond range. And why is that important? Because the longer the qubit can stay stable and reliable, the easier it is for it to do computation. I think for a lot of people, quantum falls into that category of like fascinating and amazing, but also incredibly difficult to understand. A quantum computer is simply able to take a problem space that is massive, just unbelievably massive. And if you think of a classic computer, it has to go through it. Like, is it one? Is it two? Is it three? It would take forever. A quantum computer can actually in parallel look at all the possible answers, you know, pretty much in parallel and run through it.

6:56That's what makes the thing so powerful. And why is that important? Because you're describing physics, you're describing chemistry, and chemistry and materials make up, you know, our everyday lives. You know, the clothing we wear, the food we eat, you know, it's like if we could make massive improvements in that, we can, And like I said, help massively with sustainability, climate control, better fertilizer, feeding more people. I mean, that is all science at the end of the day. Quantum computers are very good at science. What has been the biggest challenge in terms of getting to where we are now?

7:27Yeah, I'd say for our project, it really was proving the physics. In our case, we had to take a theorem from 1937 that really looked like it would be very, very good for quantum computing. But the first thing was doing the theory and the applied physics to actually prove it. And then once we did that a few years ago, how do you actually take that and harness it and turning it into a qubit, which is, you know, the unit of a quantum computer? That's been really the hardest point of that. You've taken a different approach to other quantum developers, haven't you, with the topological qubits? We have been on a 20-year journey at Microsoft.

8:05So it's a fundamentally different approach. But what we realized very early on is that to get to true utility-scale quantum computing requires three things. I have to have something that's reliable, I've got to have something with a decent size, and it's got to have a fast speed. And the cool thing is with this technology, not only do we have the reliability that I mentioned, but the qubits are one-hundredth of a millimeter, which means we can fit a million qubits in literally the palm of your hand, which is pretty amazing. Now, at this point, we are moving from proving the physics that, again, took us a very long time to reach, and into engineering, where we push scale for the compute going forward.

8:43But it's a fundamentally different approach, but we think it's worth it. You faced quite a lot of criticism from the community when you announced the mayor on a one particle last year, with people asking whether you'd really done what you said you had done, which was kind of to create a new state of matter. Do you stand by that? Absolutely, 100%. And one of the things I would say is that we're a commercial program. We're not a university academic program. But we still believe in scientific rigor. That means that we're publishing all of our results we did with Majorana I. We're publishing for Majorana II as well.

9:19And I would say that the biggest change between Majorana I and Majorana II, we actually changed the superconductor. We moved from aluminum into lead, which is, frankly, another new innovation. Most people do not know how to use lead to build chips, but we had to figure it out. And that's given us significantly improved results, which, again, we are publishing with this announcement. So what do you say to your critics who last time around said, we don't believe you can prove that you've actually done this? Well, like I said, we really look into scientific rigor. We welcome the debate. That has always been part of physics.

9:55Back to 100 years where Einstein and Bohr were debating each other on physics. it's always been part of this particular discipline. So we absolutely welcome all that kind of feedback. Again, the key thing I would tell people, go read the papers and look what's there. Go talk to the experts that we have given deep information on. And how much closer does this bring Microsoft to a working quantum computer? So this pulls in our timeframe, and that's one of the things that we're announcing into what we believe is the year 2029 for utility-scale computation. Because like I mentioned, one of the things that we've done is we spent a long time on the physics because this was a hard one.

10:30We did have to actually prove the Majorana theorem and then turn around and prove we could turn into a qubit to get this kind of Goldilocks zone that was required. So now that we've done that work, we're switching over to more engineering. And we believe with this, yeah, 2029 is looking very good for utility-scale quantum computing. Jason Zander there, some big news. I mean, I guess one of the most interesting things is that they say they're going to have this million qubit chip by 2029. I mean, that's pretty close, isn't it? And when you think of things like, you know, quantum computers being able to break encryption, I mean, this is quite a sort of shrinking of the timeline, isn't it?

11:10We're going to have to kind of get ready for that, aren't we? Look, it's easy to say, isn't it? I'm going to be doing X, Y and Z this time, and who knows whether you will be. But it's a very confident prediction to make. And, you know, if it is true, And if Microsoft reckons it can get there in three years, then you can bet that the other big companies won't be that far behind. Then it means that we could be entering the next decade with, you know, a level of supercomputing power that we've never had before. And that is both really exciting and also quite terrifying, isn't it? And of course, it's going to be an announcement and a claim that's going to be subject to a huge amount of scrutiny, isn't it?

11:48Yeah, and rightly so. And, you know, Jason Zander did say, didn't he, that's what we expect. That is how science works. And hopefully the ultimate outcome of all of that is a better product, isn't it? The BBC's Zoe Kleinman there. And we will follow reaction to this news closely. And if you've thoughts on the announcement, do get in touch.

12:12Every story is a technology story in one way or another. And on The Interface podcast, we decode the tech that's rewiring your week and your world. On this week's episode, we'll look at whether or not Google is about to destroy the internet, can the Pope save us from AI, and is Roblox finally facing a reckoning? Listen on bbc.com or wherever you get your podcasts.

12:51This is Tech Life on the BBC World Service with me, Chris Vallance, and you, our listeners. Some of the subjects we cover can generate a lot of responses and sleep is proving to be one. A month ago, we ran an item about tech to help those with a condition called sleep apnea. Podcast listener Jeff Pickard in the UK emailed us to say he suffered from sleep apnea for many years and hasn't had any success with a traditional solution, a CPAP machine and mask. Well, Jeff, we hope you do find a solution that works for you. And thanks for listening. And if you want to share your thoughts about anything on this week's show, please do get in touch.

13:30Our email address is techlife at bbc.co.uk, or you can send us a text message or voice note on WhatsApp. Our number is plus 44 330 1230 320. Please remember to include your name and where you live. And still to come, a sharp-eyed observer of AI's progress gives us his take on the tech.

14:01Now, whether you call it soccer or football, later this month, you're going to be hearing a lot of this.

14:10The World Cup kicks off on the 11th of June, and with it comes all the passion, joy and tears of following your favourite team. So no pressure on the officials then. Tech Life's Shona McCallum has been given exclusive access behind the scenes at FIFA's headquarters in Switzerland to look at some of the new tech that will be rolled out for the competition. The FIFA World Cup. It's a tournament that's defined generations since 1930, and not just by goals, but by moments that have captivated millions of fans all over the world. But putting the ball in the back of the net isn't just about skill and ability.

14:49There's now technology monitoring every kick of the game. And here in Zurich, I've been given access to FIFA's global headquarters to see the tech in action before this summer's competition. The man in charge of FIFA's research team is Nicholas Evans. He's been paying close attention to the one thing every match needs, the football. It is a normal football, but this one has a sensor inside, which is placed opposite the valve. This is the sensor that we talk about which sends information at 500 hertz so that's 500 times per second which is 10 times more than what you see on broadcast images The sensor is a unit of three different measurement devices It's called an IMU, an inertial measurement unit which has an accelerometer, a magnetometer and a gyroscope and in the simplest of terms it can tell us what the ball is doing in a 3D space The sensor will track each touch and can help make decisions about potential offsides or things like handballs It will also mean that broadcasters can give football fans information Like who kicked the ball hardest or what was the fastest goal strike So here we have a screen which you can see and you can then try it for yourself But basically you can even see the faintest of touches here Obviously if you bounce the ball you'll see a much bigger touch and this just gives you an idea of the sensitivity of the ball.

16:19Often at times the images won't be clear enough or will be very difficult to determine whether a player has touched the balls. This is very clear. The team here are happy the ball's gone through rigorous updates and testing since the last World Cup in Qatar. At a match there will be between 15 and 20 of these smart balls ready and charged with each one lasting six hours if being used in play. Beyond the ball, there's a 3D scanner being deployed. All 1 ,248 players from the 48 teams in the tournament will step into it and then be turned into an avatar. Their full body composition will come into play to improve the accuracy of the offside's decision-making, meaning it'll be easier for officials and fans to see which body parts cross the line.

17:09In charge of FIFA's innovation is Johannes Holtzmuller. Great that you had the opportunity to come here and also to test some of our technology. Here at the home of football, he ensures that the use of tech and data is kept up to date to improve decision making on and off the pitch. Would moments like the hand of God or the 1966 goal still happen today, given that we've got access to so much more data? It's a very good question and we ask this question ourselves as well. I think it will be different because we still have decisions for interpretation. I think the ones that are factual, here now technology can help and say, OK, the ball has crossed the line or not.

17:50I mean, a very good one was Frank Lampard. So coming from the German side, since I'm German, of course it was a very good experience for us. Looking from the English side, I would say, oh my God. So it was clear for everyone that the ball has crossed the line, but the referee couldn't have a second chance to see it in a replay. While most technology being introduced for this summer's tournament is aimed at improving the accuracy of the game, there is one addition that offers fans the chance to see decision-making from another perspective, the referees. A first for the World Cup this year, the RefCam.

18:25So as you can see, it's basically just mounted on a relatively simple construction or can be attached to the earpiece. it has an additional receiver so it's not considered to be too cumbersome for the referee to wear Obviously pressure is everything in a World Cup it means so much to the players, the fans and those controversial moments have made history in the past to be able to see that in those exchanges is something really special Yeah, absolutely It's often hard as an outsider to judge what's happened and this might just help in an altercation or when players are arguing, it just gives you that real sense of emotions and maybe it just helps you to understand why certain decisions are being made in a certain way.

19:11This technology is now allowing us to really immerse ourselves into the game, which just wasn't possible technically many years ago. So if you're not able to travel to the World Cup to support your team, you're still getting closer to the action on the pitch than ever before. And that's Tech Life's Shona McCallum, always on the ball.

19:39Three years ago on Tech Life, we asked if artificial intelligence could be a force for good. So among others, we spoke to a leading voice in the industry, Gary Marcus, a scientist, bestselling author and AI entrepreneur. Marcus is well known for scrutinising and criticising some of the hype around AI, particularly generative AI. And while we spend a fair bit of time on this programme discussing ideas for how AI can help us, we also think it's good to check in with those concerned about its current direction. I caught up with Gary at an event at the University of Edinburgh in Scotland. I asked him if he was noticing a shift in the public's attitude towards AI.

20:16Was the shine wearing off? There's more evidence of an AI backlash every day. Like just a couple of days ago, Eric Schmidt, who used to be the CEO of Google, was booed giving a commencement address. Like it's a very different picture. Like two years ago, three years ago, everybody was excited about ChatGPT and generative AI. But over time, people have come to see that there are limits, as I kept warning people that there were. Do you think they're feeling threatened? Well, some people are threatened in terms of their jobs. Other people are threatened by data centers that are just sort of ripping up the whole planet.

20:49So there are lots of people that feel threatened in one way or another. You're here in Edinburgh. We're sitting in a room with a magnificent view over the whole city. Unbelievable here. Unbelievable, indeed. To give a take part in a discussion, I suppose we could say. And I wanted to sort of read the topic of the discussion. Scaling compute is AI industry jargon for spending more and more money on data centers and chips inside those data centers. The big tech firms spend nearly a trillion dollars on compute this year. more than a dozen times the cost of the Manhattan Project, will it turn out to be the worst bet in history?

21:24Well, will all this AI spending turn out to be the worst bet in history? Well, so I'm going to give a talk, and that's the argument I'm going to make. I mean, we don't know how it's going to turn out. But what I would say is it's very speculative to spend, you know, a trillion dollars or almost a trillion dollars this year, and almost a trillion dollars next year and so on. To make that work, you have to have like$1.6 trillion a year in revenue from large language models, according to one estimate that I've seen. That means making four times as much as Alphabet made in its biggest year. So NVIDIA that makes the chips that are in the data centers, they're doing great.

22:01I call them the guys who are selling shovels in a gold rush. But the question is how much gold is there? And if you need to make$1.6 trillion a year to make your money back. You need a lot of gold there. Is AI a bubble? I guess that's the easy question. And the hard question, if it's a bubble, when's it going to burst? I often think of Wile E. Coyote in the Bugs Bunny cartoons, where he goes over the side of a cliff, but he doesn't actually fall until he looks down. And so some of this is about psychology. When will people realize that maybe the economics here don't make sense? But they are actually starting to.

22:38The fact that more people are talking about it is part of a shift in the psychology. The fact that there's a backlash is a shift in the psychology. It also means like younger people may not be the customers that they were expecting and so forth. And so I do think eventually it's going to deflate. We will still have large language models, but because every company is making basically the same thing, they'll become cheap, but they won't become a big source of revenue. And a lot of companies are making these data centers and so forth, I think are going to lose money. What I'm worried about is that it's going to get so bad because they borrowed so much money that the banks are going to be in trouble.

23:09And we're going to have a repeat of 2008. And we're going to hear this phrase about too big to fail. Governments are going to bail out these companies and the taxpayers are going to pay for all of this in the end. That's my biggest worry. One of the things you've been a sort of fairly consistent critic of is generative AI. I wanted to read something you wrote about it. It says, when all is said and done, my best guess is that generative AI will have done significantly more harm to society than good. Although there are some practical use cases such as coding, it is an inherently unreliable technology.

23:43It's ripping apart our educational system, our information ecosphere, and flooding the zone with non-consensual deep fake porn. There are some good things about generative AI though, aren't there? I mean, we can talk about scientific discoveries. We can talk about saving people from mundane jobs. that it's not quite as negative as you painted, is it? Well, first notice that in the actual quote that you read, which I would not change a word of, that I actually point out some positives, right? I say coding is a positive, and I think the strongest use case of generative AI is coding. It helps coders write faster.

24:23There are questions there about the security of the code, the stability of the code. In terms of scientific discovery, the best contributions to science have not come from generative AI, which is stuff like chatbots, but from tools like AlphaFold that are purpose-built for very specific problems. So you're making a difference between, say, the kind of technology that would spot early-stage cancers in cancer screening, which is one of the uses that's frequently sort of touted as a positive of AI, and the kind of large language model, the chatbot-type uses, you know, summarize this document, give me 10 ideas for a pitch for a new business project.

25:04That's right. I think the best use cases in medicine are not going to come from chatbots. And we're ripping up our environment to build chatbots. And we're also kind of destroying our social infrastructure for the chatbots. It's really the chatbots that I most dislike in all of this. In principle, I love the idea of using AI for science. That's part of why I went into the field. And AlphaFold is really actually genuinely useful, but it's not a chatbot. What's the risk with AI large language models that worries you most? I mean, people talk about unemployment, people talk about, you know, non-consensual imagery.