WEBVTT 00:00:00.000 --> 00:00:02.040 Okay, so you know how everyone's always saying 00:00:02.040 --> 00:00:04.200 stay ahead of the curve. Like you got to know 00:00:04.200 --> 00:00:06.419 what the next big thing is, right? Well, that's 00:00:06.419 --> 00:00:09.119 what we do here on the deep dive. We take these 00:00:09.119 --> 00:00:11.460 complex topics and break them down so you can 00:00:11.460 --> 00:00:13.839 actually, you know, understand them without getting 00:00:13.839 --> 00:00:16.420 lost in all the technical stuff. Exactly. No 00:00:16.420 --> 00:00:20.539 jargon overload here. Yeah. So today we're diving 00:00:20.539 --> 00:00:23.120 deep into something that honestly sounds like 00:00:23.120 --> 00:00:25.359 it's straight out of science fiction. We're talking 00:00:25.359 --> 00:00:29.379 about the connection between AI, you know, artificial 00:00:29.379 --> 00:00:32.060 intelligence, and something called quantum computing. 00:00:32.560 --> 00:00:35.679 And it's a connection that's way more profound 00:00:35.679 --> 00:00:38.079 than you might think at first glance. We've got 00:00:38.079 --> 00:00:40.179 some really interesting source material here 00:00:40.179 --> 00:00:42.979 suggesting that AI is actually becoming a key 00:00:42.979 --> 00:00:45.799 player in making quantum computers a reality. 00:00:46.500 --> 00:00:48.520 In fact, one of the sources even goes so far 00:00:48.520 --> 00:00:50.420 as to say that some of the most exciting progress 00:00:50.420 --> 00:00:53.140 in this field is coming from what they call a 00:00:53.140 --> 00:00:56.020 relatively unknown approach. And that's what 00:00:56.020 --> 00:00:57.759 really intrigued me, this idea that there's this 00:00:57.759 --> 00:01:00.500 whole new path emerging, and AI is right at the 00:01:00.500 --> 00:01:02.420 heart of it. Yeah, it's pretty mind -blowing 00:01:02.420 --> 00:01:04.120 stuff when you really start to think about the 00:01:04.120 --> 00:01:07.219 implications. I mean, quantum computing, at least 00:01:07.219 --> 00:01:09.700 in theory, has the potential to be absolutely 00:01:09.700 --> 00:01:12.079 revolutionary, right? Like, calculations that 00:01:12.079 --> 00:01:14.579 would take our best supercomputers, you know, 00:01:14.879 --> 00:01:17.379 years, centuries even, quantum computers could 00:01:17.379 --> 00:01:20.060 potentially do in like... minutes. That's the 00:01:20.060 --> 00:01:21.739 promise. And it's a promise that's attracting 00:01:21.739 --> 00:01:23.640 a lot of attention, particularly in fields like 00:01:23.640 --> 00:01:25.819 finance. I mean, think about it, the ability 00:01:25.819 --> 00:01:29.140 to make complex financial calculations at speeds 00:01:29.140 --> 00:01:31.879 that are just unimaginable with current technology. 00:01:32.519 --> 00:01:35.239 It's not hard to see why places like Wells Fargo, 00:01:35.420 --> 00:01:38.739 Citigroup, HSBC, Bank of America, they're all 00:01:38.739 --> 00:01:40.799 pouring resources into this. Yeah. One of our 00:01:40.799 --> 00:01:43.420 sources even quotes this guy, Himim Israel, from 00:01:43.420 --> 00:01:45.840 Bank of America, saying it's bigger than fire. 00:01:46.120 --> 00:01:48.060 It's a pretty strong statement, and it really 00:01:48.060 --> 00:01:50.200 highlights the financial motivation driving a 00:01:50.200 --> 00:01:52.260 lot of this research. Definitely. Though the 00:01:52.260 --> 00:01:54.500 sources do also acknowledge, kind of humorously, 00:01:54.659 --> 00:01:56.700 that there's a lot of hype around quantum computing 00:01:56.700 --> 00:01:58.959 too. Like, some people think it'll magically 00:01:58.959 --> 00:02:01.180 solve all the world's problems overnight. Right, 00:02:01.180 --> 00:02:03.359 like we'll suddenly have cures for all diseases, 00:02:03.780 --> 00:02:05.579 and we'll be able to predict the future with 00:02:05.579 --> 00:02:07.859 perfect accuracy. Or even bring back, you know, 00:02:08.060 --> 00:02:10.379 Elvis. Well, maybe not that. But yeah, there's 00:02:10.379 --> 00:02:12.439 definitely a need to temper some of the more 00:02:12.439 --> 00:02:15.240 outlandish claims. But even if we strip away 00:02:15.240 --> 00:02:18.740 the hype, The core idea of dramatically accelerating 00:02:18.740 --> 00:02:21.719 computations, particularly in fields like finance, 00:02:22.039 --> 00:02:25.539 that's still incredibly powerful. Totally. So 00:02:25.539 --> 00:02:28.020 that's what we want to explore in this deep dive. 00:02:28.840 --> 00:02:32.199 How exactly is AI playing this crucial role in 00:02:32.199 --> 00:02:34.819 making quantum computers work? And specifically, 00:02:34.819 --> 00:02:36.900 we're going to focus on one of the biggest hurdles 00:02:36.900 --> 00:02:40.740 in this field, errors. So why are errors such 00:02:40.740 --> 00:02:43.349 a massive problem in quantum computing? Well, 00:02:43.469 --> 00:02:44.949 to understand that, you have to understand a 00:02:44.949 --> 00:02:46.650 little bit about how quantum computers work. 00:02:47.050 --> 00:02:48.689 They rely on these things called qubits, which 00:02:48.689 --> 00:02:50.830 are like the quantum equivalent of bits in a 00:02:50.830 --> 00:02:53.069 regular computer. But the thing about qubits 00:02:53.069 --> 00:02:56.090 is that they're incredibly fragile. They're super 00:02:56.090 --> 00:02:58.110 sensitive to any kind of disturbance from their 00:02:58.110 --> 00:03:00.349 environment. And the more qubits you have in 00:03:00.349 --> 00:03:02.530 a system, the more likely they are to be affected 00:03:02.530 --> 00:03:04.870 by these disturbances, which leads to errors 00:03:04.870 --> 00:03:08.069 in the calculations. So it's like... The more 00:03:08.069 --> 00:03:10.469 complex the calculation, the more likely it is 00:03:10.469 --> 00:03:12.629 to get messed up by these tiny little errors. 00:03:13.030 --> 00:03:14.969 Exactly. And that's been one of the biggest challenges 00:03:14.969 --> 00:03:16.870 in building larger and more powerful quantum 00:03:16.870 --> 00:03:19.289 computers. Because if you can't control the errors, 00:03:19.830 --> 00:03:21.930 the results become unreliable. It's like trying 00:03:21.930 --> 00:03:24.509 to build a house of cards in a windstorm. OK. 00:03:24.689 --> 00:03:28.050 That makes sense. So how does AI come into play 00:03:28.050 --> 00:03:30.909 here? Is it actually able to fix these errors? 00:03:31.250 --> 00:03:33.969 It's not so much about fixing the errors directly 00:03:33.969 --> 00:03:36.349 as it is about learning to predict and manage 00:03:36.349 --> 00:03:39.039 them. And one of the most exciting examples of 00:03:39.039 --> 00:03:42.219 this is the work being done by Google AI. They've 00:03:42.219 --> 00:03:44.879 developed this system called Alpha Qubit that 00:03:44.879 --> 00:03:48.020 uses AI to reduce errors in their superconducting 00:03:48.020 --> 00:03:50.270 qubits. So tell me more about that. Superconducting 00:03:50.270 --> 00:03:52.370 qubits, what are those exactly? Well, they're 00:03:52.370 --> 00:03:54.870 basically tiny electrical circuits that operate 00:03:54.870 --> 00:03:57.050 at extremely low temperatures, like close to 00:03:57.050 --> 00:03:59.509 absolute zero. And at those temperatures, they 00:03:59.509 --> 00:04:01.789 start to exhibit these quantum mechanical properties 00:04:01.789 --> 00:04:04.949 that make them suitable for use as qubits. OK, 00:04:05.110 --> 00:04:09.159 so Google is using AI. to work with these superconducting 00:04:09.159 --> 00:04:11.740 qubits. How does that actually work? Are they 00:04:11.740 --> 00:04:14.379 like training the AI on how these qubits behave? 00:04:14.539 --> 00:04:16.720 Exactly. They've trained their AI algorithms 00:04:16.720 --> 00:04:19.660 on a massive data set collected from their Sycamore 00:04:19.660 --> 00:04:21.560 chip, which is one of their quantum processors 00:04:21.560 --> 00:04:24.519 that uses these superconducting qubits. And by 00:04:24.519 --> 00:04:27.620 analyzing this data, the AI learns to identify 00:04:27.620 --> 00:04:30.509 patterns. in the errors that occur. So it's like 00:04:30.509 --> 00:04:32.649 the AI is studying the behavior of these qubits 00:04:32.649 --> 00:04:34.910 and figuring out what causes them to mess up. 00:04:34.970 --> 00:04:37.730 Right. And once the AI has this understanding, 00:04:38.470 --> 00:04:41.089 it can then use that knowledge to provide active 00:04:41.089 --> 00:04:43.490 feedback to the quantum system. It's like having 00:04:43.490 --> 00:04:46.029 a really smart assistant constantly monitoring 00:04:46.029 --> 00:04:48.250 the qubits and making tiny adjustments to keep 00:04:48.250 --> 00:04:50.250 them on track. So it's like anticipating the 00:04:50.250 --> 00:04:52.410 errors before they even happen. In a way, yes. 00:04:52.589 --> 00:04:54.949 The AI is constantly analyzing the state of the 00:04:54.949 --> 00:04:57.310 qubits and making subtle adjustments to the control 00:04:57.310 --> 00:04:59.410 signals that govern them. And the result is that 00:04:59.410 --> 00:05:01.529 the error rate actually goes down. That's pretty 00:05:01.529 --> 00:05:04.529 amazing. So how much does this AI actually improve 00:05:04.529 --> 00:05:06.310 the error rate? Like, does it eliminate the errors 00:05:06.310 --> 00:05:08.949 completely? Not completely, but it does make 00:05:08.949 --> 00:05:11.730 a significant difference. Our source even mentions 00:05:11.730 --> 00:05:14.089 that there are these red curves in their data 00:05:14.089 --> 00:05:16.810 that show the improvement achieved by the AI. 00:05:17.129 --> 00:05:18.930 But it doesn't sound like the AI is actually 00:05:19.580 --> 00:05:22.019 reducing the fundamental error rate of the qubits 00:05:22.019 --> 00:05:24.300 themselves, right? It's more about mitigating 00:05:24.300 --> 00:05:26.420 the effects of those errors. That's an important 00:05:26.420 --> 00:05:29.000 point. The AI isn't changing the inherent properties 00:05:29.000 --> 00:05:31.660 of the qubits, but what it's doing is making 00:05:31.660 --> 00:05:34.680 the system more robust and reliable. And that's 00:05:34.680 --> 00:05:36.699 really important because it means that you can 00:05:36.699 --> 00:05:39.319 perform longer and more complex calculations 00:05:39.319 --> 00:05:42.000 without the errors completely overwhelming the 00:05:42.000 --> 00:05:44.839 results. I see. So it's not a magic bullet, but 00:05:44.839 --> 00:05:47.439 it's a significant step forward in making quantum 00:05:47.439 --> 00:05:50.899 computers more practical. Absolutely. And what's 00:05:50.899 --> 00:05:52.519 particularly exciting about Google's approach 00:05:52.519 --> 00:05:55.050 is that it seems to be very generalizable. This 00:05:55.050 --> 00:05:57.550 means that the AI's ability to reduce errors 00:05:57.550 --> 00:06:00.709 works not just on the specific quantum states 00:06:00.709 --> 00:06:03.290 it was trained on, but also on new states that 00:06:03.290 --> 00:06:05.470 it's never seen before. And why is that important? 00:06:05.930 --> 00:06:07.990 Because it means that the AI's error correction 00:06:07.990 --> 00:06:11.290 capabilities can potentially scale up as quantum 00:06:11.290 --> 00:06:13.730 computers become larger and more complex. And 00:06:13.730 --> 00:06:16.019 that's crucial. if we want these machines to 00:06:16.019 --> 00:06:18.779 be able to tackle real -world problems. That 00:06:18.779 --> 00:06:20.939 makes sense. But you mentioned earlier that you're 00:06:20.939 --> 00:06:23.560 personally a bit skeptical about superconducting 00:06:23.560 --> 00:06:25.660 qubits being the best way to build quantum computers. 00:06:25.800 --> 00:06:28.720 Why is that? Well, it has to do with the inherent 00:06:28.720 --> 00:06:32.279 challenges in scaling up these systems. Superconducting 00:06:32.279 --> 00:06:35.769 qubits are incredibly sensitive to noise and 00:06:35.769 --> 00:06:38.629 disturbances. And maintaining the right conditions 00:06:38.629 --> 00:06:41.430 for them to operate properly becomes increasingly 00:06:41.430 --> 00:06:44.209 difficult as you add more qubits. So it's like 00:06:44.209 --> 00:06:46.569 a logistical nightmare trying to keep all these 00:06:46.569 --> 00:06:48.970 qubits happy and working together. Pretty much. 00:06:49.290 --> 00:06:51.209 And that's why some researchers are looking at 00:06:51.209 --> 00:06:53.629 alternative approaches to building quantum computers. 00:06:54.149 --> 00:06:56.170 And one of the most promising alternatives is 00:06:56.170 --> 00:06:58.970 something called neutral atom arrays, or as it's 00:06:58.970 --> 00:07:01.959 sometimes called, atoms in tweezers. atoms and 00:07:01.959 --> 00:07:04.339 tweezers. That sounds like something out of a, 00:07:04.339 --> 00:07:05.899 well, a science fiction movie. How does that 00:07:05.899 --> 00:07:08.519 even work? It's actually a really elegant approach. 00:07:08.939 --> 00:07:12.800 It involves using lasers to trap individual atoms 00:07:12.800 --> 00:07:16.560 in these very precise ordered arrangements. Each 00:07:16.560 --> 00:07:18.920 of these atoms can then act as a qubit. So you're 00:07:18.920 --> 00:07:21.480 basically creating this perfectly regular lattice 00:07:21.480 --> 00:07:23.879 of atoms, all held in place by these laser beams. 00:07:24.160 --> 00:07:26.240 So instead of these delicate little electrical 00:07:26.240 --> 00:07:28.980 circuits, you're using actual atoms as your qubits. 00:07:29.360 --> 00:07:32.120 Exactly. And that has some really significant 00:07:32.120 --> 00:07:35.040 advantages. For one thing, atoms are inherently 00:07:35.040 --> 00:07:37.459 identical, so you don't have to worry about variations 00:07:37.459 --> 00:07:39.879 in their behavior like you do with superconducting 00:07:39.879 --> 00:07:42.680 qubits. And because atoms are so incredibly small, 00:07:43.040 --> 00:07:44.800 you can pack a lot more of them into a given 00:07:44.800 --> 00:07:47.360 space, which means you can potentially build 00:07:47.360 --> 00:07:49.860 much larger and more powerful quantum computers. 00:07:50.139 --> 00:07:52.720 Okay, so that makes sense. But why haven't we 00:07:52.720 --> 00:07:54.600 been building quantum computers this way all 00:07:54.600 --> 00:07:57.240 along? If it's so much better, why did everyone 00:07:57.240 --> 00:07:59.300 focus on superconducting qubits in the first 00:07:59.300 --> 00:08:02.860 place? Well, it all comes down to control. Manipulating 00:08:02.860 --> 00:08:05.800 individual atoms with such precision is incredibly 00:08:05.800 --> 00:08:08.500 challenging, but that's where AI is starting 00:08:08.500 --> 00:08:11.199 to play a major role again. It's enabling us 00:08:11.199 --> 00:08:13.779 to achieve levels of control that were simply 00:08:13.779 --> 00:08:17.160 unimaginable before. So how is AI being used 00:08:17.160 --> 00:08:20.519 in this atoms and tweezers approach? One of the 00:08:20.519 --> 00:08:22.819 biggest breakthroughs has been using AI to precisely 00:08:22.819 --> 00:08:25.480 position and control these atoms. It's actually 00:08:25.480 --> 00:08:27.860 pretty similar to how Google is using AI for 00:08:27.860 --> 00:08:30.120 error correction with superconducting qubits. 00:08:30.279 --> 00:08:33.240 The AI is trained on data about how the atoms 00:08:33.240 --> 00:08:36.200 behave, and then it uses that knowledge to control 00:08:36.200 --> 00:08:38.440 the lasers that are manipulating the atoms. So 00:08:38.440 --> 00:08:41.159 it's like the AI is learning to be this incredibly 00:08:41.159 --> 00:08:44.379 skilled atomic puppeteer. Exactly. And the results 00:08:44.379 --> 00:08:46.220 have been pretty remarkable. One of our sources 00:08:46.220 --> 00:08:49.340 even describes this amazing atom tomography image 00:08:49.340 --> 00:08:52.519 that shows over 2 ,000 atoms arranged in this 00:08:52.519 --> 00:08:55.379 almost perfectly ordered array. 2 ,000 atoms? 00:08:55.440 --> 00:08:57.379 Wow, that's more than double the number of qubits 00:08:57.379 --> 00:08:59.980 in IBM's largest quantum computer, right? That's 00:08:59.980 --> 00:09:01.860 right. And it really highlights the potential 00:09:01.860 --> 00:09:03.980 scalability of this approach. But I guess the 00:09:03.980 --> 00:09:05.820 big question is, have they actually done any 00:09:05.820 --> 00:09:08.360 calculations with this 2000 -atom system yet? 00:09:08.519 --> 00:09:11.440 Not yet. The main achievement so far has been 00:09:11.440 --> 00:09:13.559 demonstrating this incredible level of control 00:09:13.559 --> 00:09:16.659 and the ability to create these large, well -ordered 00:09:16.659 --> 00:09:19.720 arrays. But the progress in this area has been 00:09:19.720 --> 00:09:22.500 really rapid. And a lot of experts believe that 00:09:22.500 --> 00:09:24.279 performing meaningful calculations with these 00:09:24.279 --> 00:09:27.179 systems is just around the corner. So it's still 00:09:27.179 --> 00:09:29.559 early days, but there's a lot of excitement about 00:09:29.559 --> 00:09:32.840 the potential of this atoms and tweezers approach. 00:09:33.139 --> 00:09:35.480 And it seems like AI is playing a crucial role 00:09:35.480 --> 00:09:38.080 in making it all happen. Absolutely. And what's 00:09:38.080 --> 00:09:40.299 even more intriguing is that AI's involvement 00:09:40.299 --> 00:09:43.100 in quantum computing isn't just limited to these 00:09:43.100 --> 00:09:45.720 practical engineering challenges. It's also starting 00:09:45.720 --> 00:09:47.600 to raise some really profound questions about 00:09:47.600 --> 00:09:49.840 the fundamental nature of reality itself. OK, 00:09:49.840 --> 00:09:51.399 now you're really piquing my curiosity. What 00:09:51.399 --> 00:09:53.360 do you mean? Well, it has to do with the concept 00:09:53.360 --> 00:09:56.379 of quantum randomness. One of the core principles 00:09:56.379 --> 00:09:58.799 of quantum mechanics is that certain aspects 00:09:58.799 --> 00:10:01.120 of the quantum world are inherently unpredictable. 00:10:01.659 --> 00:10:03.799 You can't know for sure what the outcome of a 00:10:03.799 --> 00:10:06.419 measurement will be, only the probability of 00:10:06.419 --> 00:10:08.120 different outcomes. Right. I remember learning 00:10:08.120 --> 00:10:10.379 about that in physics class. It's like the universe 00:10:10.379 --> 00:10:13.259 is playing dice at the subatomic level. Exactly. 00:10:13.759 --> 00:10:16.620 But here's where it gets interesting. If AI can 00:10:16.620 --> 00:10:19.320 become really good at predicting and controlling 00:10:19.320 --> 00:10:22.220 the behavior of quantum systems, even better 00:10:22.220 --> 00:10:24.360 than our current theory suggests is possible. 00:10:24.649 --> 00:10:27.870 It might imply that what we perceive as randomness 00:10:27.870 --> 00:10:30.950 isn't truly fundamental. It might just be a reflection 00:10:30.950 --> 00:10:33.490 of our limited understanding. So you're saying 00:10:33.490 --> 00:10:35.789 that AI might be able to see patterns in order 00:10:35.789 --> 00:10:38.269 in the quantum world that we're currently blind 00:10:38.269 --> 00:10:41.090 to. That's one possibility. And if that's the 00:10:41.090 --> 00:10:43.629 case, it could have some pretty profound implications 00:10:43.629 --> 00:10:45.710 for our understanding of how the universe works. 00:10:46.269 --> 00:10:48.629 It's like, if AI can predict the role of the 00:10:48.629 --> 00:10:51.110 dice, maybe the dice aren't really random after 00:10:51.110 --> 00:10:54.200 all. Wow, that's a mind -blowing thought. So 00:10:54.200 --> 00:10:57.120 not only is AI helping us to build better quantum 00:10:57.120 --> 00:10:59.879 computers, it might also be giving us a whole 00:10:59.879 --> 00:11:02.299 new way of looking at the universe. It's certainly 00:11:02.299 --> 00:11:04.799 a possibility that's worth considering, and it 00:11:04.799 --> 00:11:07.399 really highlights the incredible potential of 00:11:07.399 --> 00:11:10.740 this convergence between AI and quantum computing. 00:11:11.080 --> 00:11:14.019 It's a field that's moving incredibly fast, and 00:11:14.019 --> 00:11:16.139 the discoveries being made are challenging some 00:11:16.139 --> 00:11:18.440 of our most fundamental assumptions about the 00:11:18.440 --> 00:11:20.799 nature of reality. Okay, so let's recap what 00:11:20.649 --> 00:11:23.350 we've learned in this deep dive. AI is playing 00:11:23.350 --> 00:11:26.250 a crucial role in advancing the field of quantum 00:11:26.250 --> 00:11:29.269 computing. It's helping to overcome the significant 00:11:29.269 --> 00:11:32.250 challenges of errors, and it's also enabling 00:11:32.250 --> 00:11:35.409 the development of these new, more scalable architectures, 00:11:35.570 --> 00:11:38.690 like neutral atom arrays. Exactly. And it's all 00:11:38.690 --> 00:11:40.870 happening much faster than anyone anticipated. 00:11:41.289 --> 00:11:43.210 It's safe to say that quantum computing is going 00:11:43.210 --> 00:11:45.730 to be a major force in the near future, and AI 00:11:45.730 --> 00:11:47.909 is going to be right there at the forefront driving 00:11:47.909 --> 00:11:50.480 the progress. So one last thought for our listeners 00:11:50.480 --> 00:11:54.299 to ponder. If AI can become so adept at controlling 00:11:54.299 --> 00:11:57.500 and manipulating quantum systems, what does that 00:11:57.500 --> 00:11:59.740 say about the limits of our current understanding 00:11:59.740 --> 00:12:02.139 of the universe? Is it possible that what we 00:12:02.139 --> 00:12:04.179 perceive as fundamental randomness is actually 00:12:04.179 --> 00:12:06.580 just a complex tapestry of order that we haven't 00:12:06.580 --> 00:12:09.419 yet learned to decipher? And if AI can unlock 00:12:09.419 --> 00:12:11.940 those secrets, what other incredible possibilities 00:12:11.940 --> 00:12:14.139 might it reveal? Thanks for joining us on this 00:12:14.139 --> 00:12:16.240 deep dive. It's been a pleasure. Till next time.