WEBVTT - What Happens When Quantum Computing Actually Works? 0:00:02.720 --> 0:00:11.920 Bloomberg Audio Studios, Podcasts, Radio News. 0:00:18.000 --> 0:00:20.079 Welcome to Marion Talk's Money, the podcasts in which people 0:00:20.079 --> 0:00:23.599 who know the markets explain the markets. I'm Maren Sumset Web. 0:00:23.920 --> 0:00:26.079 This week I was begining with Daniel don Steinberg, co 0:00:26.200 --> 0:00:28.560 founder and partner at eden Base, an investment firms that 0:00:28.640 --> 0:00:32.000 invests in frontier deep tech. I invited Daniel on the 0:00:32.040 --> 0:00:34.640 show because I had a really interesting conversation with him 0:00:34.640 --> 0:00:38.480 about quantum computing and just how close it might be. 0:00:38.760 --> 0:00:41.599 During our conversation, we talk about ai future of and 0:00:41.640 --> 0:00:44.960 what it really is. We also talk obviously about quantum 0:00:45.000 --> 0:00:48.920 computing and what it might mean for our futures, and 0:00:49.000 --> 0:00:53.080 we also cover exactly how one might deal with the 0:00:53.120 --> 0:00:59.120 future that quantum might bring. Spoiler alert read some philosophy. Daniel, 0:00:59.200 --> 0:01:01.160 Welcome to marind Money Pleasure. 0:01:01.240 --> 0:01:02.000 Thank you for having me. 0:01:02.600 --> 0:01:05.399 Will you tell me a little bit about what Edenbased 0:01:05.480 --> 0:01:06.600 does and then we'll take you from there. 0:01:07.120 --> 0:01:10.319 Well, in Bases, three partners got together, all with very 0:01:10.360 --> 0:01:14.280 different backgrounds, but sharing a common theme, which we were 0:01:14.319 --> 0:01:18.399 all entrepreneurs and we wanted to invest our own money 0:01:18.640 --> 0:01:22.000 but also create a platform for two things. One to 0:01:22.600 --> 0:01:25.440 allow us to hire a bigger team to make decisions 0:01:25.440 --> 0:01:28.720 in a world that we thought was becoming incredibly complicated. 0:01:29.200 --> 0:01:32.920 And secondly, one of my partners created Technician for David 0:01:32.959 --> 0:01:37.280 Cameron level thirty nine, which is the fintech ecosystem. And 0:01:37.560 --> 0:01:40.160 when I really explored what he had done, I realized 0:01:40.200 --> 0:01:44.240 that ecosystems solved a few of the challenges I saw 0:01:44.319 --> 0:01:46.920 with what was about to happen. And so Edenbase is 0:01:47.000 --> 0:01:50.640 a is an investment vehicle and also an ecosystem where 0:01:50.640 --> 0:01:54.840 we help with adoption of new technologies. We call it 0:01:55.000 --> 0:01:57.160 investing in a future we can be proud of, which 0:01:57.200 --> 0:02:00.200 actually came out of AI weirdly in our three years go. 0:02:00.640 --> 0:02:04.320 It's not sustainability, it's a future we can be proud of. Okay. 0:02:04.680 --> 0:02:07.920 I like that. That's better than sustainability. Is so vague, 0:02:07.960 --> 0:02:10.560 isn't it? Okay? So the thing I really wanted to 0:02:10.560 --> 0:02:11.960 talk to you about it and I think I invited 0:02:12.000 --> 0:02:14.760 you want to talk about, was quantum computing. But before 0:02:14.760 --> 0:02:16.120 we get onto that, I wanted to talk to you 0:02:16.120 --> 0:02:18.280 a little bit about AI because when we were chatting 0:02:18.280 --> 0:02:19.400 the other day, right you and I and we were 0:02:19.400 --> 0:02:21.920 talking about various things, and I said to you, is 0:02:22.160 --> 0:02:25.520 AI the correct name at all for what we think 0:02:25.560 --> 0:02:27.800 of as AI, because what we're using at the moment 0:02:27.919 --> 0:02:31.239 is lllm's Large language models. And anyone listening who isn't 0:02:31.400 --> 0:02:34.359 entirely sure how LMS work, please go back and listen 0:02:34.440 --> 0:02:36.840 to some of our previous podcasts if you've had lengthy 0:02:36.840 --> 0:02:39.960 conversations about all these things. So when we talk about AI, 0:02:40.280 --> 0:02:43.880 we're talking about these large language models, and they're not artificial, 0:02:44.000 --> 0:02:47.639 they're man made, and they're also not really intelligent, are they. 0:02:47.680 --> 0:02:50.800 So we we're using the wrong phrase in the beginning. 0:02:51.520 --> 0:02:54.560 I think we use the wrong phrase because that's how 0:02:55.000 --> 0:02:59.840 we referred to technologies such as these, which are we 0:03:00.160 --> 0:03:06.079 building artificial intelligence And the history of compute, certainly for 0:03:06.120 --> 0:03:11.440 the last twenty thirty years, has been exponential, and m's 0:03:11.680 --> 0:03:15.560 are a continuation of that exponential and therefore they fit 0:03:15.639 --> 0:03:18.600 into the bucket of artificial intelligence. But you're right, the 0:03:18.600 --> 0:03:21.640 word artificial is not correct and the word intelligence is 0:03:21.680 --> 0:03:24.520 not correct for what we have built. We have built 0:03:24.919 --> 0:03:28.799 incredibly powerful technology. I mean, all the technologies we've built 0:03:28.840 --> 0:03:31.440 have been incredibly powerful at the time. This happens to 0:03:31.520 --> 0:03:37.120 be even more so, and it's an opportunity and threat 0:03:37.480 --> 0:03:39.880 for what we have as our society and the way 0:03:39.920 --> 0:03:42.920 we work is enormous. But yes, you're right, they don't 0:03:42.960 --> 0:03:46.480 fit what we would classify as intelligence, which is a 0:03:46.640 --> 0:03:53.480 biological characteristic. These are programmed characteristics. The way they mimic 0:03:53.560 --> 0:03:57.720 intelligence is they have access to an enormous amount of data, 0:03:57.880 --> 0:04:02.680 most of which was generated by humans, and therefore the 0:04:02.720 --> 0:04:08.920 ability to write cogent arguments based on human intelligence is 0:04:08.920 --> 0:04:09.640 what they do. 0:04:10.960 --> 0:04:13.040 I have thee conversation with one of my nieces the 0:04:13.080 --> 0:04:15.960 other day about how you can tell what is Ai 0:04:16.000 --> 0:04:19.440 writing and what isn't Ai writing? And I was explaining 0:04:19.440 --> 0:04:22.279 to her that, of course, the way that Ai writes, 0:04:22.320 --> 0:04:24.279 the way that an LM write is actually just the 0:04:24.279 --> 0:04:28.560 way that old people write because they're using our writing. 0:04:29.040 --> 0:04:31.680 There used to be a hallmark of my writing style was, 0:04:31.720 --> 0:04:35.120 you know, short sentences and dashes, This is how I write. 0:04:35.240 --> 0:04:37.919 So is it Ai writing or is it someone in 0:04:37.960 --> 0:04:38.600 their fifties? 0:04:38.800 --> 0:04:40.760 Well, a good question. You know, it has access to 0:04:40.800 --> 0:04:44.480 a lot of information, and one advantage I have I'm 0:04:44.520 --> 0:04:46.240 writing a second book at the moment is the fact 0:04:46.240 --> 0:04:49.120 that I've written a first book. It knows my first book, 0:04:49.360 --> 0:04:51.960 and so when I ask it to write in my 0:04:52.120 --> 0:04:55.680 and my co author's style, it can copy that. However, 0:04:55.720 --> 0:04:57.560 I do find the style we wrote the first book 0:04:57.560 --> 0:04:59.279 in is not the what style we writing the second 0:04:59.279 --> 0:05:03.680 book in, but it can mimic lots of styles. Whilst 0:05:03.960 --> 0:05:08.240 it is certainly based on the data that we have, 0:05:08.640 --> 0:05:11.839 it is able to create new things, and so if 0:05:11.880 --> 0:05:15.920 you explain exactly what you want from it, it will 0:05:15.960 --> 0:05:18.920 do pretty much exactly what you do, as long as 0:05:18.920 --> 0:05:21.320 that's within a framework that it can understand. 0:05:22.040 --> 0:05:24.920 When you look at where we are at the moment 0:05:25.080 --> 0:05:29.560 with llms, and we think about, say, for example, the 0:05:29.600 --> 0:05:32.280 legal profession, and that's the place where everyone says, well, 0:05:32.400 --> 0:05:35.000 lllms will fully replace lawyers in no time at all. 0:05:35.800 --> 0:05:37.560 How do you see it playing out? 0:05:37.680 --> 0:05:41.440 Well, I think that's a very plausible scenario. Going back 0:05:41.440 --> 0:05:43.760 to my book that are in twenty twenty and twenty one, 0:05:43.960 --> 0:05:47.600 we use the legal profession as something that llms are 0:05:47.680 --> 0:05:51.240 about to wreak, havocate, as they are in accounting, as 0:05:51.279 --> 0:05:53.680 they are in management consultancy and a lot of those 0:05:54.160 --> 0:05:57.840 rule based roles. The reason that will happen is, I 0:05:57.880 --> 0:06:01.159 think for two reasons. The first is that even though 0:06:01.520 --> 0:06:04.320 you may argue or not whether it is better than 0:06:04.360 --> 0:06:07.320 the best lawyer. It certainly has access to more information 0:06:07.400 --> 0:06:09.960 than the best lawyer, because however much I read, I've 0:06:10.000 --> 0:06:11.839 got one hundred and eighty six hours a week, I 0:06:11.920 --> 0:06:15.599 can't read everything. The second thing, and we talk about this, 0:06:15.960 --> 0:06:20.200 which is, firstly, if you're a legal firm today, why 0:06:20.200 --> 0:06:23.480 would you hire junior lawyers to do things that lllms 0:06:23.480 --> 0:06:27.000 can do? And if you do, it's almost like you're 0:06:27.000 --> 0:06:29.480 doing it for the future of the industry rather than 0:06:29.560 --> 0:06:32.600 your personal success, because you might as well replace it. 0:06:32.640 --> 0:06:35.560 And we say, if that happens, then in fifteen twenty 0:06:35.640 --> 0:06:38.320 years there are no junior lawyers. The second thing is, 0:06:38.680 --> 0:06:42.960 as we become more LLLM focused, the only thing that 0:06:43.000 --> 0:06:47.600 can unpick what lllms are writing eventually becomes another LLM, 0:06:47.800 --> 0:06:52.599 and therefore you migrate to that. So if the court system, 0:06:52.680 --> 0:06:57.960 for instance, decides to replace certain parts of it with AI, 0:06:58.279 --> 0:07:02.960 then people will be writing AI to deal with the AI, 0:07:03.320 --> 0:07:05.960 and eventually we will have a different legal system which 0:07:05.960 --> 0:07:09.560 will be based on what computers are able to comprehend, 0:07:09.840 --> 0:07:12.880 like a code for a computer, rather than what humans 0:07:12.880 --> 0:07:15.400 are going to be able to understand. I see that 0:07:15.480 --> 0:07:18.240 as as very likely the other thing that we find, 0:07:18.840 --> 0:07:21.160 and I've been enforcing this at even Base for a 0:07:21.160 --> 0:07:24.360 long time, is we do not send contracts to lawyers 0:07:24.400 --> 0:07:25.920 if we can do it ourselves. I used to do 0:07:26.000 --> 0:07:28.520 that with my first startup many years ago. So what 0:07:28.640 --> 0:07:31.240 it meant is our legal bin is probably five percent 0:07:31.320 --> 0:07:33.480 of what it used to be. And that is also 0:07:33.560 --> 0:07:35.480 a problem for the legal industry. 0:07:35.560 --> 0:07:38.760 As long as you are fully confident in your LLM. 0:07:39.120 --> 0:07:43.720 It depends on the risk factor. So an example is 0:07:43.760 --> 0:07:45.840 I invested in a Spanish company two and a half 0:07:45.880 --> 0:07:48.720 years ago and I was putting a small investment in. 0:07:48.800 --> 0:07:51.440 It wasn't worth sending to a lawyer, but I put 0:07:51.480 --> 0:07:54.800 it through an LM and it took the Spanish contract. 0:07:54.840 --> 0:07:56.160 And this is two and a half years ago, so 0:07:56.160 --> 0:07:57.600 they weren't as good as they out there, and it 0:07:57.640 --> 0:08:01.080 gave me a plausible scenario not have been all right, 0:08:01.320 --> 0:08:04.280 and I was able to rather just signing it and saying, well, 0:08:04.320 --> 0:08:06.560 I don't care. I was able to write to them 0:08:06.560 --> 0:08:09.560 with a few questions and they answered those, so it 0:08:09.640 --> 0:08:12.480 was better than doing nothing. And as they get better, 0:08:12.800 --> 0:08:18.760 your risk appetite becomes greater. Secondly, the risk appetite changes 0:08:19.400 --> 0:08:23.120 as you are able to deal with things yourself, and 0:08:23.160 --> 0:08:24.600 I think that's really powerful. 0:08:25.680 --> 0:08:30.120 So what are the limits to the LM model. And 0:08:30.160 --> 0:08:31.760 one of the things that we've been looking at, and 0:08:31.760 --> 0:08:35.080 we've done a few podcasts on this is the idea 0:08:35.320 --> 0:08:39.760 that you can use an SLM, a small language model 0:08:39.840 --> 0:08:42.400 instead of an LLM, and a couple of years ago 0:08:42.679 --> 0:08:45.480 that didn't work so well. But the new study at 0:08:45.520 --> 0:08:49.440 from Stanford University and someone at pameel Iberium has been 0:08:49.480 --> 0:08:52.320 through it, pointing out that these small language models that 0:08:52.320 --> 0:08:54.319 you can run on your laptop, you can even run 0:08:54.320 --> 0:08:57.199 it on your phone, and you compare the performance of 0:08:57.280 --> 0:09:00.280 those with an LLM. The LMS, of course what we 0:09:00.320 --> 0:09:03.280 need the massive data centers for you compare these and 0:09:03.320 --> 0:09:06.000 in fact the results are pretty much as good, pretty 0:09:06.040 --> 0:09:09.480 much as good. So if you want a fairly straightforward 0:09:10.080 --> 0:09:13.800 AI model, and you want it to be private and 0:09:13.840 --> 0:09:16.679 contained in your own laptop or computer, and you want 0:09:16.679 --> 0:09:19.080 to use less energy, which is also than crucial, then 0:09:19.120 --> 0:09:21.160 you can just have an SLM on your laptop. You 0:09:21.160 --> 0:09:24.840 don't need an LLLM, which rather suggests that in some 0:09:24.880 --> 0:09:27.080 parts we might be going down the wrong path with 0:09:28.080 --> 0:09:30.800 full focus on LM's massive data centers, etc. 0:09:31.559 --> 0:09:33.800 Well, I haven't read that report, so I can't send 0:09:33.800 --> 0:09:37.080 it to you comment particularly on that. Now, what I 0:09:37.120 --> 0:09:40.320 can say is two things. One is straight lines are rare. 0:09:40.720 --> 0:09:44.440 So when I was starting to code many many years 0:09:44.440 --> 0:09:48.880 ago in the derivative trading industry, I've built derivative trade technology, 0:09:48.880 --> 0:09:51.640 and the early aunties we worked on mainframes and the 0:09:51.679 --> 0:09:54.120 only people that had desktops and the people that wrote letters. 0:09:54.480 --> 0:09:58.120 And within five to ten years as I finished my career, 0:09:58.160 --> 0:10:01.600 before I started my first startup, everybody had a desktop 0:10:02.000 --> 0:10:05.880 and you were doing everything locally. And when cloud came along, 0:10:06.000 --> 0:10:08.679 Actually I'm not bad at reading the future, I was surprised. 0:10:08.720 --> 0:10:10.760 I said, why do people need the cloud? You know 0:10:10.840 --> 0:10:13.400 we can do it all locally. And then what I 0:10:14.040 --> 0:10:18.400 failed to recognize was the onset of mobile phones, and 0:10:18.440 --> 0:10:21.079 mobile phones had limited power, and therefore the cloud was 0:10:21.160 --> 0:10:25.360 very useful. And so there are phases in which SLMs 0:10:25.440 --> 0:10:27.760 will be very, very useful. But I don't see that 0:10:27.840 --> 0:10:33.360 necessarily meaning that cloud huge cloud data centers aren't valuable. 0:10:33.640 --> 0:10:37.120 The other thing, of course, is as we build new technologies, 0:10:37.440 --> 0:10:42.440 and SLMs are based on the current LLLM technology, which 0:10:42.480 --> 0:10:45.240 is based on binary compute, and it is based on 0:10:45.320 --> 0:10:48.440 GPUs and it is based on existing storage. A mobile 0:10:48.760 --> 0:10:52.000 phone could do that. And the segue into quantum for instance, 0:10:52.320 --> 0:10:56.760 is a quantum computer is not a binary system. Daniel 0:10:57.400 --> 0:11:02.880 explain quantum computing to me, quantum computing is a break 0:11:02.920 --> 0:11:05.560 from the past. That's the first thing, and the sort 0:11:05.559 --> 0:11:07.960 of the most comprehensive way to describe it is. One 0:11:08.040 --> 0:11:12.800 hundred years ago mathematicians created quantum mechanics and quantum physics, 0:11:12.920 --> 0:11:16.520 and it described the way the universe operated extremely well, 0:11:16.800 --> 0:11:19.280 better than they had before, and they were able to 0:11:19.280 --> 0:11:25.040 write beautiful mathematics and described how subatomic particles worked. In 0:11:25.120 --> 0:11:29.200 nineteen seventies, Richard Feynman the physicists suggested that we could 0:11:29.360 --> 0:11:34.000 actually control them, and then in the eighties IBM started 0:11:34.040 --> 0:11:38.160 trying to control them. And what that meant is they 0:11:38.200 --> 0:11:43.720 were able to instead of operate binary bits i e. 0:11:44.080 --> 0:11:47.720 Ones and zeros, they were able to operate in effect 0:11:47.960 --> 0:11:53.160 probabilistic bits, so they were all combinations of ones and zeros. 0:11:53.480 --> 0:11:56.480 And what that meant was we were able to do 0:11:56.559 --> 0:12:00.280 two types of compute or in theory, because no one's 0:12:00.280 --> 0:12:04.439 built one that is big enough to do anything really meaningful, 0:12:04.679 --> 0:12:07.400 or until very recently, they haven't. It does two types 0:12:07.440 --> 0:12:10.800 of things that binary computer is very bad at. The 0:12:10.880 --> 0:12:14.760 vers is optimization. So imagine a good example might be 0:12:14.800 --> 0:12:17.120 a table plan. If you have five people sitting around 0:12:17.200 --> 0:12:19.600 a table, you have one hundred and twenty five combinations. 0:12:19.600 --> 0:12:22.360 If you have ten, you have three million. If you 0:12:22.520 --> 0:12:26.000 have twenty five people, you have one with twenty five 0:12:26.160 --> 0:12:27.880 zeros on the end of it a combination. 0:12:28.240 --> 0:12:30.480 This is why it's so hard to make table plans. 0:12:30.800 --> 0:12:36.080 Exactly, it's impossible because twenty five people is more combinations 0:12:36.320 --> 0:12:39.200 than the biggest computer in the world could compute in 0:12:39.240 --> 0:12:41.760 the history of the universe. If we'd started it in 0:12:41.760 --> 0:12:43.320 the beginning, it still wouldn't have come up with. 0:12:43.360 --> 0:12:45.439 It gives you a new respect for wedding. 0:12:45.200 --> 0:12:50.240 Planners, doesn't it exactly? Exactly? That's optimization. The second thing 0:12:50.320 --> 0:12:55.080 it's good for is simulation. And simulation is where we 0:12:55.320 --> 0:13:01.600 have the ability to simulate natural mods. So, for instance, 0:13:01.679 --> 0:13:05.480 the way that a chemical reaction happens, the way that 0:13:05.720 --> 0:13:10.280 a material actually behaves at the moment. With our data, 0:13:10.280 --> 0:13:13.480 we can approximate it because it's ones and zero's a 0:13:13.520 --> 0:13:16.439 bit like a CD and a vinyl A cd is 0:13:16.520 --> 0:13:19.480 ones and zeros a. Vinyl is the true wave. Quantum 0:13:19.559 --> 0:13:22.240 is operating at the vinyl level, and it is able 0:13:22.320 --> 0:13:25.439 to model the way that a new drug will interact 0:13:25.480 --> 0:13:27.840 with your body. I would use the word perfectly, but 0:13:27.920 --> 0:13:31.560 we don't know because we haven't We haven't understood the 0:13:31.600 --> 0:13:34.520 next levels down and the next ovesalguma and a much 0:13:34.760 --> 0:13:39.800 better resolution than you could with a binary compute. So 0:13:39.840 --> 0:13:46.040 for things like material discovery, drug discovery, how chemical reactions happen, 0:13:46.240 --> 0:13:50.200 like the nitrogen bond to make ammonia, we use in 0:13:50.520 --> 0:13:55.080 seven percent of the world's energy creating ammonia for fertilizer. 0:13:55.360 --> 0:13:58.720 If we could understand actually how that bond was structured 0:13:58.760 --> 0:14:02.520 and how you could break it, potentially you could find 0:14:02.559 --> 0:14:06.200 out how it works and use a fraction of that energy. 0:14:06.640 --> 0:14:09.480 And the reason it's so important is optimization and simulation 0:14:09.600 --> 0:14:13.280 are two of the areas that we are very bad 0:14:13.720 --> 0:14:16.520 dealing with as humans because we don't have the tools. 0:14:16.679 --> 0:14:19.840 And we're about to create systems that can do that 0:14:19.880 --> 0:14:22.320 for the first time. So if you ask the quantum companies, 0:14:22.360 --> 0:14:24.560 they say they actually have machines that work. Now. 0:14:25.080 --> 0:14:26.760 Okay, So when I was about to ask you what 0:14:26.800 --> 0:14:30.120 you meant by about but you say that, the companies 0:14:30.120 --> 0:14:34.440 would say, this is here already usable stable, it's. 0:14:34.320 --> 0:14:37.160 A great it's not a black and white. So there 0:14:37.200 --> 0:14:39.520 is something called a cubit, and a cubit is the 0:14:39.520 --> 0:14:43.080 equivalent of a bit in computing. To run a quantum computer, 0:14:43.200 --> 0:14:46.760 you need cubits and you need to connect cubits together. 0:14:47.240 --> 0:14:50.880 And what confuses the picture more and I think people 0:14:50.960 --> 0:14:54.480 have to understand this. When people use cubit, you have 0:14:54.560 --> 0:14:57.880 to ask is it a physical cubit or a logical cubit, 0:14:57.920 --> 0:15:01.360 and they don't say which is which. The difference is enormous. 0:15:01.760 --> 0:15:06.040 So a physical cubit is a working cubit. A logical 0:15:06.080 --> 0:15:10.520 cubit is one that you can control and use effectively. 0:15:11.000 --> 0:15:13.840 And there is enormous difference, maybe one thousand to one. 0:15:13.920 --> 0:15:16.480 So certain companies will say they've got a one hundred 0:15:16.560 --> 0:15:20.360 or one thousand cubit machine, but that's not a logical 0:15:20.440 --> 0:15:26.000 cubit machine, and therefore it's to a degree not that powerful. 0:15:26.120 --> 0:15:29.240 And other companies like Quantinium will claim they've got a 0:15:29.360 --> 0:15:33.600 fifty two cubit machine, but it's a logical cubit machine, 0:15:34.000 --> 0:15:36.360 which means it's very very powerful. 0:15:36.640 --> 0:15:39.360 I don't understand the difference between these two cubits. Do 0:15:39.440 --> 0:15:42.040 I need to understand the difference between these two cubits. 0:15:42.520 --> 0:15:45.040 You only need to understand it if you're investing in 0:15:45.080 --> 0:15:47.840 the space, because if you're matching two things that are 0:15:47.840 --> 0:15:50.800 not equal, then you need to know that there is 0:15:50.880 --> 0:15:54.200 a difference between them. So if I open the paper 0:15:54.240 --> 0:15:57.440 and it says company A has a thousand cubits and 0:15:57.520 --> 0:16:00.200 Company B has ten cubits, which is what they do 0:16:00.240 --> 0:16:04.560 in the press, those ten cubits may be logical cubits, 0:16:05.040 --> 0:16:09.280 and the thousand cubit machine is physical cubits, and they 0:16:09.360 --> 0:16:12.440 have zero logical cubits, And that is the only reason 0:16:12.480 --> 0:16:14.840 to understand it. You don't really need to understand what 0:16:14.880 --> 0:16:15.480 that means. 0:16:15.800 --> 0:16:19.000 Okay, So the logical cubits are the much more powerful ones. 0:16:19.280 --> 0:16:23.880 They're the ones that are error corrected and work reliably. Okay, 0:16:24.080 --> 0:16:27.280 that's what you need to know. The point is, up 0:16:27.400 --> 0:16:32.720 until twenty twenty four, no company had announced a logical cubit. 0:16:33.280 --> 0:16:36.760 In twenty twenty four, someone announced one, and then in 0:16:36.800 --> 0:16:40.480 early twenty twenty five they announced ten. Continuum, who are 0:16:40.480 --> 0:16:45.320 the half British half American company announced fifty at the 0:16:45.480 --> 0:16:49.479 end of last year, fifty is starting to become meaningful. 0:16:49.880 --> 0:16:54.240 The one hundred started to become useful. And so if 0:16:54.320 --> 0:16:56.960 you look at that progression of one to ten to 0:16:57.440 --> 0:17:01.520 fifty over an eighteen month span, you're looking at one 0:17:01.680 --> 0:17:05.119 hundred within the next year or so. And then you 0:17:05.200 --> 0:17:07.760 have a machine that is useful. 0:17:07.920 --> 0:17:11.720 Useful for what initially? So we've got a hundred logical cubits. 0:17:11.920 --> 0:17:13.560 See I know or I know what these things are. 0:17:13.640 --> 0:17:15.879 Now we've got one hundred logical cubits. We've got a 0:17:16.000 --> 0:17:20.159 machine and it's useful. But what will it do for 0:17:20.240 --> 0:17:23.360 us initially? What will it change? What will we understand 0:17:23.440 --> 0:17:25.080 that we didn't understand before. 0:17:25.880 --> 0:17:28.760 I don't think we will particularly be able to crack 0:17:28.840 --> 0:17:34.600 anything new because traditional compute, especially traditional compute with AI, 0:17:35.040 --> 0:17:39.959 is phenomenally powerful. And therefore, even though these things are meaningful, 0:17:40.320 --> 0:17:45.679 can they do anything more significant in a wide range 0:17:45.840 --> 0:17:49.200 than where we are with AI today? Probably not. As 0:17:49.240 --> 0:17:52.359 we start moving up to the hundreds, you know, the 0:17:52.560 --> 0:17:56.720 three hundred, four hundred, that is the point where we 0:17:56.840 --> 0:18:01.480 will be able to potentially discover new drugs or discover 0:18:01.600 --> 0:18:05.200 new materials. Those are the sort of numbers at which 0:18:05.240 --> 0:18:06.719 it becomes very interesting. 0:18:07.240 --> 0:18:09.720 Okay, and this is partly down to sheer speed. 0:18:09.920 --> 0:18:12.879 It's not the speed, it's the number of variables that 0:18:12.920 --> 0:18:15.680 you can hold. So each cubit can hold a variable. 0:18:16.080 --> 0:18:19.240 If you can only hold fifty variables, it's limited what 0:18:19.440 --> 0:18:23.360 the model can do. If you can hold one thousand variables, 0:18:23.520 --> 0:18:28.119 it's quite powerful because it's factoral, So every cubit doubles 0:18:28.400 --> 0:18:33.600 the model in effect, like that five, ten, twenty five 0:18:33.640 --> 0:18:36.439 that I was talking about. It's a factoral number. So 0:18:36.560 --> 0:18:41.080 adding cubit it sounds like it's linear. It's not. It's nonlinear. 0:18:41.480 --> 0:18:44.439 The numbers that people bandy about, and let's put it 0:18:44.400 --> 0:18:48.760 in a bandieback, somewhere in the early thousands you will 0:18:48.760 --> 0:18:53.280 be able to reach q day, which is breaking encryption. 0:18:54.040 --> 0:18:59.480 So those numbers are maybe six seven years away at current. 0:19:00.240 --> 0:19:04.240 The day when everyone's bitcoin becomes useless, right, worthless, because 0:19:04.320 --> 0:19:06.639 while the encryption is broken. 0:19:06.840 --> 0:19:10.040 It's very complicated with bitcoin. Actually it's not such an 0:19:10.280 --> 0:19:16.000 obvious thing. The underlying technology is not necessarily impacted. What 0:19:16.240 --> 0:19:20.359 is impacted is the wallets. You could potentially break the 0:19:20.359 --> 0:19:24.040 wallets if you hadn't upgraded them, So it's complex. So 0:19:24.280 --> 0:19:26.680 the people that have access to their wallets, could upgrade them. 0:19:26.680 --> 0:19:29.200 The problem with bitcoin is you've got millions and millions 0:19:29.240 --> 0:19:31.840 of bitcoin out there. For instance, it's a Toshi wallet 0:19:31.840 --> 0:19:34.919 of a million that cannot, as far as we know, 0:19:35.040 --> 0:19:37.879 cannot be accessed, so those are vulnerable. The thing I 0:19:37.960 --> 0:19:40.280 always say to the bitcoin community. What I say to 0:19:40.320 --> 0:19:44.200 them is if I have a quantum computer that can 0:19:44.480 --> 0:19:48.760 break bitcoin, Firstly, who is going to sell it to me? 0:19:49.600 --> 0:19:54.320 Because whoever develops it will be under enormous regulations won't 0:19:54.320 --> 0:19:56.600 be able to just buy one. And secondly, if I 0:19:56.640 --> 0:19:59.600 wanted to cause serious damage, I would not be going 0:19:59.600 --> 0:20:02.640 after bitcoin because it's too small an industry. I would 0:20:02.640 --> 0:20:06.399 be going after taking down some regional banks and bring it. 0:20:06.440 --> 0:20:08.480 If you want it to shorter market, which is the 0:20:08.480 --> 0:20:11.640 only way you make money on it, it's shorting. So 0:20:11.640 --> 0:20:16.640 so bitcoin is probably in the top thousands of things 0:20:16.680 --> 0:20:19.600 you would go for, but you will be small to care, 0:20:19.720 --> 0:20:23.600 too small to care. The bitcoin community cares because all 0:20:23.680 --> 0:20:25.800 their wealth is tied up in a single asset. They 0:20:25.800 --> 0:20:28.840 tend to be very hyper focused on that asset, and 0:20:28.880 --> 0:20:32.359 they're aware of the danger. The problem with most people 0:20:32.520 --> 0:20:34.880 is they are not aware of the danger because they're 0:20:34.880 --> 0:20:38.639 not technical, and so most businesses are not considering the 0:20:38.720 --> 0:20:42.119 quantum threat. Now, my argument would be they probably don't 0:20:42.200 --> 0:20:45.200 need to because most of them operate on Microsoft or 0:20:45.240 --> 0:20:48.480 Google or whatever, and they're very aware of the quantum threat, 0:20:48.520 --> 0:20:52.320 and so most of their technology will be upgraded automatically. 0:20:52.920 --> 0:20:55.840 Okay, So we get to the point out maybe ten 0:20:55.920 --> 0:20:58.159 years whenever it is when we get to this QDA, 0:20:58.400 --> 0:21:00.760 and the first thing that it can have then is 0:21:00.840 --> 0:21:03.040 we get to the point where that level of encryption 0:21:03.119 --> 0:21:06.720 becomes pointless. But companies such as Google et cetera already 0:21:06.720 --> 0:21:10.840 protected against that. So what next? What's the next thing 0:21:10.880 --> 0:21:14.879 that we see happening that quantum computing will change around us? 0:21:15.640 --> 0:21:18.240 I think QDA will be scary. I think what next 0:21:18.359 --> 0:21:22.800 with quantum is it's a very different animal to AI. 0:21:23.280 --> 0:21:25.560 We have to look at what it does. It's not 0:21:25.800 --> 0:21:29.880 likely to be a consumer application for the foreseeable future 0:21:29.960 --> 0:21:33.640 for two reasons. First, it's very complicated to use, and secondly, 0:21:33.680 --> 0:21:38.159 it's going to be supply restricted for two reasons. One, 0:21:38.320 --> 0:21:41.800 they're very difficult to build, and secondly, the regulators are 0:21:41.840 --> 0:21:44.600 looking at quantum already and they're not going to allow 0:21:44.640 --> 0:21:47.640 you to buy or access a quantum computer. I think 0:21:47.680 --> 0:21:51.840 what changes is different. What changes is assuming they are 0:21:51.960 --> 0:21:58.000 supply restricted, and assuming they do what we've discussed, which 0:21:58.040 --> 0:22:05.480 is optimization and discovery. Both of those are fundamental to 0:22:05.560 --> 0:22:10.520 business optimization in terms of one percent saving in logistics 0:22:10.640 --> 0:22:15.159 or in farming, or in weather prediction or any of 0:22:15.200 --> 0:22:20.560 these things is a huge competitive advantage. The second thing is, 0:22:20.920 --> 0:22:24.280 in the world that we live where intellectual property is protectable, 0:22:24.920 --> 0:22:28.960 the second person that discovers something has no value. It's 0:22:28.960 --> 0:22:31.399 the first person that discovers it. So if I'm a 0:22:31.400 --> 0:22:35.280 pharmaceutical company and how I have access to the biggest 0:22:35.280 --> 0:22:39.280 and best quantum computer and no other pharmaceutical company does, 0:22:39.880 --> 0:22:45.959 my competitive advantage is enormous because I'm out there discovering 0:22:46.040 --> 0:22:49.160 drugs on a daily basis or an hour however long 0:22:49.200 --> 0:22:52.400 it takes to run the thing, or an hourly basis, 0:22:52.520 --> 0:22:54.760 and none of my competitors do. And if they're two 0:22:54.880 --> 0:22:57.560 years behind, then I've got a two year free run. 0:22:58.000 --> 0:23:02.080 And because these systems are so difficult to build, you 0:23:02.119 --> 0:23:05.200 can't just go and find another competitor and say, okay, well, 0:23:05.680 --> 0:23:08.359 this drugged company has a four hundred or one thousand 0:23:08.440 --> 0:23:11.160 logical cubit machine. Will you build me one? They won't 0:23:11.200 --> 0:23:11.719 be able to. 0:23:11.960 --> 0:23:14.720 So does that lead us into a new age of monopolies? 0:23:15.119 --> 0:23:18.040 I suspect again the regulators will be on it. But 0:23:18.520 --> 0:23:21.800 it leads you into a winner takes all scenario. It 0:23:21.840 --> 0:23:24.840 turns a lot of companies into, in effect, the tech companies. 0:23:24.920 --> 0:23:27.160