A (2:49)

A tad before my time, but I could fake it on Qbert. I don't know if we're talking quantum mechanics. I didn't have time to audit quantum mechanics over the weekend. But we'll give it our best shot here.

B (3:00)

Yeah, one of my favorites on the Atari back in the day. Yet we are not going to fashion ourselves as sort of quantum computing experts, to say the least. My primary memory of quantum computing was I was at that's some sort of Microsoft press day sort of thing. And at the end we went to their sort of quantum computing lab and I had just arrived from Taiwan and was pretty jet lagged and I think I fell asleep in the lab, which was unfortunate. But I mean the way I think about quantum computing is. I mean you have this crazy state where in classical computing you're dealing with bits. And the whole thing with transistors is they're either on or off. And the fact that it can be on or off is how the whole thing is built. It's built on this entire. All of logic can be devolved down into ones and zeros. And the whole chain of how that happens is really interesting. But the long and short of it is that the way we do computing is that computers break everything down basically into ones and zeros and run these very simple calculations that are. There's just stacks and stacks of logic on top of this to make this possible and we can sort of solve stuff. And one of the big shifts over that we've talked about in the context of AI is the shift from general purpose computing to this GPU accelerated computing. And what that does is instead of having these relatively deep pipeline general purpose chips, we have simpler cores or so an Nvidia card has like 4,000 cores on it or something. Or the actual ones that use for AI have way more than that. And it does these simpler calculations and it does them in parallel. So it's doing a bunch of similar calculations all at the same time. And this is parallelization. Quantum computing is like parallelization, but it's like just in a completely. It's like completely orthogonal. It does all the calculations at once at the same time. So it's kind of insane because a qubit unlike a regular bit, it's not one or zero. It can sort of be all of the above at the same time.

A (5:11)

Right. That's what I couldn't really understand is like how you could possibly build systems that are anywhere approaching reliability.

B (5:20)

No, you can't qubits. That's part of the breakthrough here. So you have these qubits that can be sort of at all states between 0 and 1. The problem is like, you go to measure it and then it flattens down into either a 0 or 1. And so.

A (5:37)

Right.

B (5:37)

It changes. Right. And, and there's all these crazy things like different qubits can be entangled. So when one moves, the other one moves sort of X, Y, z. I think the big, the, the more, the bigger breakthrough here is you have like a physical qubit and then you have what's called a logical qubit, which is the qubit you actually do computation with. And one of the challenges, you have physical qubits that are very hard to stay in state. And what they've done is basically has shown that you can scale up the number of physical qubits such that they correct each other and keep each other in place. So that the logical qubit which sits on top of the physical qubits is sort of stays in state long enough for you to do calculations on it. And they basically shown that we can add more physical qubits to make it more stable, which is a big sort of breakthrough because usually more physical qubits, they get less stable, but it's basically they've figured out this error correction mechanic where ones fall out of state, the other ones that are part of the logical qubit correct the ones that fall out of state, and they're all correcting each other such that the logical qubit stays in its state long enough that you can use it for sort of a calculation. That right there, I think, is the main, is sort of the main breakthrough here, because that's sort of a prerequisite to doing anything, is that you have something that you can sort of operate on consistently and you have a path towards it being sort of more useful, more useful over time. So it's a genuine breakthrough. It's also one that's like five years on from the last sort of breakthrough or whatever it is. This is a field that, it has applications. I mean, the, the, the classic one everyone talks about is breaking certain kinds of encryption. One thing that, like a hard problem in what you do with a lot of encryption is you multiply two numbers together and then you get an N number. But, but to devolve that back to the numbers that created it, you have to factor it, right? I don't remember factoring from like middle school math or whatever, where you're sort of breaking apart a number into its different component Pieces so you can figure out what the variables are. I hated factoring back in the day. But instead of calculating every single possibility, which you would do with classical computing, this idea that you could sort of do all the possibilities sort of simultaneously and figure out what it is and in theory means you could get to a state where certain sorts of encryption algorithms, you could crack them very quickly and pretty quickly. Right. But this doesn't apply to all encryption algorithms, just some of them. Some of them are more easily broken. There are workarounds sort of against this. It's not like encryption is going to be dead. And also we're years and years away from this being possible because we're talking like super small. So that's one. There's like certain search applications, particularly on unstructured data, where again, you're just like, right now there's lots of applications where we just sort of do it serially. We try every possible thing and then we eventually find the right one. And that's where they come out with these predictions. Oh yeah. To do this with classical computing would take seven septillion years or whatever it is, because that's assuming you're like going through and trying every single one. And what if you could try a bunch of ones sort of simultaneously, all at the same time in the different metaverses or whatever it might be. So there's real applications. This is real progress. It has basically no implication on the world today as it is, or in sort of the, you know, the media to medium future. I mean, I guess the long term goes to infinity, so it does its applications there.

A (9:11)

But long term goes to septillion, Ben?

B (9:14)

Absolutely. Yeah. So, yeah, no, it's a breakthrough. It's cool. It's not really pertinent, I think, to anyone and not really pertinent to the sort of stuff we talk about on Sharp Tech or on Stratachery. So and I probably even in this very high level fuzzy overview got details wrong. So don't kill me for it.

A (9:34)

But I appreciate the effort and I appreciate the question from Chris because I've seen sporadic mentions of quantum computing over the past few years. My wife is actually working on a project related to post quantum cryptography for her consulting work. And on Sharp China, there's always stories about China pursuing quantum computing and the US trying to do quantum computing. And so I wasn't necessarily sure. I mean, I was coming in pretty green here. But it's good to know at a high level that a lot of it's related to cryptography. And in the near term, a lot of it is sort of irrelevant to the tech story.

B (10:13)

Right. It's not all cryptography. It's like certain ones, like there's, there's a very famous algorithm like, called RSA that's used like for web browsing and stuff like that that's like trivially broken. And so that would have to be like solutions done to that. There's other ones, like hashing algorithms that aren't really impacted at all. It's not an insurmountable problem. Like if someone had quantum computing or there was visibility to it being a thing in the next decade, it would be a big lift to update everything to make it resistant. But it's not, it's sometimes framed as all encryption is dead and that's not the case. And so like that's what my wife

A (10:50)

is working on, updating systems to be quantum resistant.

B (10:53)

Yeah. So like Bitcoin, like the mining algorithm could be solved faster, but it's like a speed up. It's not breaking the actual like keys where you hold your wallet. That could be broken tomorrow. So like, but, so that would have to be changed like, so, but so there's, there's, there's different out, there's different encryption methods that are impacted in sort of different ways. But again, I don't think there's any sense or worry that, that China is going to figure out tomorrow. And we thought it was like decades away. I think everyone's pretty clear. It's a long ways away. There's something here. It's certainly very compelling. The entanglement aspect of qubits being connected across even though it would be close to each other is insane. Quantum mechanics, it breaks your brain to a certain extent. That's why you throw in this multiverse sort of aspect. I mean, sure, um, you know, we're, we're recording this. It is 9:47am on a Monday morning. I'm probably not in the right mental state of mind to explore the, or explore those possibilities.

A (11:56)

Well, I appreciated Google for throwing that in there because it made the story buzzier and more viral and it's why we got a number of questions about Google's quantum computing efforts.

B (12:07)

Yeah, this will test is silly because it's like, it's like simulating quantum mechanics. It's like using quantum mechanics to simulate quantum mechanics. And like the comparison to like how classical computing, it's not actually verifiable because number one, no one has tried to do it for obvious reasons. It would probably take forever. So they're comparing it to like A theoretical output. And there's actually been in the past, I don't have it offhand, but I know there's been tests they've done and then someone's like oh actually we could do it in classical computing. They figure out some new algorithm that does it just as fast. So I don't think that's probably not the case here, but the actual test has no bearing on like anything in real life or whatever.

A (12:42)

Again, solving real challenges that we face today. Okay.

B (12:45)

I think there is a real breakthrough here and it's really cool. I also, my at least view right now is this has no real impact on, on sort of day to day life. So it's, it's a fun story but nothing to get too worried about.

A (12:57)

Okay. There was one other tweet that I wanted to read here before we move on. A listener named Sam sent us this note from Trung fan who tweeted this in response to the Willow freakout that everyone was having. He wrote, Google is really the new Bell Labs just using that monopoly search money printer to throw off new inventions. Waymo AlphaFold, the Transformer Gen AI paper, Willow Quantum computing chips and triple unskippable back to back to back pre roll YouTube ads. Ben, do you agree that if any company has a claim to the Bell Labs legacy, it's Google?

B (13:36)

Absolutely. I mean I think we've talked about that on this podcast before and this, I mean, I guess it says something about me and I'm the wrong person to ask about quantum computing where this question gets me much more excited and I'm like, oh yeah, my sort of brain gears turning, let's roll up our

A (13:53)

sleeves and talk through it?

B (13:54)

No, because it's an interesting sort of governance and philosophical question. Right. Should we worry like Google one search, congratulations. Is it actually a better net positive that instead of going after them and trying to push against a string as I frequently talk about and undo this search monopoly, we should say hey yeah, just keep shoveling your money sort of into these science projects. Should there be some sort of similar to the Bell Labs? Like part of, part of the settlements the government had with ATT over the years is they couldn't sort of go into those industry. They could only use those inventions for their own. And it worked out because things like transistors and information, you know, what's the word I'm looking for? All the stuff they were working on was relevant to phone service and so they got a benefit but obviously basically birthed modern computing. And it's hard to argue that that Wasn't a big sort of net win like Bell Labs was basically the R&D department of the US. This is a point that Bernard Hobart made in his book in our interview a couple weeks ago. It was basically the R and D lab for the US Government. The US government could basically not have to spend all the money to do like public R and D. They could just tolerate the ATT monopoly and say, you do all the money for R and D. And again, US sort of technological power is predicated on a lot of inventions that came out of Bell Labs. Now, is there some sort of disconnect here where what Google does ought to be shared more broadly? Transformers is a good example of what we want. It was a public paper. Anyone could pick up on it and do it, and they did. And the fact that it's not sort of patented is a real victory. Should there be something similar with, you know, with a Waymo or whatever it might be? Maybe. I mean, and is it really a criticism that, oh, they're inventing stuff that's not being sort of monetized or, you know, that, you know, I think that there's this. I'm not completely sure how to structure this. Like, do you just sort of back into a Bell Lab situation? Can you actually do it with purpose? I certainly tend to agree that a lot of the modern antitrust crusaders spend a lot of time on challenges that they're not going to solve and that unfortunately have the outcome of not really helping anyone, including potentially killing stuff like this. So I think that that's a fair critique. But hey, it is an interesting philosophical discussion. What do you think?

A (16:25)

First of all? Yes, I think that Google is the closest we have to a modern day Bell Labs. I mean, if we could get a look at the internal projects from every company in big Tech, I think Apple would be the most entertaining to see. You know, the Apple car and various failed ideas over the last 15 or 20 years. But Google, if everything they were working on was just released to the public domain tomorrow, I would guess based on what I've read over the last couple years, that their tech would have the most benefits to society at large and technological progress. And it's almost more interesting to have a company like that existing today than it was during the Bell Labs era, because now it feels like the government, I don't know that the government is capable of pushing forward on the technological front, whereas 50, 75 years ago, I would trust the government more in that area. And so it seems like we've sort of outsourced A lot of technological innovation and research and R and D just generally to the private sector. And there are real benefits to that. And I don't know if that went away. I'm not sure who would be pushing forward if it's not a company like Google that's sitting on just a mountain of cash and can afford to take moonshots in a variety of areas. Like I do think that's a real benefit to consider in this conversation.

B (17:51)

Yeah, well, I mean governments can still kill stuff though. And I mean you folks seem to be going after Google. Just leave, leave for Google alone.

A (17:58)

Well that look, the theme of the podcast, there are trade offs in every area. There's a real cost to competition and innovation with Google's success. But they also are funding crazy moonshots and empowering Hartmut Nevin to do his worst out there and push forward on quantum computing.

B (18:18)

So I mean the quantum computing.

A (18:20)

See where it all that's there are

B (18:21)

other like folks working on like a. Microsoft's been working on it for a long time. The Apple mention is interesting because Apple generally does not do this kind of things. And this was actually a comparison that I thought was interesting between Nvidia and Apple that stood out to me when I was reading Tae Kim's book. Is Apple's pretty relentlessly focused on inventing four products if it's not obviously commercializable again, an Apple car. It wasn't theoretical is could this be a product? And once it was determined it's not going to be a product then we're not going to just keep doing sort of research.

A (18:59)

And that's for the record is why Apple would be so entertaining because it's like these are things that we might have used and that might have gone to market whereas Google, I feel like it's a little bit more abstract and you'd have to be like a real developer or engineer to appreciate the brilliance that exists there and could benefit the world.

B (19:19)

Right. And from a shareholder and sort of analyst perspective it's like what are you wasting money on? What's the point here?

A (19:26)

Totally. Yeah. Well, we'll see. In any event, to keep it moving, Stu says it seems like you're purposely avoiding talking about these drone sightings. Probably because anything you say would be pure speculation. Stu. That is correct.

B (19:43)

But also, also no, no, I know I'm going to push back on Stu. I am not a fan of assertions of sins of omission. Like I have not considered talking about drones on Sharp Tech. I'm not purposely avoiding anything. It doesn't seem really relevant to my topics or what I do. So, no, no. Asserting motivations to something that was not done. That's, that's, that's my feedback to Stew.

A (20:09)

Well, we recorded last Wednesday night in the United States, Thursday morning your time, and this was a story that was still very much developing. So anything we said back then would have been a complete speculation. But in any event, the good news

B (20:27)

is today it's still complete speculation.

A (20:29)

Still complete speculation. That's why I didn't foreground it in any of our Anderil discussion. But he says it got me thinking about the technology of drones, which is apparently pretty much commoditized parentheses to the point that you can't really tell from photos who made a drone. And then he says, it also occurred to me that the drone is kind of like the iPad. It seems like it should be a useful platform, but outside of military applications or real estate photography, it doesn't really have much practical use. What are your thoughts on drones more broadly? Also, feel free to speculate about these drone sightings because they are kind of weird. Ben, do you have any thoughts on that? Take the main reason I included this is because I loved calling drones the iPads of the sky, at least as far as consumer tech is concerned. But do you have thoughts?

B (21:23)

Yeah, I disagree. I think drones are very compelling. I mean, the military one I think can't be understated. Like, drones are pretty clearly the future of military. They're not the future, they're the present.

A (21:36)

All right, and that is the end of the free preview. If you'd like to hear more from Ben and I, there are links to subscribe in the show Notes, or you can also go to SharpTech FM. Either option will get you access to a personalized feed that has all the shows we do every week, plus lots more great content from Stritechri and the Stratecheri Bundle. Check it out, and if you've got feedback, please email us at emailarptech fm.