Amazon Healthcare, Oculus & VR/AR, Google Quantum Supremacy? (#10)

Hi everyone. Welcome to the A6&Z Podcast. I'm Sonal, and I'm here today with the 10th episode of our short-form news show, 16 Minutes, where we cover recent headlines the A6&Z way. This week, we cover multiple topics— Amazon's virtual care, what it means for incumbents and startups, and the future of healthcare. Then we do a Quick Take on Oculus Connect and where are we with VR and AR. And finally, we discuss the leaked paper/news that Google may have achieved quantum supremacy. What does it mean? Where are we with quantum computing? But first, please be sure to subscribe to 16 16 Minutes separately in your podcast app. Just search for it there because we will no longer be dropping this show here, at least regularly, that is, in the main A6&C podcast feed. So go subscribe now. I always include links and relevant references or background that's mentioned in the show notes, and you can find those at a6andc.com/16minutes. And finally, please note that none of the following is investment advice or intended for investors. Please be sure to see a16z.com/disclosures for important information. So the first segment of this week's episode is about the news that Amazon just launched something called Amazon Care, which was described as a virtual medical clinic for employees. And what that really is is a combination of telemedicine and targeted in-home follow-ups. And I think it's really significant that it's for employees because previously Amazon joined up with JPMorgan and Berkshire Hathaway. And the reason they did this is because they're like, we need to move the needle on how expensive healthcare is as employers covering this. Like they have 1.2 million employees combined. So they basically banded forces to like figure out this problem. And Amazon had also acquired PillPack, which is an online pharmacy. And so all these pieces are kind of coming together. And so let me now welcome our A6NZ experts, A6NZ biopartners, Jorge Conde and Julie Yu.

It's great to be here, Sonal.

Thanks for having us.

So let's talk about this news. Why do we care?

Well, I think the reason to care about this is really, I'd say, as you pointed out, the cost of healthcare continues to go up. They just published a report that the cost of employer-based healthcare just topped $20,000 for a family plan annually, and the employers bear about 70% of that cost. So that's one important sort of trend that is driving a lot of this. The other one, which is to some extent related, is consumers are being asked to bear more and more of the cost of their care out of pocket as well. So the average consumer is investing something north of $6,000 a year to their and their family's healthcare. And so I think what's particularly interesting about this is you have a large employer that also has the ability to deploy technology coming up with a solution for their employees initially. But if history is any guide here, this is also a testing ground.

I think another interesting thing of significance from the incumbent lens is we often hear from whether it be payers or providers or even self-insured employers as they're having strategic dialogue about their healthcare operations, oftentimes the question comes up, what does Amazon's presence in the healthcare market mean for me? And oftentimes those conversations end by saying, well, at least Amazon's not delivering healthcare.

Oh no.

And so that excuse has now been officially taken off the table. And by the way, it's, it's worth noting that Amazon is not the first employer to do this. I think it's more, you know, Amazon's position in the market as a key consumer-facing brand. That's really what has been the source of fear for many of these organizations.

Well, even something like AWS, Amazon Web Services, was something they built internally for their own use that then became something they opened up to a ton of startups. And so what it ended up enabling was this whole new wave of companies being built. So if they do open up these services in the future, who knows what's possible? Exactly.

There are at least two vectors that you could see this taking. One is, you know, Amazon actually turning this into a direct-to-consumer service or taking more of that AWS approach and going B2B and saying, you know, we're now going to be selling a benefit.

One of the things that's really interesting here is that they're providing care, but they're doing that through a contract so that they haven't hired, you know, nurses and physicians.

Right. It's a part— it's a company called Oasis or something.

A company called Oasis.

Oasis Medical Group.

So what you're also pointing out then with the partnering side is that they essentially right now are buying versus building, or at least partnering versus building. And that's really interesting to me because Amazon at core is a tech company. So why wouldn't they be building this tech themselves? Do you have any thoughts on this whole build versus buy aspect of their moves?

There's a virtual care component, which certainly is technology enabled. There is a home visit component, which is human clinical services. And then there's the pharmacy component, which is really what people believe to be pill pack sort of repackaged. And I would say the first and the third of that certainly are tech. And Amazon has built the first and then acquired essentially the third. I think it's really that middle piece why people were skeptical that Amazon was going to get into healthcare is that the clinical service delivery operations is really the most complex part of that. You know, having on-the-ground services to enable nurses to walk into people's homes and have the right equipment. So that piece intuitively makes sense for this organization to actually sort of bring through a partnership. But that is also, you know, something that people will be watching very closely is will Amazon actually backwards integrate into those clinical services ultimately themselves?

I think there's also an assumption one can make here in that second bucket that you lay out, Julie, that there is an incredible lift that one could get from technology in terms of helping with the coordination of care, scheduling, scaling it, and scaling it. Yeah, because I mean, I think one of the things that is incredibly true about healthcare is we know it's very analog still. And so the things that get done on computers—

fax machine, medicine people call it.

Yeah, fax. Exactly. So I think, I mean, it'd be interesting to see how and if Amazon is helping to essentially supercharge or power Oasis in this whole delivery.

I have a question, though, just a pushback, which is this is all very tech-centric. What would the counterpart be to sort of saying the realities of how healthcare really works?

Yeah, it's a great point, Sonal, and I think some of the challenges that one would kind of see around a business like this are, you know, the very reasons why sort of tech-only companies haven't been able to scale in this market historically. So certainly one is how do you sort of integrate yourself into the supply chain of the broader healthcare landscape? Because at the end of the day, the services that Amazon is providing now are primary care. They're not doing anything related to specialty care, certainly anything that's related to sort of procedural, you know, surgical services that would require you to go into a hospital, etc. And when you look at actually the distribution of spend, primary care is really a small component of that. And so in order to truly get at the even the cost equation, let alone the overall how do you manage a patient's entire healthcare experience, this is really only scratching the surface.

So then honestly, is there room when you have huge players like this, like for startups?

I mean, at the end of the day, we always say that every small niche in healthcare is at least $1 billion wide. It's such a massive market. Primary care alone is a $220 billion market opportunity. And so I believe that it will not be a winner-take-all market, especially given some of the characteristics of how you need to scale a business like this. And when we look at these sort of novel primary care models, there's a few questions that we will ask. One is, how does this service get distribution? That is oftentimes the hardest part of building businesses in healthcare is you're really fighting against very, very heavily ensconced incumbents. And to get around those channels or build new channels is very, very difficult. I love, by the way, Alex Rampel's line.

Line, which is that the game for startups is to get distribution before the incumbent gets innovation. That nails that.

Exactly. And so that's sort of one thing to think about. The second thing is, how does whatever upstart company sort of integrate itself into the broader supply chain of health care? And that could be through referral relationships. How do you build a referral network and sort of overcome some of the barriers to becoming essentially a new market entrant? And then the third piece is scale on sort of a geographic basis. And health care is one of those sort of businesses where just because you have a contract with a payer in a given sort market or state, that means nothing for reducing friction as you're trying to enter into other geographies. And so I think that step-by-step, that brute-force model of how you sort of build reach is challenging. Amazon, if they were to bundle this as an offering with something like an Amazon Prime, you know, distribution becomes a no-brainer. I think the downside for them will arise from, you know, what about all of the data and how will my privacy be affected?

What is the future that it takes us to? We've been talking about the news so far. Like, what's the big picture of what happens next in healthcare when there's things like this new virtual care?

I think it portends what the future of healthcare might look like. In a lot of ways, it's becoming much more of a decentralized thing. And so you can get healthcare in a retail setting, whether it's a minute clinic, you can get healthcare at home, whether it's telepresence, whether it's home visits. I think what we're going to see is, you know, in many ways healthcare, you know, is leaving the building. Healthcare is leaving the pill bottle. In many ways, healthcare is leaving the traditional idea of, you know, having a medicine back in that You know, now we have technology like iPhones that can do an increasing number of ways to detect health and disease. You know, an Apple Watch can now act as an EKG. And so, you know, there's so many things that are coming together that I think are going to hopefully have a very positive impact on our healthcare system.

Healthcare's leaving the building, the bag, and the bottle. So bottom line it for me.

Yeah, I think what Amazon is doing will be what, you know, large employers will start to do. I also think that it is a sort of a forcing function for people to recognize that this is now a thing, that this will be a category of how care is delivered, certainly at the lower acuity ends of the spectrum versus more specialty care and higher need. And startups have plenty of market opportunity left across the surface area of healthcare to innovate in similar ways from the bottoms up.

Well, thank you guys for joining this segment of 16 Minutes. Okay, so for our next segment, I asked a16z general partner Chris Dixon for his quick take on Facebook's annual developer event Oculus Connect 6, which just took place this past week, and what it means for VR, virtual reality, and AR, augmented reality. And Dixon, as you go through and just quickly summarize the news, quickly share the broad implication of each.

Yeah, so Oculus made a number of announcements. To me, the TL;DR is they continued to advance VR and AR at a very rapid rate. There's the Oculus Quest, which is their marquee device right now. It's really kind of a game changer, and I think really has the potential to go mainstream. And they actually said at the event that they are selling them as fast as they can make them. They announced support for— the Quest will now run PC VR games, which unlocks a whole bunch of of sort of new content. On the technical side, they've gotten full hand tracking working, which means in the future, won't need to have the controllers in your hands, and also will have a much more granular ability to interact with the virtual world. I think of this in the larger theme as what you're gonna see over time is you get rid of the controllers, and then you see the headset shrinking and shrinking, and eventually just simply be glasses that can do both AR and VR functionality, and this is an important step along the way. They also announced the Oculus Quest has some simple AR functionality where you can see through It's called Passthrough, and they announced the Passthrough Plus, which is kind of a more advanced version of that. I think kind of signaling that they are thinking very much about kind of AR and VR converging. Really, they're very similar technology where, you know, VR is when the real world is fully occluded and AR is when you can see partly the virtual and the real world.

Just to break Passthrough down. So when you're playing like VR game or just in a VR environment right now, to see the physical world, you actually have to take your headset off and then put it back on. And so the significance of that is that when you talk about blending VR and AR, your point that you actually can be immersed in a world and also have virtual world overlaid over the physical world.

Yeah, so one of the limits of VR now, like, I'm a fan of VR, but like after an hour, there are some things that are not optimal comfort-wise. And so as you chip away at all those things, I think it becomes a much more— first of all, a much more social. Like, that's one thing about VR. Like, right now they have this cool thing where you can screencast, and so people can see on their phone what you're seeing, and that makes it more social. But make it more inclusive, make it more open so people feel like they're still in the room but they're having the full immersion. I mean, it's— there's two sides of the coin. On the other hand, the whole benefit of VR is the deep immersion, so. Right. You want that, but you also want— I think you want a balance. So that's what they're trying to do with like the AR features and just the obvious things. They also announced some more futuristic stuff. So for example, one thing that they've been working a lot on are what are called varifocal lenses. Can you say— explain that? Yeah, just very briefly. It turns out there's a number of ways in which people see in 3D because, you know, you have two eyes and your brain puts together the difference in what you see. And VR is able to replicate that very well. If you want to try it, there's an app called BigScreen, which we're investors in, where you can watch 3D movies and you get a much more richer 3D effect than you do in the movie theater because you have true eye separation. But it turns out there's other ways in which your eye sees depth. I'm not an expert on this, but I'll try. So what your eye expects is that when an object is farther away, the light bends in a certain way. And right now in VR, it doesn't happen that way, right? Because there's a screen in front of you, it has an infinite focal distance is what the technologists would say. And so one of the things they've now gotten working in the lab is something where you actually can have many dozens of layers of focal distance in the same headset. I would expect the next device to have what's called foveated rendering, which is basically tracks your fovea, which is where your eye's looking, and render— so it turns out the way you look in real life is when you're looking at something, you see high-res in the thing directly in front and then low-res around it, and this is your way of your brain kind of saving compute cycles. And so VR will do something similar where you only render the center.

Wait, so foveated rendering, if I understand correctly, it basically trades off the broader field of view with higher resolution for where your eye is actually focusing and not for focusing on things in the periphery as a way to trade off the available processing power.

Yeah, so for example, the Oculus Quest is running on a mobile processor, unlike the Oculus Rift, which you have to physically plug into your PC and has much more GPU power. And so as a result, if you look, you can see, like, less lighting effects, just simpler textures, things like that. And that's because the GPU is running at full speed, and it's a mobile GPU. I mean, it's amazing that they could do this on a mobile GPU. It is amazing. The GPU is doing both inside-out positional tracking, which is itself a hard machine learning problem, and it's rendering a full game engine and everything else. And that's just one of many things that are being advanced. And so a big question when you have a new technology is, are you kind of following a linear or exponential path? And I think we're following an exponential path, or we will be soon once we get the developer ecosystem really going. The missing piece right now is enough users that developers can invest. Right now, the major kind of game studios and other kinds of software developers don't see it as a big enough platform. At some point we'll hit that tipping point. There's probably 10 million active users or something. Needs to be in the billions and tens of billions. But all signs point to this technology really getting rapidly better and Facebook driving the vast majority of it. There are other companies doing cool stuff, like Valve is doing some very cool stuff. Microsoft tested HoloLens and AR and things. I think they're kind of betting more on enterprise use cases in AR.

Okay, so that's a quick update at the device level and where we are, where we're not in the trend. Bottom line it for me, Dixon.

The bottom line is this is going to be a major new platform, I believe, and we're seeing evidence of it. I think we went through the trough of disillusionment in VR and we're coming out of it. And the rate of improvement is very quick. I get the feeling watching Oculus Connect the way I used to watching the Steve Jobs Apple keynotes, like today feels more incremental, frankly, on the mobile phone side. But here there's like sort of genuine big breakthroughs.

Fantastic. Well, thank you for joining this week's episode of 16 Minutes. Thank you. So in this last segment, we're going to talk about the recent news. And honestly, I wouldn't normally put this in the news because it's really journal article, but enough media headlines covered it that I want us to talk about it, which is this paper, apparently Google claims quantum supremacy. But here's the funny part. The paper was accidentally released by some of their collaborators at NASA who put it on their public website. So then it was really quickly taken down. But in that period, a ton of media headlines hit about it. And just to quickly summarize the news. So a group at Google achieved so-called quantum supremacy with a 53-qubit superconducting device. And a qubit, just for definition, is a quantum bit. And they're doing this with a quantum processor called Sycamore. And just a quick definition on what generally people people may define as quantum supremacy is just, very simply put, it's the point at which a quantum computer surpasses classical computers for some specific set of problems in a given domain. And then there's also some other recent news and movements. But what we really want to talk about today is tease apart the facts from the fiction. Scott Aaronson, the professor at UT Austin who writes a lot about quantum computing, in a blog post he wrote about the Google News, he had this line that said, "Fact-free pontificating about what it means has predictably proliferated." So we're here to unproliferate That voice you just heard is a 16Z Biogeneral partner, Vijay Ponde, who has a PhD in physics, which is very relevant here, but also early in his lab at Stanford did pioneering work on a distributed computing protein folding project called Folding@home. And it's pioneering because it really pushed the boundaries of computing and was very early to using GPUs. And by the way, just to quickly define what quantum computing is, simply put, it's a type of computing based on nature's operating system of quantum mechanics. So unlike classical computers, which are limited to human style binary code, quantum computers can do things that classical computers can't do or can't do fast enough. And this is really important, especially as Moore's Law continues to achieve seemingly impossible feats. At some point it really hits its limits, and that's where quantum computers come in. But people have been talking about them for ages, and they also always tend to go right to the saying, oh no, once this happens, all our cryptography will break too. So what's your quick take there, Vijay?

So what's interesting here is that There are some people who think that this is nothing, and that's wrong, actually. And then some people think this is the end of the world, and that's wrong too. So is quantum computing real? Is it here? Yes. Now, does this mean that cryptography is broken? Does this mean that classical computers aren't useful? That part is not true. And I think supremacy is a term that really maybe is a little bit grander than it needs to be.

Yeah, actually, can we talk about this? Because people get really caught up in it. And I think it's one of the funniest jokes I've heard about the term quantum supremacy. Someone once told me that it's a way for Google to convince management that they're doing good work. Work because it gives like the managers like, how do you even do milestones and really cutting-edge work like quantum computing? But the other thing that is, you've actually coined this concept of a quantum intercept when we wrote about our investment in Rigetti Computing. That's right. And I've also heard the term quantum inimitability.

And also like Rigetti has used the term quantum advantage. I think quantum supremacy is a big way to say that you can run a calculation on a quantum computer faster than whatever that calculation would be on a classical machine. It doesn't have to be a calculation that any of us would ever want to run or need to run.

It just can do it better, basically.

It can do it better.

And just to be clear, this can be very different for very different types of problems in different domains. That number of qubits required to do it is very different across all of them. Absolutely. You told me about the part that isn't hype. Can you tell me the part that doesn't work about this? Okay, so here's what you can't do.

It definitely isn't a general-purpose machine. You couldn't run any sort of calculation on it. But, you know, GPUs are like that too, in a sense. And then the second thing that it isn't is that people have been saying, oh, now cryptography's broken and Google, like, can do this and that. That is not the case. When people talk about a quantum computer, they mean different things, and this is part of the problem. And so when some people talk about quantum computers, they mean error-correcting or noiseless quantum computers. And what Google and IBM and Rigetti are working on is sort of near-intermediate, sort of noisy quantum computers.

Are you talking about NISQ, like noisy intermediate-scale quantum? Yes. And in layman's terms, Is it basically the idea, like how to basically get useful information out of a crazy quantum machine?

It has a little bit of noise, but there's a lot of things you can do to work around that. And so in that sense, it's maybe almost a little bit more similar to analog computing in the early days where noise doesn't have to be a problem, but you have to have clever algorithms. And there's several clever algorithms that run on these NIST-style machines, but cryptography is not one of them. And so here's why this is interesting. You know, you talked about the early Folding@home days When we first coded on GPUs before CUDA or languages for GPUs were common, we were one quarter the speed of a CPU. And a lot of people in my lab were kind of disappointed about this. I was excited because what happens is that when you have two curves and one is growing faster than the other, with the GPU curve growing faster than the CPU curve, just the fact that you could get this to work means that in a relatively short period of time, it will become dominant.

It's interesting you mentioned from CPU to GPU because people have talked about also TPUs, like tensor processor unit, but then after that, a QPU, a quantum processing unit?

Well, the QPU and a quantum computer are one and the same. So quantum computer is a very different approach to computing in that instead of doing calculations with a typical processor, you're essentially running a quantum experiment with these multiple qubits. And why would you bother?

So just concretely, what are the early applications where this might play out?

Yeah, the early applications that people are thinking about are one, in areas of actually doing quantum mechanics calculations, but also there's a whole burgeoning area of quantum machine learning. Basically, the nonlinearity of quantum mechanics allows a much simpler version of quantum machine learning to reproduce what you normally have to do in a much more complex way with classical machine learning. And of course, you know, in the world I care about in engineering biology, now we can finally engineer it in some aspects that we can't today. So quantum mechanics is the world of molecules as it is. And so being able to finally calculate that is entering into a world where we can apply engineering-like principles into areas that we just can't right now.

So then let's go back to this idea of quantum supremacy. So give me your quick take on— do you think it's just BS and some hypey thing, or is it like a real thing?

I think it's real that they've been actually able to run a bona fide algorithm on a machine of that scale. So I think that's impressive. And I think what this hopefully does is take people who are quantum skeptics and make them realize that there actually is something here. The other part that I think is really, really important to emphasize is that it's not like we're jumping from one Moore's Law curve to another that's going a little faster. If Moore's Law says the power of the machine doubles like roughly every year, but a quantum machine, its power is exponential in the number of qubits for certain types of applications. And so where GPUs, we just had a faster version of Moore's Law, from CPU to QPUs, the paradigm shift, it's a completely different game. This is something like, it's a completely different law.

Well, to break down concretely with the implication for people building software and anything that's relying on computing, it means that if that law plays out, the surprise of when people hit that moment of quantum supremacy, intercept, advantage, whatever label we put it, will happen much faster than you think. Nothing happens, and all of a sudden, oh my God, it's there.

That's exactly right. You know, people are making analogies for this to Sputnik. You know, Sputnik was just a very simple satellite, but there was a lot more to come.

Scott Aaronson, he actually likens it to the Wright Brothers, and it's the perfect example. It's exactly what you said. It's not here, but it's also not not here, because the Wright Brothers, there was all these false But then they actually did learn how to fly. You and I just talked about this a couple of weeks ago on 16 Minutes. There's always a hybrid phase when you're doing really cutting-edge stuff.

Absolutely. So we talked about these different types of machines. There's like the noiseless machine, there's the NIST machine. You could actually think about a third type of architecture where you combine some aspects of quantum and classical. And when you think about when you run on a GPU, you never run just on the GPU. It's a GPU and the CPU working together. Exactly. And some parts are very on the CPU, some parts on the GPU. CPU-QPU hybrid makes a ton of sense, especially now. Fantastic.

Thank you for joining this segment.

Great. Thank you.