Some Future Day

Microsoft and the New Tech Stack: Artificial Intelligence, Blockchain, and Gaming | with Yorke Rhodes
===
[00:00:00]
Marc Beckman: York, good morning. Thanks for joining some future day. It's great to see you today.
Yorke Rhodes: Great to see you too.
Marc Beckman: So I got to ask you for a guy that's been involved in blockchain to the extent that you have as well as for the length that you have When is blockchain going to have its chat GPT moment? When will you know, when will Blockchain become ubiquitous.
When will it reach the masses?
Yorke Rhodes: It's a great question. I'm glad you started there because that's definitely something that I've been sort of predicting for quite some time. When I got into blockchain in 2015, which is really where I went down the rabbit hole, I spent about six months really thinking about like how has technology evolved in the web stacks.
And, um, how, you know, how should we think about this technology as it relates to that? [00:01:00] Um, and if you, if you ever listen to Tim Berners Lee, who's one of the, essentially a lot of people regard him as the father of the internet tech, tech stack. Um, and I just watched an interesting, um, interview with him when he was at CERN, uh, as a, as a young researcher.
Uh, when he actually came up with this idea that you could hyperlink from, you know, one thing to another. Um. Uh, the, the comments that he makes about the internet, that he really feels like where sort of it fell down was actually an identity, uh, and, um, so that sort of pushes you towards well what, what form of identity, what does identity mean in a decentralized substrate like the internet?
Um, how do we think about those technologies in a way that, uh, allows a consumer to be self representative and in self control over, uh, Uh, their personal data, um, and actually have a system that, um, if you're going to [00:02:00] represent a human being, uh, that is self sovereign, that's actually incorruptible by regime change and things.
And so, um, from a tech stack perspective, you know, this is now been well studied, um, over, you know, since, since, uh, blockchain came onto the scene. Um, you know, blockchain has taught us that there's a way to have a technical substrate which gives you self sovereignty. Uh, and so that's pretty interesting, right?
And it actually fills in a hole that Timsburgers League recognized, um, and has recognized for quite some time. Um, so... That's a foundational technology. Um, blockchain also teaches us that, um, you can transact value on the internet, uh, in a way that you know, is very hard today. Um, where you're transferring, you know, digital assets essentially of various forms, um, that have a lot of attributes associated with that, and a lot of those attributes can be valued.
Um, and [00:03:00] then thirdly, um, this idea, uh, that you can have an identity and own your own data. Sort of closes the loop on this concept of digital assets and ownership. Um, and so you now have this tech stack, which, you know, almost a decade in, right, has evolved and matured. Um, and I was having this conversation with someone yesterday.
It's like, well, where are we on the journey if you compare this to the internet and e commerce and the scaled up applications that we see, particularly in the e commerce space? Um, And, you know, only a couple of years ago, I would have said that in the blockchain space, we were like in the sort of the pre internet stage, right?
Like before consumers became aware of the internet, which largely was around 1990, right? Microsoft had actually an interesting opportunity to 95.[00:04:00]
Basically replaced a proprietary networking stack with TCP IP, which is the stack of the internet, um, as a default. Um, so now when consumers, this is 1995, Microsoft shipped Windows 95. It was actually just the anniversary recently of that. Might have seen, uh, some videos of, um, Microsoft executives dancing around on stage to the Rolling Stones song, Start Me Up.
Which, by the way, Microsoft licensed for that purpose. Um, specifically, um, Yeah, consumers, so like anybody who was buying a PC at that point that had Windows as an OEM, you know, software stack installed on it. Uh, was getting actually proper access to the internet without any additional configuration, even though, like, the broadband telecom infrastructure wasn't quite there yet, right?
And so, that, to me, if you go back and you look at that, what does that represent? You have [00:05:00] multiple browser companies at that point in time that enabled you to get access to the internet, including, um, Netscape, and Mozilla, and Microsoft, and, and others. Um, that is the internet stack, and that's largely the internet stack that we have.
Still today, right? It's, I mean, yes, HTML has evolved, right? But, but broadly, it's that tech stack. Um, so, and then you had the five years of, you know, the, the building on top of that, right? Which, which then wound up in being the, you know, the dot com boom, essentially. And then, obviously, at the end of the 90s, the dot com bust.
Um, and what happened in 2000? What happened in 2001 was people realized this technology was mature. Right? And if you realize that this technology was mature, you can go build scaled up, massive consumer businesses on top of that. So, where we are today in blockchain is, I think we're at this point, [00:06:00] it feels like to me, that like, we sort of scale, we've solved identity, we've solved self sovereignty, we've, we've solved, uh, uh, privacy, essentially, through various approaches, including zero knowledge proofs.
Um, zero knowledge proofs are scaling very fast, uh, and blockchains are scaling very fast, right? And so, like you say, okay, well, scale, privacy, self sovereignty, uh, and then also wallet technology is maturing pretty dramatically to help protect consumers. And then you say, well, we have browsers like Opera Browser, Brave Browser, uh, the Chromium browser doesn't quite have, uh, Web3 in it yet, but Microsoft Edge, um, actually has a very early, uh, build of a crypto wallet browser, uh, inside of Edge.
What this effectively means is that we've now got this new internet stack. Right. And so to me, I [00:07:00] propose that we are actually sitting at the timeframe around 2000, 2001, when this new internet stack emerged and starts to become the default thing that's available, whether consumers care or not, right? Just like in 1995, consumers didn't care, but the stack was available.
Um, the, the next wave of mass consumer adoption or development and then adoption. Uh, can happen, right? And so, one of the things that I've been challenging myself and, and other people, um, that, that I speak with is, what can you imagine can now be built on top of this new foundation? Um, And that's where I think we are.
I think we're at this stage where you could build massive consumer applications on top of this new tech stack, uh, which is the Evolved Internet Test Stack, which many people would call a Web3 enabled, uh, internet tech stack.
Marc Beckman: So it seems to me, and part of the conversations I've been having with [00:08:00] my circle is that with the advent of artificial intelligence, we might have an acceleration in the crypto space in web three, but there is also this. Inherent tension, almost like a philosophical tension between the concept of crypto and or blockchain with AI. Um, for example, centralization, right, versus decentralization. So I'm wondering in your opinion, can artificial intelligence help crypto evolve and or can crypto help artificial intelligence evolve in certain ways? And you mentioned zero proof knowledge, zero knowledge proof, um, I think like this might be a good segue to get into the, into the.
Discussion about [00:09:00] zero knowledge proofs and perhaps how crypto specifically could help AI evolve in certain ways. But I'll leave it to you, York. I just threw out, you know, a big, I threw out a lot at you. So,
Yorke Rhodes: Well, I mean, so just going back to your question about ChatGPT Moment, um, I think the tech stack is poised, right, to get there. I would also make... Um, just a, uh, amendment to what I said as well, because I think the new tech stack also has AI infused in it, right? And so everything that we're doing now in consumer applications, both at Microsoft and every other, uh, large tech provider is looking at how to infuse AI into Productivity tools, right?
Whether it's Microsoft Word, Excel, PowerPoint, uh, et cetera, it is a productivity enhancer, right? So that becomes a tool that becomes part of this tech stack that, that you build on top of. Um, and, you know, largely if you look at what, Chat. G p t is, it is a [00:10:00] utility function for every other application that's very accessible, which is what chat g p t, um, made possible.
Um, so I think in the, you know, in the context of, um, uh, you know, the, the stack, that's what we largely have now, right? We have this very capable foundation plus this amazing utility that you can really think about as the, the future of, um, how you build applications. Um,
Secondly, um, to your point about, like, this, um, either tension between Web3 technologies and other technologies like AI, or benefits of, um, technologies, um, across ecosystems. I think, you know, one of the most amazing, uh, things that the Web3 ecosystem has provided to core researchers, including folks at Microsoft Research and others, is... The, and [00:11:00] cryptocurrencies are an example of this. We have a worldwide infrastructure. That allows you to create self sovereign digital identities, allows you to create owned assets by those identities, allows you to transfer value, right? So it's a demonstration of a scaled up worldwide infrastructure that can support those concepts.
That is a playground for anybody doing fundamental research in encryption technology, whether it's, you know, one-way hashes or zero knowledge proofs or, uh, homomorphic encryption, uh, or fully homomorphic encryption. These are all like, um, encryption technologies or identities, right? Like how do we think about identity and authentication differently based on this capability?
versus the old world of what we considered identity, which was really just logins to things like Facebook and Google and Microsoft, etc. Um, so [00:12:00] first of all, this technology foundation, one of the most interesting things about it is it actually has helped to speed The iteration of these technologies, uh, particularly in the, and you see this in, uh, encryption technologies, um, particularly around looking at fully homomorphic encryption and zero knowledge proofs.
And the reason it's speeding the iterations on those is because it's a Applied use of those technologies, right? And this is one of the challenges with fundamental research. Fundamental research is obviously required, but if you're not doing applied engineering with the outcomes of fundamental research, you don't iterate fast.
And so this Web3 substrate has allowed these technologies to iterate really quickly, and a whole community of users, the Web3 builders, see these technologies and recognize their power and are pushing the envelope. Right on, on those things that were only a couple of years ago, um, still just fundamental [00:13:00] research, um, things.
And so big benefits of having an applied substrate like that is value to other technologies, right? And so, uh, encryption is an example of that. And that's why, you know, again, we can talk about zero knowledge proofs and what that means in a minute. This applies to AI as well. Right. And so what is, what do we know that AI needs, uh, in order for AI to be trustworthy and trustful, it needs provenance, right?
And what do we fully understand, um, from a, uh, technical foundation perspective from the Web3 world? It's provenance, right? And it's not just provenance, but it's various different ways of providing, uh, consumer privacy, uh, enterprise data privacy. Uh, as well on top of, uh, a substrate that gives you provenance.
And so I think one of the added, uh, values that Web3 will help do is actually more quickly help to realize how to do provenance at scale, [00:14:00] uh, in AI. Uh, datasets and AI training and AI outputs. Um, I think that's a huge benefit. Without Web3, people would just be trying to figure out, well, what does it mean to do provenance?
How do I figure that out? How do I do that with my Azure Active Directory centralized directly of users, right? With Web3, that becomes a conversation that is, okay, AI runs everywhere. AI doesn't. necessarily, unless you're doing it in enterprise context, um, have a, uh, authentication directory, right?
Associated with user access to it, like chat2dt, you just use it, right? Um, how do I think about provenance in that context? How do I prove provenance in that context? Um, and so that's a, an advent, I think, that accelerates that journey for, for AI, um, in that way. And then I think... Yeah,
Marc Beckman: I imagine from there, like, other issues extend off of not just provenance, but issues of, [00:15:00] you know, whether or not the machine is properly trained, right, like issues that you mentioned earlier as it relates to trust, but also bias and, you know, other ethical concerns are there too. So is that where zero knowledge proof can, can come into play?
Yorke Rhodes: Yeah, so, um, zero knowledge proof essentially allows you to provide, uh, a proof of something without revealing the underlying things, um, and so that doesn't seem obvious, right, how that's valuable in a, in an AI context, but if you think about training data, you Um, you want to prove maybe the absence of bias in training data, right?
You don't need to show the actual data because a consumer would never know how to consume all that data. But you'd want to provide a trust score that says this ran through some algorithm, right? That algorithm did an analysis of the, um, the bias. You know, and the intent of bias and things like that, that you would do, um, on, on a [00:16:00] training data set.
And it got this score. That is an okay score. Here's a trust certificate, potentially with the score, that says, uh, the training data was unbiased or unbiased to a certain level, right? Um, and, um, that type of thing is quite useful, um, in a consumer context because anybody who wants to know could look at the, the trust certificate and the trust score.
And know, without having to look at the data, that it actually has the appropriate characteristics, right? An unbiased data set, right, as part of, uh, you know, the definition of how, how the AI model got to the answer it got to.
Marc Beckman: So it's kind of interesting because it also touches on the topic of proof of. Humanity or proof of being alive. And I know that, um, Microsoft has a major investment in Sam Altman's company, OpenAI, [00:17:00] Sam's also, um, one of the founders of WorldCoin, which as your probably very well aware is now using biometrics, uh, you know, vis a vis a retina scan as a, um, essentially a way to, um, apply zero knowledge proof, right?
So can you explain how that works? And then I think that's like a nice link between, um, how zero knowledge proof, um, can come into play from a practical. All right? Beyond just the technology.
Yorke Rhodes: Um, so Whirlcoin is a, uh, kind of an amazing combination of technologies, which allows you to, um, at scale, um, use biometrics to create self sovereign identities. Um, the, the, there are some challenges with the implementation. Um, and one of those, um, is, uh, you know, in an, in an encryption [00:18:00] context, the thing that.
You never want to do is hide what you're doing from an encryption perspective, right? You want to be able to, um, show people the algorithms you're using, right? That those are safe quantum, quantum resistant algorithms, for example, um, and that your tech stack is very, uh, well defined in terms of what it's doing with, with the, the data that it's ingesting.
In the case of, um, WorldCoin, you kind of have two problems. Um, one, you have. Uh, you know, a, uh, essentially a VC funded company that, yes, Microsoft's an investor in, um, OpenAI, part of Sam Altman's world. The, uh, on the Whirlcoin side, you have this technical, uh, product, which is the, uh, a physical product, which is called the Orb.
And the Orb is what does biometric enrollment and creates, you know, the outcomes associated with that. Um, the Orb uses, um, techniques to ensure [00:19:00] that, um, Uh, only one identity can be created from a biometric enrollment, so it, it actually uses their knowledge proofs to do this, um, and effectively what that technique is, is, uh, essentially is, once you've, once you've created an identity, it gets registered, um, sorry, there's a, a doorbell I gotta get here, um, let me, let me pause for a second and then, um, we can, we can, we can start that
Marc Beckman: Good. Good.
Yorke Rhodes: uh, once you create an identity. [00:20:00] Alright, sorry about that.
Marc Beckman: That's all right. So John, John, wha should we just pick up again? Like should we just let York break back into, you know, what world how WorldCoin is using biometrics to essentially prove that the person behind it is alive and Is that, should we just break back into that, Jörg?
Yorke Rhodes: [00:21:00] Yeah, absolutely.
Marc Beckman: but Jörg, break, when you go through it, um, I think it's important to talk about the retinal scans and the orbs and, you know, also my mind's going, like, that's great B roll footage, um, but it's an interesting concept, right? I'll let you just start again from, you know, WorldCoin, but it's a very interesting concept because also, I'm curious, Jörg, and I, I guess don't start here, but where does that data go?
Like, can someone access, then I realized that it's still. You know, zero knowledge. But can some, can it leak? Like, I don't want my biometric data getting anywhere. So is it like totally impermeable? All right. So let's,
Yorke Rhodes: Yeah, I'll talk about that. Okay, so, um, so WorldCoin is, uh, both a software stack as well as a hardware, uh, orb, is what they call it, essentially. So, you know, one of the concerns about the physical [00:22:00] device, the orb, is that physical devices, uh, in an encryption world... That are, you know, things like your phone, for example, which also has secure enclave technology in it.
If someone gets a hold of your physical device, then they can basically do all kinds of physical attacks on that device. So that in itself is an attack surface, right? And so the orb, right, if it's stolen, then someone can do some physical attacks on the orb, right? So that's, that's a concern. And, you know, perhaps those physical attacks are You know, breaking things open and getting into content.
Perhaps that physical attack is, Hey, somebody stole it for a couple of days. Um, they, they opened it up. They re engineered something. They injected some malware, right? Um, uh, or Trojan capabilities into the device, possibly. I'm making this up, right? Um, but that's a problem with a physical device. It's not necessarily in your control, right?
That's a vulnerability in any kind of [00:23:00] encryption and privacy, uh, system. Yeah, exactly, right? And if you look at, um, uh, in the, in the world of cloud, one of the reasons why cloud are secure is because they have a very secure physical perimeter around the data centers, right? And so that helps with people not being able to just walk into a data center and start doing physical attacks on the hardware.
So, um, that is a vulnerability in itself. The second vulnerability is the software. So what, what happens with, and these are possible vulnerabilities, right? Um, uh, and when you're doing security, you look at attack services. So the, the, what you're doing when you look into an orb is you're doing a retinal scan, right?
And so there's. There's software and hardware inside of this device that handles the retinal scan. And then what does it do with that? Typically in a biometric system, and I believe WorldCoin does the same, biometric systems, [00:24:00] um, whether you're being enrolled or whether you're actually scanning, they don't save the, like the, essentially the data that allows you to reproduce a retina.
They, they save a fingerprint associated with that. And that fingerprint is what is, uh, what is unique. So you can't go and recreate the retina to create that same fingerprint again. Um, it's basically a fingerprint, which is what, what is unique. Um, in the context of what they're doing with zero knowledge is they actually take some variation, um, probably a hash of that.
Um, and I'm, I'm being a little bit vague because I haven't read the documents recently. Um, but they largely do, do something like this, right? Where basically you, you transform, uh, this, a piece of data like, like the fingerprint of the retina into a hash. You store the hash with other scanned, um, hashes on a distributed ledger.
And now you have [00:25:00] a representation of myself through my unique retina. Um, on this distributed ledger through hash technology, and It's not the original, the ability to recreate me. It's just a fingerprint of that. And what that allows you to do is, um, ensure that there's never a second registration with the same retina, right?
So that basically creates a unique account associated with one individual based on their retina.
Marc Beckman: So what would,
Yorke Rhodes: that's done, by the way, through zero knowledge proof. You don't want access to the original content, but you do want to know, is my retina already resident in a hash table, essentially, in this distributed ledger?
Um, so that's
Marc Beckman: like by, um, by like biometrics, then York are like to, to be complete, to have like the highest level of privacy, but yet to prove that you're a living human being, [00:26:00] it sounds like biometrics are going to be the only way. To solve that problem with zero knowledge proof in mind, um, because I've, like, I've worked with a sovereign entity, uh, to get ID using zero knowledge proof, right?
Um, I actually have a, um, sovereign issued, uh, governmental backed ID from Palau. I think I mentioned this to you, but that was, um, you know, I had, I had to go through the government and, and, um, you know,
Yorke Rhodes: right, you're basically going through a KYC process, right?
Marc Beckman: correct.
Yorke Rhodes: right, and then through the government. Um, which, you know, like, if you look at what is the most recognized forms of... Id associated with person today. Right? It's some form of governmental id. Right? Whether passport or driver's license or, or what have you, right?
And so that's kind of the scaled up, um, what you would call claims, right? [00:27:00] Um, essentially in a, in a technical context, claims about this person. So for example, a New York State driver's license, uh, is a claim that York Roads, who holds that driver's license is actually okay to drive. Right, and claims are in some ways similar to zero knowledge proofs, um, because you don't have to know anything, right, about, uh, what's on that driver's license, but if I hold it, it's a second factor, so my person plus that license gives me two things, right, I am who I am, right, and that's proven for me to hold through this thing, a picture on that, that I hold, uh, and it gives me an authorization, right, to drive.
Assuming it's not expired, right? Obviously, right? So that's, um, uh, technologies in the identity space and digital identity space is something called a verifiable claim, which is basically how you recreate that scenario using decentralized identities.
Marc Beckman: [00:28:00] So, if you bring that into like the real world for the people watching and listening just to, as it relates to a person's ID specifically, um, do we really need then a, a traditional form of ID like a, a driver's license or a passport as it relates to owning and buying, let's say, objects in the future?
Perhaps, you know, a car or a home, a boat and beyond?
Yorke Rhodes: Yeah. So the, today, the government, um, a government is the issuing, uh, party behind a passport, right? Or, or driver's license. And, um, there's a presumption that you belong to that, that entity, right, and that's something, you know, like your, your residency, right, or your social security number in the context of the U.
S. Um, that same framework can, can very well work with decentralized identities. And essentially, instead of [00:29:00] having the government effectively own my identity, I own my identity and the government becomes a party that says, okay, associated with this identity, here's a credential that says that, okay, we went through this KYC process.
We know this is York Roads and York Roads has the ability to drive, right? So that credential or the driver's license becomes essentially a credential associated with my identity. Self Owned Decentralized Identity that I hold just like I hold a driver's license. Now, in a digital context, um, that's actually, you know, more powerful than the context of, you know, how do you digitize a driver's license, right?
Um, today, right, people ask you for a picture, right, of the front and the back side of your driver's license. That's not a great way to create a digital representation of, of this very important document. However, if you do it in the context of... Um, you know, a verifiable claim which is signed by the authority.
Um, New York State would essentially sign this claim [00:30:00] that says, okay, we've validated that this claim is valid for this individual who's holding it. So the concepts are Similar, the implementation of how we think about identity is what's different under the, under the, under the covers, and it creates actually, um, a more powerful, uh, way of interacting that's actually less complicated than typical federated, uh, scenarios, federated meaning, uh, issued by a, a large authority, um, the, you know, the ability to do things in the context of my, um, Identity that's less duplicative, um, is actually quite powerful.
So a great example of that is, um, you know, in the financial context with, uh, Know Your Customer and anti money laundering. Um, my identity, my digital identity, um, that has been established and that I own can have these claims associated with it where I've just gone through a KYC [00:31:00] and AML process with, say, JP Morgan, and that certificate is signed and issued by.
JPMorgan as a verifiable claim against my identity, one example. It could also be a zero knowledge proof. Those things I think are sort of overlapping and in some ways. And that claim would allow me to go to the next bank and say, You guys don't need to rerun this. I just did this a week ago, right? With JPMorgan.
Um, you know, you don't have to go through that same level of diligence. Um, I have it, right? Here's the claim. And if you really want to know that they, that, uh, JPMorgan did it, you can actually follow the trail and in fact, get a new signature from JPMorgan digitally.
Marc Beckman: And then York, what would be the benefit to something like that? If a person, let's say, is looking to operate in the metaverse, um, you know, is that a situation where in real time they can, um, [00:32:00] perhaps get a loan to purchase a digital asset while they're, you know, operating in Fortnite or, or something along that line?
Yorke Rhodes: Yeah. It, it enables, um, digital interactions. A metaverse is a, is a, is a good, great metaphor for that. Um, right, because in, in a metaverse you might, uh, want to purchase something and you don't have the enough, uh, money in whichever accounts you wanna call those accounts, whether it's, um, crypto accounts, stable coins, or, you know, Venmo, PayPal, whatever.
Right? Um, Uh, and so you want to, in a digital context, um, in the Metaverse, right, which is, Metaverses are, um, essentially global, and anybody can go into, uh, Fortnite, anybody can go into, uh, most Metaverses, no matter where they are in the world. And then, in that context, actually do a transaction digitally.
You need some verification that that person is who they say they are, right, um, that [00:33:00] essentially they're the KYC, right, and so it would enable you in a digital context to very quickly establish that yes, you've recently been through a KYC AML, um, Certification and you have it, you can present it. Um, and therefore the speed with which you could actually provide a loan to that person, first of all, you're going past the AML KYC issue very quickly.
Um, and then second of all, um, you know, you would probably want to do some other kinds of verifications as well. Like for example, um, what is this person's income, uh, or what is this person's, uh, you know, available, you know,
Marc Beckman: Proof of solvency.
Yorke Rhodes: Yeah, exactly. Proof of solvency, um, is one of the ways that, um, you know, you think about this in the context of, uh, digital assets and in the financial world.
Um, and, and, and that's also another interesting case where you could prove solvency, [00:34:00] uh, to establish a certain level of creditworthiness to get a loan without revealing a ton of underlying data. Right. And so today, proof of solvency, uh, in a physical world, just think about going and getting a mortgage is all kinds of paperwork, right?
Um, or, or any kind of loan, right? It's all kinds of paperwork from different institutions, right? Um, in the context of, uh, something like proof of solvency, That's a proof using technology like zero knowledge proofs that can prove that it has all the backing documentation, um, right, and that you have a certain level of solvency and that certificate or that proof proves that without having to actually go and look at all the source documentation again, right, um, and without having to have that source documentation available.
Um, at, you know, at the split second that the proof is being evaluated,
Marc Beckman: So I know we've gone down a rabbit hole and I want to get back to how zero knowledge proof will help artificial [00:35:00] intelligence progress, but before we go there, it seems like what you're talking about as it relates to verifiable is Personal identification as well as proof of solvency, um, through zero knowledge proof and blockchain technology can open up a new type of banking sector, could allow for people to have, you know, loans issued like on the spot for digital assets that
And you know, like I want to buy that digital sword for 10, 000 and, you know, imagine if you and I walked into Bank of America and we asked for a 10, 000 loan to purchase a sword for Fortnite, they would laugh us out of there. But these things could become actually reality. Do you agree?
Yorke Rhodes: I do. And I think, you know, like having, having lived through the, the nineties era when stock options were, [00:36:00] uh, a way that tech companies, um, you know, helped create a bigger compensation package for employees, um, banks didn't know what stock options were worth. Right. To like, they couldn't conceptualize that.
Right. And so I kind of think very similarly around like, um, different forms of digital assets. Like this is just a new form of, of value. It's not quite yet understood broadly. Although I think if you look at most, um, Thoughtful financial institutions. They very much see the benefits of, of, uh, tokenizing all different types of assets, including real world assets, because it just speeds up your ability to transact on those assets.
Um, so yeah, um, yeah, I, I do. I think there's, uh, if you go back to my point earlier in this conversation is we have this new tech stack. Right. What will we build with it? What will be the new things that get built over the next five to 10 to 15 years on top of this tech stack, which is part of [00:37:00] an inclusive AI, of AI, right.
As I said earlier, so,
Marc Beckman: So, what about the issue of, um, I touched on it a little earlier, but just to expand on it, the issue of whether or not these machines are being trained properly, can, can zero knowledge proof somehow be built into, um, verifying whether or not the, the model has been trained properly? Thank you. It seems to me like some of these large language models are, are like, you know, so massive.
Um, I don't know if CIRA Knowledge Proof is capable yet of verifying everything that's coming out of, you know, GPT 4 or, or, or other types of large language models. But, um, is there a value, um, as it relates to, to the machine learning piece of it?
Yorke Rhodes: yeah, I think you have to be, uh, uh, you know, rather than trying to like bite off the whole whole cake, right, look at a specific use [00:38:00] case and think through the use case and how there might be value, um, you know, in a, in a specific use case. Um, you know, and a great example is. Uh, you know, what are, in terms of how we show up in a digital context, um, how do you know that Mark is talking to York?
Right? That's, that's one problem. Right? And with the, with the, you know, you can see this by the way, every day, if you're following anything that's going on with AI, the ability to replicate my voice with different words. And the ability to replicate my digital presence, my visual digital presence, um, uh, to speak those words is possible today.
Um, it's just a matter of, uh, horsepower and, you know, the types of, types of algorithms they're using. And there's very accessible consumer examples of this. Um, and so that now goes to this problem, which is... How [00:39:00] do I know what I'm looking at is not synthetic? And how do I know, how do we get to this representation of these words coming out of this synthetic version of, you know, someone, someone's presence, right?
And obviously that's a very significant concern, right, from a destabilization of democracies and other You know, very bad uses of, of, uh, outcomes potentially. Uh, so then the question becomes, you know, how did you get here, um, from the perspective of, of the inputs and the models and the presentation of those, and that's effectively provenance.
We talked a little bit about provenance earlier, um, as well, and what that means is you want to understand how this digital experience was created that I'm looking at. Um, what were the tools? What were the transformations? What were the inputs? Were those actually [00:40:00] York's words from some valid source? Was the image that was created of York, uh, whether it's still or moving?
Um, what is the origin of that particular image, um, that, that is, you know, that gives it validity, right? As, as an actual human, uh, image taken with film, um, or with a camera. Um... And not a synthetic, right, uh, outcome. So both of those things are, um, uh, you know, you have to essentially go back to the source of the data, look at how the data is being transformed, have proofs associated with, um, The SORUs have proofs associated with the transformation that shows that the transformation was valid and okay.
Um, and this involves, you know, zero knowledge proofs, um, signatures, um, you know, all along this chain of custody associated with how this digital experience [00:41:00] of a person was created.
Marc Beckman: So it seems to me that with the election cycle about to heat up again, this idea of deep fakes in politics is, you know, really important. Um, I heard an anecdote, interestingly, um, it was a friend of mine in the tech space that was talking about, um, C2PA. Is that, is that it, Jörg? Where essentially a signet, what, it's C2PA?
Yorke Rhodes: C2PA, yeah.
Marc Beckman: if I'm getting this right, it's essentially, um, it allows for the, um, creator technology like a camera to have a signature on the original content, um, to show that it's, it's real, that it's, it's authentic, um, and that politicians, because of deep fakes, may consider now, literally, um, uh Going into the marketplace, having conversations, and literally recording everything that they're [00:42:00] saying to show the time and the date that they said it, and then using, um, this type of signature to authenticate what they said so they could push back on deep fakes.
Have you heard anything about this yet?
Yorke Rhodes: So C2PA is actually a, uh, an organization that was, um, co founded by some camera companies and some tech companies and, uh, so I think Adobe is part of it, Microsoft's part of it, uh, Sony, Nikon, a couple other camera manufacturers, and the idea, exactly as you described it, is we want to ensure that we know the origin of what we're doing.
Of content in a film, right? Whether it's video or moving, um, still or video, um, uh, context, right? And so, um, the C T P A actually describes essentially a signing mechanism on that content, which packages the original content plus a bunch of metadata, um, directly from the camera. That's, [00:43:00] you know, Not manipulable with, with the copy out of what I said earlier, which is, right, as soon as hardware is out in the wild, hardware is an attack surface, right?
But inside, the concept here is that hardware can hold a signature, can hold a secure enclave, can sign with this, with the key that's in the secure enclave, that allows you to create an envelope, a package, which says This is the original photo, and this photo was taken with this other associated metadata, the things that you would typically see from a camera, plus things like GPS, right, time and location, which helps you establish the origin, the provenance of that particular piece of film.
Um, Now, what happens to film, uh, you know, as you're probably aware, is if you take an original 12 megabyte, um, photo, that's not what gets published on a website, it's not what, right, it gets transformed before it's published on a
Marc Beckman: [00:44:00] going to be edited for newspaper publication, digitally, etc., right? So,
Yorke Rhodes: so you have to,
Marc Beckman: that signature then, once it's,
Yorke Rhodes: so it actually gets, it gets an additional signature, so, um, that's why Adobe is involved, right, so Adobe has Photoshop and. and other tooling which does digital transformations of media content, right? And so, in the context of something like Adobe, modifying, whether it's cropping or doing something else, maybe enhancing the colors, right, associated with a particular picture, that transformation, cropping or modifying the colors, is then a validated transformation.
The software would actually sign it with the software's key and say, this is a valid transformation. Here's the inputs before the transformation. Here's the transformation algorithm that we used, and here's the output after the transformation. So now it becomes a validated signed transformation by this piece of [00:45:00] software.
Um, so that becomes part of the provenance associated with this piece of digital media. Um, and that gives you, if you think about that, wind that all the way forward to the concept of a deep fake. If it's completely synthetic, you won't have that original signature on the content, right? You won't have the validated transformations of that content into, uh, into the final product.
And so provenance of things that are synthetic that are generated by AI is extremely important, right? Um, and, um, that is, you know, again, that going back to what's the value case of having a scaled up applied engineering substrate like Web3? Those concepts are very well understood, um, in the context of Web3.
How do we have validation of media by consumers without the necessity to have a federated authority, um, that is, you log into to go [00:46:00] find it, right? Um, so, um, and, you know, if you sort of wind that back to the early days of the web and e commerce, These types of trust certificates, which in our scenario here, would roll up essentially all of the various proofs associated with those transformations into the final product.
That trust certificate would be something, say, maybe you would see it on the screen here and be able to click on a little I in the corner, which very much like a website or an e commerce site, would give you Here's the trail of trust associated with the provenance of this data, which shows you where it comes from.
Marc Beckman: So beyond that, when, like staying focused on the concept of trust,
Yorke Rhodes: One more point, by the way. Um, in context of, um, political campaigns, what you typically see on the bottom of a political ad is this ad is authorized by, right? Um, so that could be an additional signature on the content, which if you, say, click on, in a digital context, you click on [00:47:00] this little eye and you get the trail, part of that, Signature would be that it's that it's a valid piece of content authorized by a political figure.
Marc Beckman: So Jörg, the, the, the concept of, like one of the pillars of cryptography is the idea that we have a higher level of trust. Is that, is that a fair assessment? A trust of the data that's passing between the two parties?
Yorke Rhodes: Um, trust that the data is tamper evident. Um, right. So, uh, it doesn't necessarily change your trust in the data. It changes your trust in who created it, how it was handled, right? What was the custody of that? Um,
Marc Beckman: So I was,
Yorke Rhodes: of a choice, right? Like, um, you know, it's a slippery slope here. But, um, like the concept of what.
Marc Beckman: Well, what I'm getting at is like, sometimes it's uncontrollable. I was reading this paper over the weekend [00:48:00] about anamorphic cryptography. Are you familiar with this concept? So it's
Yorke Rhodes: Describe
Marc Beckman: so, so there's a professor at Columbia that wrote this paper. The concept is anamorphic cryptography, and the idea is like, what happens when Like he was writing it in the context of like a dictator stealing, um, encrypted messaging that would be very important, you know, to a different government or something like this.
So the idea is essentially creating so, so the. One entity, the creator of the, um, the data, of the content, would actually have two private keys, and this would be a solution in protecting government secrets from, you know, bad actors, or, or, he was talking about dictators in this, in this instance. [00:49:00] It's almost like a, like a, yeah, it's like a double factor authentication process, but through cryptography.
Have you, have you heard about this at all? It's pretty interesting.
Yorke Rhodes: watermark. I will go and read about it. I had not heard about it, but thank you for the reference. Um,
Marc Beckman: I'll send you, I'll send it over
Yorke Rhodes: yeah, essentially, so this concept of like hidden digital watermarks is kind of what that, you know, that would be my summary of what that's trying to solve for. Um, uh, is, you know. That's actually a pretty common thing.
I haven't read about this specific approach, but like embedding content into other content
Marc Beckman: Yeah, exactly.
Yorke Rhodes: you know, where it came from is essentially a digital watermark.
Marc Beckman: And it plays into Zero Knowledge Proof, too.
Yorke Rhodes: yeah, yeah, I can, I can see that, right? Like I want to prove through that representation, I can provide a proof that, you know, this is authentic, right?
Um, uh, from the [00:50:00] perspective of, uh, that second, you know, the second key that's part of that.
Marc Beckman: So it's gotta be a really exciting time to be at Microsoft right now. In June of 2023, McKinsey issued a report estimating that artificial intelligence technology could add 2. 6 trillion to 4. 4 trillion dollars in annual value to the global economy by making an array of processes more efficient and effective.
So, that's really interesting. Evercore predicted that AI could add 50 to 100... billion dollars to Microsoft's annual revenue by 2027. Um, and this past February, in February of 2023, your CEO, Satya Nadella, said, I've not seen something like this since I would say 2007 2008 when the cloud was just first coming [00:51:00] out.
you, Jorg, is how is, how do you expect Microsoft to capitalize On artificial intelligence first, and then, um, to follow up really beyond the obvious, uh, for the people that, you know, are into tech and following tech, like, which new products can we expect and what utilitarian type of features will these products have?
Yorke Rhodes: Uh, great, great question. So I think, first of all, um, you know, describing AI as a utility is complete, completely valid, right? It's, uh, how does, how do you leverage this new technology to make you more productive in the things that you're trying to do? Um, this fundamentally changes, like, writing and blog writing and the productivity that you can actually achieve, whether you're a student or, or a writer, um, or a researcher.
Um, you know, just having the facility of this assistant. which helps, you know, uh, do things that you would normally do yourself. Um, and then you [00:52:00] become more of an editor versus, uh, a writer. Um, that's quite useful. And I think we're going to see a lot of that. Um, there was a great lecture by, um, uh, Ethan Mollick, I don't know if you know that name, uh, he's a Wharton professor, um, and he, um, teaches, um, entrepreneurship and, um, he's obviously very down the AI, uh, AI rabbit hole, um, and one of the questions he asked actually, um, in a session that, uh, participated in as part of my NYU instruction, uh, last week was, uh, How many people in the audience have spent more than 10 hours working just with things like ChatGT and OpenAI and, you know, essentially the tooling, whether it's mid journey or, or, you know, writing tools and things like that.
Um, and it's a really interesting, uh, data point. Um, and, and, uh, With it kind of pushes you to say, well, am I using this technology enough? Right? It's available today. Why are not I not using this technology [00:53:00] more? Right? And so, um, as I go into the teaching in the fall semester, I'm going to totally expect that my students are using AI to help them be much more efficient in writing blogs and things like that.
Right? And so what does that mean from a, from an instructor perspective? I'm gonna, yeah. Assume that the content is going to go up in terms of, um, right output, right? The, the sophistication of the content, um, should ideally be, be higher, right? And the level of the level of, uh, writing should be higher because now easier for you to get to the end product.
That's just productivity, right? And so, um, you know, these, this type of utility will show up in every productivity tool that we use. You know, whether it's Word, PowerPoint, Excel, um, you know, or, you know, any kind of other modeling tools, um, we're going to see the add on and the benefits of AI associated with that.
And in a programming context, it's actually been [00:54:00] happening for quite some time. If you look at the, the GitHub Copilot, um, capability, it's been in the market for a while. Um, It helps, what I just described about blog writing, it does that for programmers. It helps programmers more quickly arrive at something that's a reasonable implementation of what they're trying to achieve.
Um, and then, obviously, you go through and modify things you need to tweak. Um, and, additionally, um, uh, Ethan Mollick actually used this example. Um, he's not a programmer, but he's able to actually write Python code
Marc Beckman: It's
Yorke Rhodes: to... Right. To show visualizations based on the data that he's working with because of AI.
Right. And so there's, there's this very transformative and, um, sort of, uh, productivity, uh, transcendent capability that AI gives us, which is why I think the term utility is really important. Uh, utility, because it's [00:55:00] not like you're going to create necessarily a new product, right? You're actually creating value through the technology in existing.
Productivity applications.
Marc Beckman: So, so what does that do for, for like innovation as it relates to mankind? Um, like, are you saying that the new language for coding is English? Or will be English,
Yorke Rhodes: Yeah, there's definitely people who are saying that, right? Like, the fact that you can construct a sentence um, that can be transformed into code and the language of choice of that code can be many, right? Um, is quite powerful, right? It, it now, obviously when you have your tools on something, your, your fingers on something powerful, you have to have some context of like, what is that thing doing?
And what does it mean? Um, at least in the early stages. So I think we will continue to see lots of examples of that. Um, [00:56:00] I mean, there are examples of, of, uh, that on the internet, like there's a, uh, I don't know if you're familiar with the, um, baby AGI. Um, application.
Marc Beckman: I've heard of it, but I've never used it.
Yorke Rhodes: Okay, so there's a, what baby AGI in terminology, you know, put terminology aside, but AGI stands for Artificial General Intelligence. The concept of baby AGI was, how do I use AGI, AI as it exists today to try to get, you know, sort of closer to that. And what baby AGI is, it's basically taking multiple steps.
Uh, feeding that into AI, AI creating some output, taking that output, feeding it into the next step of things that I need to do to, you know, get to an end in the state I'm trying to get to. Um, and this interesting and also quite viral, uh, thing that happened on the internet in the last month or two, guess who wrote it?
[00:57:00] It was written by a VC, like, not by a programmer.
Marc Beckman: Right. This is my point. So how will that impact? Like, so, so if, if the argument is that now English will be the programmers. Common language that means that it opens up to people with no, no computer science background. Right. If a VC wrote right, if a VC wrote that, how is that going to advance science medicine?
It's, it could fuel job creation, it could have impacts on, you know, the film industry and, and, you know, and, and beyond.
Yorke Rhodes: Uh, yeah, I mean, if, if, so a couple of things. One, um, we're having a conversation in English. Right. An example of where a large language model in real time would be quite useful is if you don't speak English, right? We could be having a conversation that's translated on the fly, right? Um, that capability has been around for a [00:58:00] while.
It just hasn't been fast enough, right? Um, and so it's a great example for things like Zoom and Teams and, you know, discussions like this where you could have, um, language models translating from English into other languages on the fly. And, you know, you can imagine this would be this type of thing that happens in the UN, right, in the UN when someone speaks in a particular language, there's all these translators, right, that are human beings translating, um, you know, the, the answer.
So, um, you could do that more quickly and you could still have human co pilot, right, intervention, um, which says, okay, I agree with that translation, right, um, or I would tweak that translation, right, use this word instead of that word. Because the people who are doing those translations are highly trained, right, around the nuances of the subject matter areas, and so having That copilot, you know, scenario would be useful there.
So that's, I mean, that's, you know, I think we're going to see just [00:59:00] these massive changes in the assumptions that we make about communication. Um, uh, in addition, you know, to your point about the film industry, there are. So many examples on the internet of, like, short films and feature length films that are completely created, um, through someone using AI prompts, right, to generate outputs, um, and, um, a lot of techniques, actually, that have been So Performed, uh, for years just can be enhanced with, with AI to more quickly get to the outcome.
So I think we'll see, um, in almost any kind of digital media, a, uh, speed up in the ability to arrive at the outcome. And then film and video and feature length films, uh, and film production and games, um, and metaverse experiences. Those are all... 3D context, digital media experiences.
Marc Beckman: So, so will Microsoft, is this part of the [01:00:00] evolution of Microsoft? Will Microsoft take on this new type of, um, almost like a new form of Microsoft, a new entity of sorts? Um, with open, uh, open AI really, you know, arguably leading the path, um, of, of, uh, uh, You know, AI, artificial intelligence and Microsoft having such a significant ownership stake in open AI is the, is the Microsoft company evolving again?
Are we seeing this new type of entity unfolding before our eyes?
Yorke Rhodes: So not speaking as a Microsoft spokesperson, but like speaking as somebody who's observing, um, uh, I think there's, there's two lenses to that. One, um, you know, Microsoft's already spoken about having this capability in all of its productivity applications. Um, that's, that's one. Um, uh, and, and as an example, there's a.
Microsoft, I think it's called Microsoft Create, that's now part of the browser, [01:01:00] um, that actually uses Dolly, right, to, to create imagery and to help with, like, you know, things you might want to do that type of imagery for, um, you know, to your point about advertising, um, you know, that speeds, or production of advertising content, speeds that up, um, so Microsoft will evolve its technology landscape, right, from, uh, every application, right, we'll, we'll, I want to take advantage of uh, these, this utility and these capabilities.
So, I would fully expect to see across, across the board at Microsoft uh, you know, just working on where does it make sense to infuse AI into particular products. And that could be the things that we think about that are Microsoft products like Office. Um, it could be in Teams, it can be in LinkedIn, it can be in Dynamics, it can be in core Azure functions, um, right?
Um,
Marc Beckman: It's an exciting time.
Yorke Rhodes: yeah, [01:02:00] there's, there's a lot of landscape. I mean, I
Marc Beckman: I heard Elon comment that, um, uh, although he was an, obviously an early investor in OpenAI, I heard him comment that Microsoft controls OpenAI. Did you hear that?
Yorke Rhodes: think, I think Elon likes, first of all, likes to hear himself talk, and second of all, loves to be provocative. Um, so, um, I, I had not heard that, but it's not a comment I wouldn't put past him, right? Um,
Marc Beckman: But, but another, um,
Yorke Rhodes: by the way, one other data point. So, um. Microsoft has a lot of its own products, but the products that are built on Microsoft technologies is actually vastly bigger, right?
And so those are all of our partner partners, right? People who we help build, help understand, you know, the Microsoft landscape of tooling where they can build applications on top of that, whether that's a cloud native startup or whether it's, you know, other types of companies. Like, SAP obviously is a big partnership with [01:03:00] Microsoft, even though Microsoft has Dynamics, which is a somewhat competing product in a slightly different market segment.
Um, the partnership and the ecosystem is extremely important to Microsoft, and if you look at some of the most recent partnerships that we're doing, forget OpenAI, which is just a huge unicorn type of, you know, multi billion dollar partnership, but look at down from there, right? The smaller partnerships.
They all, in some way, play Um, actually, um, include, uh, the value case of AI and those technologies, whether they're a startup or, or in any particular sector, right? So we're going to see that, uh, again, this is an enabling technology. AI is now a part of the toolkit. Right. The next thing you do for the next five to 10 years is going to have AI as part of that toolkit.
And then, you know, as I said earlier, this, the tech stack of the internet has evolved to include essentially AI and in the
Marc Beckman: well, [01:04:00] we, we,
Yorke Rhodes: scenario, web 3.
Marc Beckman: we, you know, we spoke about the value of Web3 in, you know, the gaming metaverse earlier, as it relates to Fortnite, proof of solvency, proof of, of, you know, being an actual living human. Um, etc. But part of the transformation that Microsoft seems to be taking on also is the community that it's buying into or bringing along with regards to the, um, 70 billion dollar Activision Blizzard acquisition.
And all of a sudden, um, unlocking these massive communities that are very loyal in Call of Duty and World of Warcraft and beyond. So what's the impact? Uh, to the Microsoft ecosystem, when you bring these huge communities, very loyal fans, great intellectual property assets too, what's the impact on, on Microsoft from that perspective?
Yorke Rhodes: You know, it's [01:05:00] interesting, uh, like obviously Activision is a very large acquisition, assuming it's approved and goes through everywhere. Um, um. And again, speaking as an outsider, it is, um, part of Xbox, right, and so that's a very big division at Microsoft that's specifically focused on gaming, uh, and, um, you know, it's, it's a very sizable acquisition in, in the gaming space, so it will have a transformative impact, um, you know, on what we imagine as Microsoft gaming in terms of, um, the different studios that, that build product in that space.
I think, um, you know, we briefly spoke about this as well. It's like both digital ownership and AI have a massive future in, uh, the context of gaming, um, related to asset ownership and things that, that gamers collect that, you know, today they don't really own, they're really renting them, um, is one, and then [01:06:00] secondly, um, as you look at the utility of, of AI Even in its most nascent form around creating digital media, that has got to transform the work associated with bringing games to production, right?
And so I think that's going to have a big impact as well. I mean, like, typical AAA game production cycle is four years. And it's just a ton of digital content. It's very, it's very similar to, and more complicated in some ways, than creating a feature film. It's now you have, it's a multiplayer game, right?
So how do you, how do you think about that?
Marc Beckman: Right. Um, so, so you mentioned like If the transaction is ultimately approved, I think it's making some progress in certain regions, but, you know, your comment, um, obviously puts my mind into where we are here in the United States. Um, it seems like the United States [01:07:00] government, to a certain extent right now, is like anti Biden.
Big tech. I'm not even talking about on the regulation side. We could get into that conversation in a second as it relates to the SEC and the confusion, um, to cryptocurrency and all, but do you feel like there's a general United States government sentiment surrounding big tech right now where they're just...
Kind of against it.
Yorke Rhodes: And I think, uh, it's a good question. I think the United States government's a complicated, uh, set of bodies, right? Um, that has, uh, with individuals, right, which, um, in an ideal world are responding to their constituents, um, you know, in places like Congress and, uh, and the Senate. Um. The, and I think you hear probably a lot more noise, um, generally than, you know, what may be actually the reality.
Um, and I think as, as [01:08:00] outsiders, you know, to, uh, to that, I think a lot of people probably ascribe different motivations. Um, and so I, you know, I just I think, you know, frankly, politics is very complicated, right, in that way, and ultimately it boils down to people and what people believe and what their constituents want, um, uh, and, um, so I don't know.
I don't, I don't have a definitive answer there. I do think that, um, if you compare the United States to other geographies that have embraced, uh, different types of technologies, um, you can see a concern about, like, Slowing down innovation, um, and slowing down, uh, progress, right, which is what innovation drives.
Um, where, if you, if you think about it from a competitive landscape perspective, other geographies seem to be farther ahead in certain categories. You know, that being said, like, there are [01:09:00] definitely, like, consumer protection issues across all these technologies we've been talking about, um, that, that, you know, someone's got to look after, right, um, and typically that's been, you know, various parts of, of governments, um, you know, in the US or other geographies in the world, um, um, so, yeah, I mean, I, I don't, I don't really have a, a strong view on that, um,
Marc Beckman: So, but trust is such a big issue. Like, as we were, you know, we're evaluating the Microsoft ecosystem today, and we've touched on so many different issues as it relates to the centralized force of Microsoft touching people's lives at home, while they're gaming, in the office, um, now through the use of machine learning and artificial intelligence.
Decentralized as it relates to blockchain technology, um, how will Microsoft be able to provide, [01:10:00] um, a certain level of confidence as it expands its ecosystem and its reach that the customer's privacy will be protected?
Yorke Rhodes: It's a, it's a great question. Um, Microsoft, one of the, I think you can find this on the Microsoft website. It says something like Microsoft runs on trust, right? And so how do you establish trust from a technical foundation perspective? Um, that's where, um, you know, you have to think, you have to be very thoughtful about, um, how vulnerabilities are introduced into both hardware and software.
Uh, and you have to put in place controls that, um, prevent those from happening. Um, you have to, um, uh, think about when a customer runs a workload, uh, what, how do you help that customer understand, um, how to create a workload that's safe and secure and doesn't reveal secrets, um, as an example, um, [01:11:00] uh, how do you ensure that Um, that customer is, uh, who's running that workload is thinking about GDPR and, you know, consumer privacy and consumer data.
Um, those are all one sort of practices and, and guidance that we provide to customers. But as a technical substrate, we think very deliberately about. What does the software supply chain look like? What does a software stack look like? How do you make sure that a software stack has efficacy of provenance?
That's actually a big area that is in the U. S. There's an executive order specifically focused on that technical area. And that's really important, right? Because if You inject malware or some kind of backdoor into a software stack without knowing, without people knowing it, right? It's basically becomes a [01:12:00] dormant attack vector, right?
That can be awoken at any point in time. So, hence why the U. S. government is, um, has created an executive order around that topic. Um, Whether it is, um, uh, the, how do the workloads run in a trustful context, right? How do you ensure that when keys are generated to create encryption around, uh, data at rest, data in transit, um, and data in essentially in memory or being computed on, um, what are the tools that you put in place to ensure that, that Encryption key, right, is not vulnerable, right?
And how does that encryption key generate it? Um, who holds that encryption key? Who has access to that encryption key? Um, and, you know, enforcing things like rotation of keys and things like that. So there's a lot of policy and control, um, around systems that, [01:13:00] um, really create, uh, a, an environment that is a trustful environment to operate.
Uh, code in, right, which, and underlying every application, right, that, that we talked about earlier, whether it's Office or Activision or LinkedIn or Dynamics or third party workloads like SAP, it's all code, right? And so, that is a technology foundation that you have to ensure from, uh, the physical, uh, Uh, uh, plant, meaning the physical, you know, guardrails around a cloud data center to the physical devices inside the center, to their firmware, to the software that runs on them, et cetera, et cetera, all the way up the stack, right?
Um, all of those things have to be part of that trust.
Marc Beckman: And yet, it seems like Microsoft is even going deeper into capturing different types of data from individuals. Your CEO, [01:14:00] Satya Nadella, referred to this time period as the co pilot era, which is really interesting, um, particularly as it relates to product now, perhaps, beyond, um, programming. So, for example, I noticed with the United States Patent and Trademark Office that Microsoft just filed for a patent surrounding an artificial intelligence powered Backpack.
Have you heard of this?
Yorke Rhodes: I saw that over the weekend also.
Marc Beckman: Yeah, so it's incredible. It's filled with sensors. Um, they provided a few illustrations, one with an individual who's skiing. I think another was in front of a Beatles poster. Um, but again, capturing tons of data so that this backpack is no longer just an object holding, you know, a child's school books and pens, but rather really serving as a co pilot of sorts.
Yorke Rhodes: So I think, well, yeah, let me, let me speak to the patent in a [01:15:00] second, but the terminology copilot is actually a really good framing for, for AI. And I think I used it earlier in the discussion, but basically it's like you have all this power. Right. That, that power is really a co pilot to the pilot and you as the consumer of the pilot, right?
And so the example I used with, um, translators, right? The translator actually is the pilot. The AI that does the, you know, the quick translations and first iteration or first draft is, is a co pilot, right? And so that's a really important, um, contextualization of how we think about the power of all of this AI utility, um, that ultimately the consumer's in control.
Right. Of what gets approved, you know, into the final product. Second point going back to the patent. So I don't know, I haven't seen anything internally. I don't know anything about the internal filings. As you mentioned, there is a public filing that was recently revealed through the patent office. What's interesting though, and my first impression was, well, what's in that thing?
Right. Um, [01:16:00] and so my first reaction is literally everything is in that thing is also in your phone, right? Every sensor that's in the backpack is in the phone already, right? So we already have walk around with a something that. It, you know, has all those same sensors, and I think the, the utility of the backpack as you combine it with other technologies is that a backpack gives you a better way to carry heavy duty computing equipment, including batteries and things like that, um, that frees up your hands.
Right, because a phone, obviously, you're manipulating with your hands. Um, and so that, I think, is a useful context in a, in a lot of different, uh, ways. There was also, I think, some recent filings as well around, like, um, different types of, uh, headsets, right? Um, you know, and the range of headsets, by the way, is, is vast, right, from things that look like, Glasses, you know, to [01:17:00] full VR headsets with, you know, an inability to see through the front of the headset like an Oculus, right?
Um, and I think the, the filing related to the headset was, and it might have been referenced in the same article that I read, was basically that, um, these, um, cameras would allow you to, um, ingest the external environment into this new headset. Um, and then manipulate imagery like you do with AR VR, um, but not have the, the AR experience being provided through your eyes, have it actually being provided by a camera, which then gets fed into a screen, which could be opaque, right?
Just like an Oculus. Now, whether that's good or not, I don't know, but the way it was described in, uh, in this presentation was that it gives you, uh, A more powerful and seamless way to integrate, um, the, the parts of the experience together that you would typically see through AR, um, because you're pulling in the external environment, but [01:18:00] the way you're doing it, it's already digitized, right?
And so it gives you a more powerful way to, to manipulate that. Um, so just tying that to the backpack is you can imagine that now you've got a backpack, which is very powerful. And you've got a headset, which is very powerful and those could collaborate together to give you basically a experience that's a completely hands free experience that completely understands your surrounding, um, kind of feels a little bit like an autonomous vehicle, right?
Or an autonomous capability that. You know, does things that we've always imagined things like glasses would do with an AR overlay, right? Um, so, you know, I, I don't know. I, you know, whether, whether that's real or not, um, you know, we, obviously today you can buy laptops that have some of that capability. I'm sorry.
Today you can buy backpacks that have some of that capability. There's a lot of backpacks that, for example, that have battery, um, battery chargers. Um, there's backpacks [01:19:00] which have sensors in the market. Um, so is that an evolution of the backpack and, uh, you know, in a, like, as we think about, um, uh, in a fashion world, IoT enabled clothing, right?
This is now, um, essentially an IoT enabled backpack, right? Um, seems like a natural evolution of possible, right, um, possible use cases, uh, in the world.
Marc Beckman: Jörg, I've had you for the entire morning, um, you know, I don't know, I was gonna get into Apt, the, um, Aptos Labs partnership, but I feel like let's, let's wrap it up, right? I feel like Right, I'm sorry. Um, before we go, like, just one thing that I do with every guest is we kind of do this, like, predicting type of thing where, um, I give you, like, you essentially say, like, in some future day, and then I, like, give you, like, a prompt, right?
So, um, do you mind doing, like, one or two of those with us? Real quick. [01:20:00] So it's just like a sentence, right? So like, for example, in some future day, my co pilot will be,
Yorke Rhodes: uh, essentially a digital twin of myself, right? With a full, um, identity as an agent that I authorize, um, which has a combination of, um, you know, this tech stack, which I talked about, which is the, um, the full Web3 tech stack, which respects my self sovereign identity. My digital ownership of things, um, and the utility of, um, AI, right?
Which can actually speak in a digital context on my behalf.
Marc Beckman: and then finally, in some future day, Microsoft, the company, will be.
Yorke Rhodes: Uh, it's a good question. You know, I think, uh, the company will have products that evolved to, um, [01:21:00] to include the utility of AI, um, across, across the product suites to the point, uh, where it is just infused in a product and consumers don't actually have to invoke it, um, in the same way. Um,
Marc Beckman: Awesome. York, thank you so much. I appreciate it.