Satoshi’s Legacy: Giving AI the sense of space

I'm quite sure you can hear me all the way in the back. We don't need any microphones, right? And you are recording? Great. I'm going to be talking about what I believe in the fullness of time will be understood as Satoshi's legacy.

I'm going to make the spicy claim that it's not going to be Bitcoin. I'm going to try to weave a story here about how Satoshi kick-started a movement that ultimately leads to decentralized machine perception, how AI will physically see and experience the world. And to get started, I want to point out that of course Rome wasn't built in a day, this is a very famous saying, but we also saw this weekend that a lot can be built in a weekend. So it's worth just thinking a little bit about the value of time, especially the value of time at civilization scale. I used to live in Beijing for seven years, and in Beijing there are more cars on the road than there are people in Los Angeles.

And if you do a quick napkin calculation, you find that every week the commuters in Beijing collectively lose about as much time as it took to build the Great Pyramid in Giza. This is some unintuitive math. If you want to just fact-check me on this, you can count 20 to 30 million people losing two hours of commute every day. So about every week we lose one pyramid worth of human productivity just in the city of Beijing. But of course cities get much bigger than Beijing.

Today I live in Hong Kong. Hong Kong is this little part here of this big thing here called the Greater Bay Area. Greater Bay Area. The Greater Bay Area has a population of 87 million today on an area that is technically smaller, again, than Greater Los Angeles. Two-thirds of the size of Greater Los Angeles, but 87 million people live there today.

By comparison, we can look at New York, the biggest city in the West. New York has 900 buildings taller than 100 meters. But in Hong Kong, where I live, there are over 4,000 buildings taller than 100 meters. And if you look at the Greater Bay from space, it looks something like this. This is Hong Kong.

This. This is Hong Kong, right? This little part is Hong Kong. This is the Greater Bay Area. From space, you see this is one city.

And this one city has more tall buildings than all of North America and all of Europe combined, right here. And what does this mean for the future of the economy? What does it mean that cities are getting this big and that so many people are going to live here? Well, we haven't started counting the AIs and the robots yet, right? So if this place will have 100 million people living by 2030, living by 2030, how long until civilization is at 100 billion decision makers?

I believe we will get to 100 billion decision makers at civilization scale within the next two decades because of AI, robots, drones, etc. And as a result, I think we're going to see the whole world economy I think we're going to see the whole world economy basically 10x within the next two decades. If that sounds nuts, you should go back and see how many times the world economy has grown over the previous two decades. 10x over the next 20 years is not crazy. But what's going to be different this time around is that a lot of the world's transactions will be machine to machine, AI to AI, and we're going to be looking at perhaps 10 to 100 million transactions per second.

This is what civilization will need to function. Of course everyone in this room, everyone that's involved with Ethereum and blockchain in general, knows that we need to have AI be able to exchange value at computer speeds, and that that also means that we need to have a form of value native to the internet. So for many, many years, for more than a decade, the intellectuals in the blockchain space have been talking about we need to have some kind of internet money for the future of machines to be able to trade with each other. But there's another missing protocol that's important to talk about. What does an AI actually want to buy?

This is a question we don't ask ourselves enough. What does an AI actually want to buy? If there's 100 billion AI out there, what will they spend their money on? Probably not NFTs, probably not even Bitcoin. What do they want to buy?

What do they need? What does an AI need? On a recent podcast, this was two to three weeks ago, Elon Musk, weeks ago, Elon Musk, I'm sure you all know this guy, was asked by Lex Riemann what will it take to win the AI race and he said well you need two things you need access to the best compute which is why he's building this big GPU cluster in Houston I think and he also says you need access to the best data. Okay let's take a step back access to the best data. Okay, let's take a step back.

Access to the best compute. There's a lot of builders in blockchain already trying to build decentralized GPUs for training, for example. This is already a very, very crowded space with people like Akash and Aether and all of these guys. But also, you need good access to data. But also you need good access to data.

But one thing that Elon said that I think we should pay attention to, he said that all of the digital information that humanity has produced so far is actually very little in the grand scheme of things and probably not enough to win the AI race. Even if you had all of the information that humanity has produced to date, it wouldn't be enough to win the AI race. But, he said, reality scales. And then he said something that should really scare us in the decentralization movement. He said that Optimus, the humanoid robot that he's building, will be the greatest source of data in the world.

the greatest source of data in the world. The greatest source of data in the world. He explained that hundreds of millions, maybe billions of these robots will be distributed across the world, recording data about the real world. That means the robot you're buying and you're putting in your kitchen knows what's happening in your home. We've known for a long time that Mark Zuckerberg knows basically everything we write on the internet, but Mark Zuckerberg never had the power to walk into our bedroom, open our bedroom drawer, and see what's in there.

Elon wants that power, and he wants that power because winning AI is at stake. Optimus will be the greatest source of data in the world. Robots collect data about the world because they need to be able to navigate. The problem is that robots and AI lack a sense of space today. This is one of the big challenges in robotics.

They don't understand the world very well. They're just starting to learn how to perceive the world. Ultimately, they need a spatial semantic representation of the world. Spatial semantic means spatial, meaning the space, semantic meaning the things. It needs to understand where in the world are the things, but it's very limited to the sensors that they have on board.

Let's go back to the problem with Beijing traffic, right? We might imagine that in the future self-driving cars will coordinate with each other and you won't even need red lights, the traffic will just go at 150 kilometers an hour all the time. Okay, if and only if the cars can actually communicate with each other, negotiate with each other, but also express space in a universal way. The GPS is not precise enough to communicate about where you are in a way that will avoid collisions. If you have two cars coming like this around the corner and you want to know which one of you needs to slow down or if one of you needs to slow down you need to know precisely where you are precisely the speed that you're moving at and the precisely where you are part is not solved by DTS.

So robots like this really cool Chinese robot by Unitree they need to be able to collaboratively understand the world, not just rely on their own sensors, but get sensor data from other machines, other cameras, other things in the world to be able to navigate. Many world governments have understood this already. The US government, for example, has rolled out something or are rolling out something they call V2X, which is infrastructure for cars to talk directly to each other but also to sensors related to the road. So what does an AI want to buy? I believe this is the answer.

An AI wants to buy And AI wants to buy collaborative spatial computing. Spatial computing is the art of teaching machines to understand the physical world. It's things like computer vision, but more than that. It's understanding its movement, it's understanding its surroundings, and you need to do this collaboratively. The hundred billion robots and AI in the future, what they will need to be able to purchase from each other is spatial computing resources, compute and data.

But collaborative spatial computing will have impact more than that too. The Vision Pro was released a while ago and when Tim Cook announced the Vision Pro, he said that just as the Macintosh introduced us to personal computers and the iPhone introduced us to mobile computers, the Vision Pro will introduce us to spatial computers. He didn't say headsets, because the headset form factor is actually not what's interesting here. What makes it a spatial computer is that it's a computer that understands its surroundings. But the Vision Pro is very big, right?

It's very big and very heavy. Why is it so big? Well, because you need to cram a whole laptop worth of compute onto your face. But with collaborative spatial computing, you can move a lot of this compute away from the face and have smaller glasses. In fact, in our lab in Hong Kong, we are right now building the world's first collaborative AR glasses where the compute are not in the glasses.

And those glasses today only weigh 50 grams. Those glasses today only weigh 50 grams. The Vision Pro is between 300 and 400 grams. So when you move to collaborative spatial computing, spatial computers get significantly smaller. So the Internet does not just need Internet money.

The Internet also needs a universal spatial computing protocol so that robots and AI can communicate with each other in a mutually intelligible way about the physical world. Elon understands this. Apple understands this. Microsoft understands this. In fact, seven of the ten biggest companies in the world right now are spending enormous amounts of money on spatial computing every year.

They understand that the GPS cannot be the future and that navigation will be based on perception. So Microsoft has built what's called a VPS, a visual positioning system. Apple has built a visual positioning system. Google has built a visual positioning system. Google has built a visual positioning system.

Tesla has built a visual positioning system. NVIDIA has built a visual positioning system, and so on and so forth. What's the problem with this? The problem is visual positioning works by comparing what you see to a digital twin of the world. But the digital twin of the world is much too big to fit inside your device, right?

The digital twin of the world. But the digital twin of the world is much too big to fit inside your device, right? The digital twin of the world does not fit in the Apple Vision Pro. So instead, what you do in all of these solutions is that you send your camera feed to their cloud and they will tell you where you are. This is what the world's biggest tech companies believe is the future of positioning.

They will get to know what you are looking at at all times for the AR glasses to be able to function when you leave your home, for the robot to be able to navigate when it leaves your home. This should be something that the decentralization movement should care about a lot. For me personally, I think this is more important than decentralizing the financial system, decentralizing the dollar. Society is functioning poorly, but functioning with the monetary system we have today. But we don't know just how bad surveillance capitalism will get if these big companies literally get to see what we're looking at at all times.

I think this is one of the biggest challenges for us as a civilization and one of the biggest challenges and opportunities for the decentralization movement. We need to ask ourselves what comes after the GPS? I believe that what comes after the GPS is deepened, and that deepened in the fullness of time will be what we understand Satoshi's legacy to be. Deepened, the concept of decentralized physical infrastructure networks, collaborative networks of people that come together to deploy civilization-scale infrastructure that is funded, owned, and operated by the people is a new chapter in human economic history. For the first time, these kinds of civilization-scale projects can be done by a collaborative group of people rather than a government or a large organization.

or a large organization. And I believe that the most impactful deep end that can be built will be the decentralized machine perception network. Why? Well, according to the World Bank, 70% of the world's economic activity is still happening in physical places that require physical labor. These places cannot be aided by AI unless AI learns how to perceive and operate in the world.

So as revolutionary as AI is already, it's only able to help a small fraction of the world economy. If we want AI to help us across the world economy, we need to teach AI to see the world and to see the world economy. If we want AI to help us across the world economy, we need to teach AI to see the world and to see the world together. So the decentralized machine perception network, I think, will be the greatest D-pin that civilization ever builds. We are trying to build it with something we call the post-mesh.

It's a a deep in for AI perception or an external sense of space for machines. The way it works is that a venue like this can choose to self-host their digital twin and their visual positioning system with an interoperable protocol that allows any visiting application, AR headset, robot, drone, whatever it is, to connect and already understand the surrounding. So you don't have to observe the whole world. You could just know as soon as you step out of the elevator how to walk to this podium. This is what the post mesh allows.

It's an external sense of space that devices can connect to. It's already live. We already have real-world use cases. We're focusing a lot on us building it are very focused right now on helping retail like grocery stores helping the staff in the retail store know exactly where every product is at all times but also use collaborative spatial computing to allow them to communicate asynchronously with invisible post-it notes in space, things like this. We believe that just for that use case, we'll get into 100,000 physical locations over the next three years, capturing on average $300 per location per month.

Quick napkin calculation, that's a cool $0.36 billion in annual recurring revenue. Quick napkin population, that's a cool point, 36 billion in annual recurring revenue. And that's just for grocery retail. Again, the World Bank says that 70% of the world's economic activity is happening in these physical places.

Retail, hospitality, actual hospitals, healthcare, warehouses, construction. So we are building the eyes and ears of AI at civilization scale. And on Wednesday, we are launching our token and thereby also launching our grant program. So starting Wednesday, we'll start writing grants for builders such as yourself that want to build something on this protocol or collaborate on the protocol itself. That's why I came here from Hong Kong to see you guys here in Tokyo, to invite you as part of the decentralization movement to spend some of your cycles thinking about how can we prevent the surveillance dystopia of optimists being the greatest source of data in the world and how can we make the decentralization movement care about not just the privacy of our money but the privacy of our spaces.

If you're interested in this you can go to auki.ai slash whitepaper to read our whitepaper or follow us at PostMesh on Twitter. Again, on Wednesday, we launch the token and our grant program. I look forward to hopefully building something cool with you. Thank you very much for your time.