Why AI and the blockchain need each-other to access the physical world
AI Oracles super empower Real World Assets on the blockchain — and super empower AI to act in the real world
In the (near) future AI Agents will become Non-Human customers for RWA Smart Contract Rails
Artificial intelligence (AI) has emerged as a breakthrough technology in the past year, particularly in the development of limited AIs designed for specific tasks. These AIs, such as ChatGPT, have demonstrated significant progress in their ability to respond believably to inputs. In order for AI to reach its full potential, it needs to be able to interact with the physical world. Mattereum’s Asset Passports (MAPs) on the blockchain can provide a bridge for AI to interact with physical assets in a trustworthy way, enabling them to identify, buy, sell, track, and move things. The integration of AI and the blockchain has the potential to revolutionize industries and change the way we live and work.*
It’s not often that you can point to the moment that the future arrives, that there’s been a phase change in reality, that the paradigm has shifted and whole new vistas of possibility have suddenly opened, but this is one of those moments.
As 2023 dawns, it is clear that the breakout technology of the last year has been artificial intelligence (AI). For all the hype about the Metaverse from certain tech companies, it has struggled to gain popular traction; self-driving cars have shown that they are years away from being reliable enough for popular adoption; nuclear fusion has achieved a crucial milestone, but is a long way from practical rollout. The blockchain has gone through its frenzied bubble phase and is now undergoing its “Great Disinfection” before finding its mature role. The blockchain’s time will come, and then it will be massive; but for the moment it’s getting purged of criminals, scams and gimmicks before being repurposed to become the world’s operating system for the future. but AI has really broken out of the pack, and it has broken big.
AIs are astoundingly better at responding believably to inputs than anything seen before
We’re not talking about general AI, the William Gibson/ Iain M Banks sentient human-plus intelligences (yet), although Google software engineer Blake Lemoine did cause a stir back in July by claiming the chatbot he was working on had achieved sentience, swiftly followed by Google firing him. It’s the more limited AIs, developed to carry out specific tasks, that have made the impact. Things like ChatGPT are astoundingly better at responding believably to inputs than anything seen before, and the improvement in image generators like Stable Diffusion, Craiyon and Midjourney has been perceptible, even since the first ones went public last spring.
This is clearly just the start of something special — AIs are going to be doing a lot more than chatting and drawing pictures. They’re evolving fast too; when I started drafting this just before Christmas I wrote a sentence here speculating about the potential for integrating AI and search engines, and before I finished editing it, in early January, Microsoft announced they are adding ChatGPT functionality to Bing. Initially it’s not going to take us to completely new areas, AI is going to help people achieve their current goals more efficiently, but even that will be revolutionary. The things were roughly performing at the level of a primary school child just a year or two back; more recently, an average 12 year old; now more like a college student — there are already think pieces being written about the death of the college essay because AIs can now write a plausible one at undergraduate level, they’ll even provide the citations.
AI needs to have ways of interacting with the physical world
With Mattereum, AIs can buy/sell physical things in the real world
All these uses for AI have been in the digital realm, but for AI to achieve its full potential it needs to have ways of interacting with the physical world.
Sure, they could operate physical devices — waldoes, self-driving vehicles, robot bodies etc, but their reach could be far greater than those would allow, and Mattereum Asset Passports (MAP) on the blockchain can give them the API to make this possible, providing the bridge to the physical world they need to make things happen. This is how the blockchain becomes the world’s operating system. To buy, sell, track and move things AI needs to identify them and be sure they are what they claim to be. MAPs provide the trustable information about physical objects, connected to warranties that are legally enforceable in 170 countries that would enable an AI to be certain that they are really dealing with a particular physical asset. With the MAPs connected to a trustable NFT on the blockchain, the AI can use these NFTs to interact with the assets in the physical world safe in the knowledge that they are actually attached to something real that will be bought, sold, tracked or moved. You can imagine AIs instructed to only use MAPs to define objects, then searching the blockchain for the ones they need and working with those. MAPs enable AI to see and grasp the real world.
There are ways that Mattereum can use AI to be more effective as a company
That’s where Mattereum can contribute to the big picture in the future; in the shorter term there are ways that Mattereum can use AI to be more effective as a company too. It’s really too early to be able to see all the opportunities that may exist for AIs at Mattereum, but it is clear that they will make a profound impact. I can make a stab at where things are likely to go though, and for certain this is an area we’ll be paying close attention to as we go forward. There are six use cases I can come up with now, and this is just the beginning — I am sure we’ll identify more as the technology evolves and Mattereum grows.
1) NeoSwap
We have already taken steps to incorporate AI into what Mattereum does; our new partnership with NeoSwap is the first step in that direction. Neoswap is an AI-powered framework for web3 commerce that finds optimal multiway trades that redistribute assets and currency in a win-win-win manner. Combining their marketplace with NFTs backed by MAPs opens the way to a phase shift in the way items are traded that completely aligns with Mattereum’s work to bring physical assets on-chain. NeoSwap uses AI to figure out, of all the near-infinite number of possible trades, which trades are most likely to be beneficial to both parties and therefore accepted (among other optimisations). In the limit, if lots and lots of goods were in the Mattereum/NeoSwap ecosystem, it could produce a sort of “magical river” where you just give away what you don’t want any more and things which surprise and delight you just turn up in your life: minimal environmental impact, and near-zero cost!
2 ) AI Certifiers
I have, let us say, an AI certifier that does comic books, and my AI certifier is very, very simple; you show it the cover of the comic book, and it goes away and it comes back with comprehensive publication history for your comic — published on this day by the following people, these folks worked on the cover, these people worked on the story, this is the illustrator, the inker, this is how many were published, this is the price history and so on. I take all of the information out of the databases that I have using the AI and I sell it to you in the form of a warranty. Right? Very, very, very low risk. As long as my AI recognizer is good. I’m just giving you information that already exists.
3) Wear Analyzer
I give the AI a picture of a handbag and it comes back and says, “mint condition, almost new condition, slightly battered, really battered, distressed and finally “Oh my god, what have you done to that thing?”. Now, that’s a little trickier because this is a judgement call and you could have somebody that disagrees and then you’ve got to have some human coming in to give a judgement about whether the AI is doing a good enough job of this or not. So, categorical identification of a thing. The first activity is an assessment about a thing based on available data, now this second activity is very, very, very, very, very fiddly, but it’s the kind of fiddly the AI is actually pretty good at. If I show it a training dataset of 50 handbags, and tell it “these ones are less worn, these ones are very worn” the AI will be able to generalise what it’s looking at. But once in a while I will show it a handbag which is intentionally distressed, but is in mint condition and the AI will say that this handbag is horribly beaten and damaged, so it’s essentially worthless. At that point we will have a problem, but that is what the warranties are for. If it’s wrong, the buyer or seller claims against the warranty and gets their money back, and the AI will learn that this particular kind of handbag is mint when it looks like most other ones do when they’ve been run over by a truck and buried for three weeks.
These last two functions are both AI recognizers and are basically very much like the systems that will ultimately offer you a bid on car insurance, or health insurance. They’re looking at the available data, they’re making an estimate of what something is. Some of those estimates are nice and objective and clear and clean and easy, others are very, very, very subjective and there’s a strong risk of humans coming in overturning them. The next three opportunities are more sophisticated.
4) Clustering
This is the holistic formation of story through material objects. So I have an AI and it knows of my clothes and knows my size and when it sees something which is in my style and made in size hippo, then it just goes out there and automatically searches for things with the appropriate info in their MAPs and tells me that what I want exists and maybe even gives me multiple options for it because I’m not sure on exactly what I want. It can then even show me multiple things simultaneously — “Okay, hey, Vinay, how would you like a Penguin costume? I got a monocle. I got a top hat. I got a karate thing. I got a cane with a spike in the bottom of it and I have these really wicked white wingtip shoes all in your size and you already have the black pants and the waistcoat” and I go, “yeah, I’d quite like a Penguin costume — buy!”
Perhaps a more common use for it would be interior design; I say I want an Art Deco furniture set for my living room, I say “hey, here’s my room, get me Art Deco furniture that will fit”. It goes off, looks at the asset passports of furniture for sale and comes back with furniture set one, furniture set two, furniture set three at 350 grand, 75 grand and 175 grand and I can choose which one I like and it buys them for me. This can work in the short term too. You rent an Airbnb, unfurnished. The AI uses Mattereum asset passports to find 12 sets of potential furnishing options you can have installed for it. You pick one and it goes out and buys all that and fits them into the place. You stay for three months. At the end you leave and the AI resells all the furniture.
AI provides the ability to customise the experience of space
AI provides the ability to customise the experience of space, the AI knows what the style of furniture you choose looks like, it knows how big the place is, it knows what furniture is available, it understands what price is good, and what it does is it creates experiences for me. I know I’m getting actual Art Deco furniture and the price is good because it’s all warrantied in the MAPs so there’s no need for me to scrutinise the details, check out the individual seller’s reputations and all those kinds of things you’d have to do now. The combination of AI and MAPs makes the whole transaction fast, safe and trustworthy. Now imagine wedding planning; so you’ve just a big empty building and warehouses of furniture and people figure out what they want and it just finds stuff and arranges everything and off you go.
What AI is doing is synthesising multiple different streams of goods into a holistic whole, which has more value than the individual components. It could do this using the NeoSwap ecosystem now, just using seller info, but adding the MAP to the mix just makes everything more trustworthy and means it could also use regular auctions, it could use street purchases, anywhere something with a MAP is being sold.
5) Prediction
This is where the AI buys stuff just as you need it, sells it when you don’t by learning and anticipating your use patterns for various things, This might be the same thing multiple times, so it buys you a bike in the spring and sells it in the autumn every year; it knows you pressure wash the car on every third Sunday of the month, so buys you a washer on the Friday and sells it on the Monday, removing the need for you to have the thing cluttering the garage in the meantime. There’s a small cost differential perhaps, but it’s the convenience cost of always having what you need when you need it and not having piles of stuff hanging around until the day you finally need it. It uses the MAP to find and verify things, and by circulating things between people when they need them removes the need for everyone to have one of everything just hanging around for the day when we need it — this reduces waste and makes the spiral economy viable.
This would be really effective for travel. I’m going on holiday to Ibiza and in my hotel room, when I arrive, is everything I need, given that the AI knows what kind of person I am and I’ve got an express list of preferences for the things that need to be there. I buy my ticket weeks in advance. The AI organises it and comes back and says “I’ve been unable to find white platform shoes in your size. You’re going to have to bring those yourself. I got you a couple of special bonus items too”, and they’ll be things you like because it recognises your taste and can extrapolate. So the work of having to look through the space of all possible objects to select what I want is no longer necessary, and I won’t have to cart everything and fret about airline baggage allowances.
This is where AI has become a synthetic engine that can reshape physical reality
This is the mega payoff because this is where AI has become a synthetic engine that can reshape physical reality in the same way it filters and reshapes images. Today the super-rich are the only people who can do this, they say they want something and point a hose of money at it until it happens; how much is not important because they no longer think in terms of value for money; they work on the metric of want=get and money simply lubricates that process beneath their level of attention. With AI suddenly that becomes everyone’s experience because it does all that work in the middle for you, which billionaires have gophers to do. It makes it a whole lot cheaper too, and can work within the financial parameters you have.
6) Modelling
This works by the AI knowing all the new and second hand goods which are available in an area and using the information to identify a gap. In New York, there’s a roaring market for this object. The object is not currently for sale in London. The AI then proposes moving a bunch of this thing from New York to London to sell it here, or suggests you commission it to be manufactured in Bradford and sold. It can work out that if a whole bunch of fashionable teens in Nigeria like this thing, then there’s probably a bunch in London and Beijing who will too. AI can do market gap analysis based on patterns that it can perceive in culture, or based on a whole bunch of people saying they want to buy something that isn’t in the market yet. In that case I make a MAP for a thing that I wish existed, I publish it as specification and say “if this existed I would buy it”; get enough people who want that and it prompts manufacturers to make something that fits the MAP and sell it to the people who want it. Everything comes with MAPs so it is warrantied and can be tracked and this helps it to keep its value, so when the fad is over, it can still be sold on at a good price.
There’s going to be more, much more
Mattereum Provides a Marketplace for AI Services
So, this is just what I can see extrapolating from what we know about AI now; there’s going to be more, much more, but we’re going to have our hands full bringing all this to life. We’re just at the start; some AI stuff will no doubt be done within Mattereum itself, but most of it will be done by partners, like we are doing with NeoSwap.
What Mattereum provides structurally is a marketplace where companies that have AI technology can monetise it by attaching it into transactions as a certifier. Imagine a bunch of kids in college; they’ve come up with some fantastically sophisticated recogniser technology. I apply it inside of the Mattereum system and we use it for something like clothing sizing. You show me a bunch of pictures of a piece of clothing with a ruler somewhere in the shot and I’ll tell you exactly what size this thing is. Rather than having to measure a piece of clothing, I download the app then just slap a ruler on a table, take three or four pictures, I upload them, and it comes back with sizing information for the clothing, and it automatically generates size warranties for every garment that goes by. They could just run that as a business because it’s far more economically efficient for customers to just lay all the clothes out and take pictures of them rather than having to pick up the garment and measure every sizing dimension.
Mattereum provides a marketplace where companies that have AI technology can monetise it
The market provides the AI creators with a way of monetizing their product because they can now take a margin on the transactions of the goods to which they’re attaching AI generated information. Without that, they wouldn’t have a market because they would have to then generate an entire ecology around their AI to be able to get access to a transaction stream, but Mattereum provides them with a ready-made market where they can very, very rapidly monetize their AI work simply by plugging it into an existing transaction stream. They’ll need to raise some capital so that they’ve got a pool of money to back the AI with when it occasionally makes a mistake and they have to pay out on certification. They need to have the funds, but if they’re good, what they’re getting is a continuous revenue stream from de-risking other people’s transactions, and they don’t need very much capital because they will want to be super reliable.
There’s a whole fascinating AI ecosystem out there waiting to happen
There’s a whole fascinating AI ecosystem out there waiting to happen, and Mattereum has the potential to form its core by using the blockchain to drive innovation through real world asset NFTs that come with Mattereum Asset Passports. The future really is here, and we aim to be part of making it more evenly distributed.
*introductory summary created by GPT3 AI
If you are working on an AI project that you think could enhance or be enhanced by the Mattereum system get in touch with us via contact@mattereum.com
