podcasts

Bringing AI Lego to blockchains, Toby Simpson of Fetch.ai

podcasts

Bringing AI Lego to blockchains, Toby Simpson of Fetch.ai

August 2020

Posted by

Jamie Burke

CEO and Founder

As an early investor in Bitcoin and Ether Jamie went ‘all in’ during 2013 founding Outlier Ventures, Europe’s 1st venture fund and platform dedicated to blockchain and Web 3....read more

Toby Simpson is co-founder and COO of Fetch.ai an open decentralised machine learning network. We talk about his three decades of work in agent based systems: creating MMORGs (massively multiplayer online role playing games) in the 90’s to biologically inspired systems as Head of Software Design at DeepMind, which went on to be acquired by Google. And how these experiences led him to create the world’s first blockchain based decentralised AI and its use cases from DeFi to removing inefficiencies in complex mobility systems.Toby Simpson is co-founder and COO of Fetch.ai an open decentralised machine learning network. We talk about his three decades of work in agent based systems: creating MMORGs (massively multiplayer online role playing games) in the 90’s to biologically inspired systems as Head of Software Design at DeepMind, which went on to be acquired by Google. And how these experiences led him to create the world’s first blockchain based decentralised AI and its use cases from DeFi to removing inefficiencies in complex mobility systems.

Posted by Jamie Burke - August 2020

August 2020

Posted by

Jamie Burke

CEO and Founder

As an early investor in Bitcoin and Ether Jamie went ‘all in’ during 2013 founding Outlier Ventures, Europe’s 1st venture fund and platform dedicated to blockchain and Web 3....read more

Key Themes:

  • How blockchain convergences with AI
  • Why Money Lego needs AI Lego
  • Agent based systems
  • Why gaming is a great training environment for AI

Listen on iTunes

Jamie Burke
Welcome to the founders of web three series by ally ventures and me Your host Jamie Burke. Together we’re going to meet the entrepreneurs, that backers and the leading policy makers that are shaping web three. Together, we’re going to try to define what is web three, explore its nuances and understand the mission and purpose the drive its founders. If you enjoy what you hear, please do subscribe, rate and share your feedback to help us reach as many people as possible with important mission that is web three.

Today, I’m very happy to welcome on the show Toby Simpson, CEO and co founder of fetch dotnet. I welcome Toby.

Toby Simpson
Thank you very much. Great to be here.

Jamie Burke
So fetch.ai is an artificial intelligence lab. You’re building an open access tokenized decentralised Machine Learning Network, obviously, there’s a lot in that. So we’re gonna, we’re gonna have to unpack it for everybody. But the promise is that this allows any organisation or indeed individual to build or configure applications on top of a digital representation of the world. And these applications are effectively driven by what you now call software agents, I guess in the world of crypto and web three autonomous economic agents as as another way to call them and these software agents autonomously search, negotiate and transact to form automatic marketplaces. The reason why I’ve got you on the show, I’ve been working with you and Herman and the fetch team since God 2017 Is that possible?

Toby Simpson
Yes.

Jamie Burke
You were one of the first projects we incubated as a disclaimer Lies probably the biggest holder of fat, the token that drives effect protocol outside of fetch foundation and you guys, in fact, I believe I was even the one that came up with the name fetch Dotty I, you know, not not to kind of

Toby Simpson
not to pay too much credit, Jamie.

Jamie Burke
But that was about it to be honest with you. So, you know, there’s there’s a lot of amazing stuff that happened at facts that I have nothing to do with. But as some listeners will already know, I’ve been obsessed with the convergence of blockchain and AI for some time, and you were our first investment into that thesis. You are some of the first people AI people to come into the blockchain space, and to start thinking about how we can transition from pretty dumb blockchains with supposedly smart contracts, but they’re actually operating on very basic logic to something that could be used Great to scale and be harnessed by the world of machine learning. And for me, if you kind of follow that line of reasoning, the idea of an agent based web, where the DNA of web three in the sovereignty of the user versus the platform is kind of baked into that, this is going to be an incredible transformational driver foot for what web three is. So really looking forward to exploring some of these things with you.

Toby Simpson
Lovely, yes, indeed. And it was that that whole convergence that these individual technologies when you look at them, when you look at them individually, they’re still pretty amazing. But when you start combining them, then out of that you get something considerably more than the sum of its parts. And that’s certainly what excited us coming from an AI and an agent background when we looked at what it was that we could get out of blockchain, but also other cryptographic technologies that are beginning to mature and find their ways into this space to

Jamie Burke
exactly so let’s kind of contextually You as a founder as a co founder, and you’re kind of journey here. Normally I kind of rattle through this stuff to get into the meat of the interview. But you know, your background and the work that you’ve done is fascinating, but also very informative, I think to unpacking some of the big concepts around machine learning that will allow us to understand, you know, what you’re doing at facha, we’ll probably go spend a little bit longer than I would normally in this section. So back in 1991, you were a producer at preacher labs, where you designed a programmed a number of different games for things like Amiga, and the big game, there was creatures where you ended up creating a series creatures, two and three. And I think this is really informative to how you’ve come to to create facts and some of the principles that drives it. Could you explain a little bit about creatures?

Toby Simpson
Yeah, that was an interesting journey, actually. Because one of the great things about software is complexity kills And one of the things I’ve been saying recently, you know, when people ask the difference between 2g and 3g, 3g and 4g, etc, I would just say it’s several orders of magnitude more software. And it’s that additional complexity. And we’ve often said, you know, the more more plumbing you’ve got, the easier it is to block up the pipes. And I learned that firsthand in the early 90s, creating computer games on the Amiga assembly language, and totally drowning in in 100,000 lines of assembly language and been unable to manage all of that complexity. And I was introduced to the idea of approaching the problem from a different angle. So instead of writing a piece of code to solve a problem, to write, or put in place, all the building blocks that would allow the problem to largely solve itself. And when I got to creatures that was, I guess, the pinnacle of seeing that process in action, in that we modelled a whole bunch of biological building blocks and join them together to create an artificial organism and the creator of that Technology often said, you know, there’s no such thing as half an organism, which I would completely agree with. And from that perspective, he says you can’t just have a brain existing by itself, you need the supporting biochemistry to act as a feedback mechanism in order to ensure that it’s able to learn by itself. And you also need the inputs and outputs to connect you to the real world so that you can affect the world around you, but also be affected by that world. And I would still argue, even to this day, must be goodness, 25 years down the line, that creatures is still the most advanced example of an artificial organism. What interests me about this, of course, is that we created these general purpose problem solvers that were able to solve a problem that they’d not been exposed to before. And I thought that was really interesting because you didn’t write an enormous amount of code. Suddenly, we wrote a large population of simple things, and they they work together to solve bigger, grander problems. And that was really interesting. I went on to explore the usage of that kind of technology for virtual people in massively multiplayer online games because you don’t want the barman to offer you another drink when the bars burning down. But you don’t want to have to define a rule that says if the bars burning down, don’t serve the drinks, because that doesn’t scale particularly when you’ve got lots of human beings involved. And actually, this kind of approach of drive driven biochemistry backed artificial organisms is incredibly effective at dealing with these things in a plausible lifelike way. So that was effectively the introduction to that agent based approach. And you can see how that maps to the modern digital economy that we’re building with blockchain type things you know, you’ve got your Oracle’s, which are conduits between the real world or the outside world and inside the blockchain space. You’ve got your smart contracts, you’ve got your little units of code that are acting a small space building blocks. But the one thing that you’ve got now, which we didn’t have back then is this ability to scale it in definitely. And that was the thing that excited me most about blockchain when I suddenly realised, well, goodness gracious me, having spent the best part of a decade trying to figure out how to build the matrix in my living room, I’m suddenly sitting on a piece of technology. That means I can build a virtual world as large as I want. And there is no theoretical limit to how big that can get. And that as we walked into the whole space with fetch was one of the things that was just so incredibly exciting that we could create the scale. And when you can create that scale in a modern digital economy, with billions of moving parts, suddenly you can begin to see plausibly how you would conduct that orchestra in order to create some really fantastic music.

Jamie Burke
Yeah, and we’re going to come back to I guess, how you arrived at the combination of blockchain and AI a little bit later. So from that creature lamps that was 1999 to 2000. He then went on to nice tech limited for another 11 years, and we worked on something called Alice server. Similarly, this was around biologically inspired agent orientated, massively multi user simulation engines. And it was all this kind of work that you were doing that led you to be kind of picked chosen by the mind. I think you were employee number one, two or three. But you joined there as head of software design, of course, deep mind went on to be bought by Google to become effectively their primary AI initiative. And but it was specifically the work that you’ve been doing in these biologically inspired learning systems. That was the reason why you were brought in deep mind. Can you explain why specifically that background was was of interest to DeepMind? And its founder?

Toby Simpson
Yeah, well, then I’ve known demos for quite some time, actually. I think it was in the mid 2000s, I went down and gave him a demo of Alice server. One of the demos we were showing people at the time was this thing I lovingly called world in a box. And you started with one agent. And then there was 248 16. And then suddenly, you had a whole world with trees, plants and shrubs appearing out of nowhere. And the most amazing thing about that is unlike most games at the time, none of it was scenery, all of it was real. So you could in fact, in theory, reverse a Jeep into a tree, knock it over, and then build a log cabin. And no part of the code would have to know anything about log cabins or trees for that to be possible, which meant that the worlds could be interacted with in the way human beings do. And that is that humans are unpredictable, you go into a space, and we think of novel new ways of interacting with it. And games are very restrictive. And you’re often reminded of why you can’t do something or the fact that that’s not a real rock. It’s actually just a piece of scenery. It says a plot device or not a block device or just something an artist creates And the advantage without crutches if you didn’t like the forest you were looking at, you could just switch it off and create another one in a few seconds. And that ability to create endless infinite environments I thought was really interesting. And it matched very well to a lot of the stuff that Dennis had been doing in the late 90s and early 2000s. And that sort of approach of taking a biological perspective on things and building things that would grow and breed and pass on things that define what they do and how they look from generation to generation that he thought was a really interesting, additional approach to have involved in that journey that they were going to be taking. So it felt like an obvious choice to get involved with that. And it was a huge privilege to be part of that that journey, particularly so close to the beginning of it,

Jamie Burke
and what were the domains at DeepMind that this approach was being applied to so obviously, you know, you developed a lot of this in the context of of gaming naturally, I guess, because it’s Probably the, you’re having these large scale environments, you probably gave you something that you couldn’t have in the real world at that point. But what kind of domains were you applying this technology to at DeepMind at that point?

Toby Simpson
Well, it’s general purpose problem solving the idea that given a puzzle that you haven’t been exposed to, an organism can, or an agent can learn how to make sense of it and survive in that environment without needing a bunch of rules to do so. And this, incidentally, is one of the reasons of course, why computer game environments are so popular when it comes to building and training modern AI, because they are rich, interesting environments to interact with. So you often see people playing or building API’s or AI units to play, say quake or to play real time strategy games because they are rich, interesting, diverse environments that you need to have an understanding of, in order to do something sensible in so gaming environments are great places to do this stuff. But this general purpose approach This approach that lends itself a little bit more to biology is a really good way of creating things that can learn by themselves with without an external input, that ability to learn without supervision, given a brand new environment and a brand new set of problems is a valuable part of the stepping stones that everybody is trying to take towards a general purpose intelligence a true one an AGI.

Jamie Burke
Right. So you left DeepMind 2013 you went to also sim limited, doing similar things. General Purpose simulation engine again, and then you moved on to a found fetch with him iron. So you met him I and prior to deep mind, I believe you’d kind of been in each other’s orbit for well over a decade. And he was also one of the first angel investors into deep mind. But how did you guys meet one another?

Toby Simpson
You know, I can’t exactly remember precisely how we met. I think there was there was another project that he was working on that funnily enough involves dams. And we met around mountain that sort of time, I’d say a good 10 to 15 years ago, and shared some of the same thoughts about the the problems with with complexity of software. And I was looking at this from from a gaming perspective, and he was looking at it from other perspectives. And we’re sort of thinking, well, at what point are we going to arrive somewhere where all of the technological building blocks are present to allow us to do things of this enormous scale that we want to do? And at that time, it sort of seems quite quite, quite impossible to be able to do that.

Jamie Burke
So I met you in 2017. I met you both. I remember the meeting. Clearly. It was a mutual contact of ours and Melissa, and it was a perfect kind of point of serendipity in that you guys were working on what I’d been envisaging was was this convergence of blockchain and AI, and at that point, actually, you were trying to solve a very particular problem. If I recall, it was around kind of traffic management systems for drones, but in a in a decentralised way.

Toby Simpson
Yeah. Now that was quite interesting, actually, because that was that was quite, that was quite some journey. Because we were we were looking at at how you could effectively treat individual drones as autonomous economic units in their own right. And how you could create a system that would effectively manage and plan and deal with a large a large fleet without the need to have a single point of failure or a single centralised control unit. So all of these things that effectively argue amongst themselves to get things done. And what became more interesting, then as we started looking at the individual drones and thinking, what actually, if these things are carrying a bunch of cameras and a bunch of other sensors, why aren’t those economic units in their own right? Why can’t they negotiate with the with the flight hardware to say, well, could you just take it A small diversion. So I can take a picture of this or sense that or do this and do that. And suddenly, there’d be all this additional utility value that would come out of a single unit that might be just flying somewhere to take a few pictures, and then a whole bunch of other components on the drone would then negotiate in order to get even more value out of that. And that would start uncovering what it is that we we tend to waste on a moment to moment basis in the economy right now, simply because there’s so much of it, that no centralised point of management can possibly organise at all. And it was that sort of journey as you go further and further down through the drones. And then you start thinking, Oh, well actually, there’s an enormous amount of additional knowledge that you can get. So you’ve got three or four drones flying around and they’ve got sensors for mobile signal strength or whatever, then you’ve got AI you can apply on top of that. So now you’ve got this learning going on and you’ve got another agent on top of the of that picking up all of this learning and deliver that value. And it’s sort of spread upwards and downwards from from from the drone into this grander agent vision, which, of course, is how we ended up at fetch the idea that it seemed wasteful to apply this extraordinary combination of technologies to just that, when in fact, not only could it do that, but you could attach autonomous economic agents to almost everything, to every vehicle that’s going around to pieces of data that have been generated IoT devices to people to pretty much anything that you could think of. And if you could come up with the scale that you would need to be able to connect them all in effectively in the same environment and the learning and AI you would need to connect them effectively, then you really looking at something interesting.

Jamie Burke
And so you know, this journey or mission around a decentralised form of machine learning, really evolved. From there, you have to kind of pretty much build down the stack to solve for a problem initially that you you’re starting to do the application layer almost and realise That at that point, a lot of the infrastructure was missing. I think before we go into that, it would be good to just get a high level understanding or definition of collective learnings. Obviously, collective learning is is the big concept behind fache. But it’s perhaps, you know, quite elusive for the average person to understand. So could you just give us a bit of a primer on collective learning?

Toby Simpson
Yeah, actually, collective learning is just one part, of course, to the whole fetch AI thing. But it is a very interesting one as you point out, because we live in a world where people are collecting an awfully large amount of data, and some of it for a variety of different reasons can’t be shared. So in, for example, in healthcare, you’ve got a whole bunch of data that can be shared because it’s confidential. And because the paperwork required in order to be able to share that is nightmarishly complex, and varies dramatically, depending on what the data is and where you are in the world. But then there’s also data that’s confidential, and you’re sitting there thinking, well, it’ll be useful. Lovely if we could generate a machine learning model from our data and everybody else’s data, but we don’t want to give our data to them, and they won’t want to give their data at us. So how are we going to do this? collective learning is the solution to that. they’d see the ability for a large number of people to contribute in a decentralised way to an overall learning model. And for that to benefit everybody. So you know, there’s a couple of examples of that, of course, one of them is, say, for example, particularly modern modern vehicles, modern modern modern cars are collecting a huge amount of data and sending them back to HQ about the performance of the vehicle. Ideally, that data is then used to figure out models to predict when vehicles might fail. So you can advise someone to go in for a service or to have something changed or altered before it becomes expensive and complex. Now, wouldn’t it be great if one car manufacturer could combine their data or somebody else’s in order to end up with a better model have benefited them both but with neither having to give away They’re confidential proprietary data. That’s where collective learning comes in, in healthcare, where you’ve got a whole bunch of different things that might, for example, allow you to build a better model for detecting or diagnosing potential conditions before you would otherwise be able to do so you’re sitting on a huge amount of confidential patient data. collective learning means that you can all contribute to a model, the model gets better and nobody else has to have to give up that confidential data. So that’s what makes that very exciting. And what’s really exciting about it has been able to do it in a truly decentralised way. Because you absolutely want to avoid central points of failure and central points of control, where any of that or combined with any of that data actually ends up resting.

Jamie Burke
And that’s obviously very important. If you look at the overall AI universe, which is generally proprietary and dominated by a handful of platforms. You know, Google being the most dominant and why We’re seeing initiatives like open AI, almost to kind of counter that. But still, this is a technology that is highly permissioned. And so currently with open AI, if you wanted to use the latest instance, GPT, three, you have to get permission in order to use it. And so the thing about fetch is this permissionless part. So in the intro, when I was talking about the idea that this was open access, tokenized decentralised machine learning, the open access part is is really key to this as well, isn’t it?

Toby Simpson
Yeah, very much so. And certainly in the case of collective loan, you could see how that would work, you know, in a decentralised way with no one entity been in control. And that’s really good because every individual is in control of how that works. But of course, also there’s there are other aspects to this, this sort of AI Lego approach that you can build autonomous economic agents that solve particular problems, or that buy up low value data and transform them to something interesting, and then you’re effectively providing a network where all of these things can be connected to each other in an effective way. So if you want something, there is a way of describing it so that you can find it. And and that overall search mechanism is another key part of this. And of course, ai exists on the outside from an applications perspective. But of course, also it exists on the inside, from from a technology perspective, because if you’re trying to find something and you don’t know how to describe it precisely, how on earth can you do that? Now turns out of course, there are some some very well understood techniques for doing that in machine learning, but allow you through dimensional reduction to effectively position things so that things that are likely to be similar are close to each other in a reduced dimensional space. And that’s really interesting because you don’t even have to be entirely accurate and then you can slip in a circle around a certain radius and capture everything that you might be interested in. When you’re effectively turning 10s of billions into just a few hundred with with something like that, that’s a very, very effective method of searching in a context sensitive way.

Jamie Burke
So why, why blockchain I think we by now understand the importance of having a decentralised form of machine learning or AI. But specifically, how does blockchain solve for that? And your Why? Why does it require a token?

Toby Simpson
Yeah. Now this is a this is a fun one, isn’t it? Because this is one of those ones, which when when we first met Jamie, I was still not I guess, 100% convinced of how all of this stuff works. I tell you what, I wish I’d met you a few years earlier. And got a bit of a head start on with this blockchain. A Well, I mean, I did it I did a talk a couple of years ago. I think it was now going as Time flies called the first role of blockchain Club, which I described as being completely unable to explain blockchain to anybody else. And I think that this is an area It’s where we don’t do that very well. We talk about blockchain, potentially just in the field of Bitcoin without really understanding what it is that the blockchain is doing. And that ability to maintain a, an overall ledger amongst a great number of parties. Without it been economically practical, in any way, shape, or form to modify the past is really the key thing about that, because what that really means is you’ve got integrity of a data structure, with no one person but in charge, and you can keep adding more people to the processing of that data structure and individuals can come and go, and yet, anybody can self service establish what the truth is, by looking at that ledger and working backwards. Now, you may think, Well, yeah, surely that’s just a that’s just a banking on a financial thing. But actually, it means that you can hold any large scale data structure together so you get scale. And when we were thinking about constructing an environment in which autonomous economic agents could begin The idea that you could just simply keep adding nodes to that network and get bigger and bigger scale by adding the capacity to add more and more agents to it. And to potentially allow those individual nodes on the network to specialise, either by location or semantically or both meant that suddenly we were able to see well, there is no limit anymore, there is no hard obvious limit to how many of these agents we can attach to this network. Now, the token of course, is the fundamental incentive mechanism that makes it more more costly to be dishonest and honest. And that was of course, why why Bitcoin actually worked. That there there was this incentive for people to run the numbers in order to be able to do all of this. And there were there were those rewards for for for doing so. So they are intimately tied with each other the the underlying token and and the network and the ability to actually Hold it all together. But of course, then you’ve got that that native token, you’ve now got a method of value exchange between all of the autonomous economic agents in the case of fetch. And you can do it practically because, and we often saw right at the very beginning as well, you know, who’s gonna pay very much for the temperature out of somebody’s back garden down a street, you know, these are potentially very low value things. If you’re paying a 10th or less cent for a piece of data like that, you know, the transaction costs have to be way less than that, for it to be practical. And when we looked at the different combinations of blockchain technology, and Bitcoin is not exactly a great example, a normal theory and when it comes to transaction costs, but there are other solutions and other methods of handling consensus where you get that high transaction throughput and you get those low transaction costs. And suddenly you can imagine these millions if not billions, of autonomous economic agents actually getting work done and been able to do These these micro transactions is all of this data and information moves around the network. So Bitcoin was the enabling piece, we already had everything else. We understood how agent based environments work. We we’d seen the emergent behaviour that you could get. And we understood a number of the AI components that we would use for this and the learning components. And we knew that small units of intelligence could combine to produce grander insights into all of this, but blockchain was that missing bit that has title together.

Jamie Burke
So in effect, now, you’ve been able to introduce, I guess, an economic layer or economic agency into agent based systems in a decentralised way. And the flip side is, you’ve also introduced an intelligence layer to the world of blockchain. And so I really like that idea of AI Lego, and how that might speak to this idea of automation. Lego, which is obviously kind of another way of talking about defy, which has exploded on top of aetherium. So, you know, can you just talk a little bit about that AI Lego concept? And I guess how how that comes into existence across multiple chains in a cross chain? universe?

Toby Simpson
Yeah. And that interoperability is actually very, very important. Because people often say in this space, and we see it a lot, and I see it, I guess, well, we’ll see it a lot less as time goes on. And people say, well, surely a theorem is the solution to everything or surely Bitcoin is a solution to everything or surely does that awareables is the solution to everything, but actually, none of that’s true. They all are different and they solve different problems. And the idea that you can effectively connect the unique capabilities of all of these different networks together in a way that’s that’s meaningful is actually really fascinating and incredibly exciting. And it also means we can unlock a lot more have the potential value by being able to use these these unique features. And this is one of the things that we realised very early on when we were talking about our fetch that actually there’s a whole class of agents that act as interfaces between both between the blockchain world and the real world effectively as Oracle’s and and and also those that allow you to communicate between an interact between all of these different chains and they’re effectively service agents. And more for that matter, that mean that the the unique capabilities of these things can be can be packaged up and attached to autonomous economic agents. And that means that there’s additional value and additional ways of exploring. And right when when we started talking about this, one of the things that came out of my work back in the data server back in 2000s, was that an agent based massively multiplayer online game engine, and the idea was that an agent could meaningfully interact with another agent in even though have never seen it before. And we have the same thing here, we’re fetch that. We want agents. In fact, we’ve got agents that are meaningfully able to interact with each other with no prior exposure. And that means that you, you wipe away in one go, all of that horrific complexity of having agents having to know about all of the details of the other agents that they’re going to interact with. And it’s that kind of thing that has allowed us to build some of the really cool stuff that we built recently. In for example, in mobility.

Jamie Burke
So in version two update, which happened quite recently, you have now enabled this interoperability with Cosmos hub, and then with aetherium via this Cosmos IBC bridge. So what what is that gonna make possible on Cosmos or a theorem that isn’t possible today?

Toby Simpson
What I’ve actually think it comes down to what what you were just talking about with with AI Lego ponies it’s more than that. I guess it’s is blockchain capability Lego, if suddenly means that all of these systems can effectively be connected as one continuous world, and that autonomous economic agents in the fetch space are reasonably able to make use of all of the features and communicate with the things that they need to aim in these other spaces. So for example, with the decentralised delivery network that we’ve got, we’ve we’ve built that the moment that the the way that works is it uses a theory on contracts, but their fetch agents existing on a flat space. And the ability to get a combine those two things is absolutely fantastic. And the same applies across the space with with all of these other networks and capabilities too. But suddenly, you’ve got all these additional pieces. And whether that’s d fi, whether that’s AI or whatever. Now you’re looking at been able to assemble all of these bits and pieces into into one space and then then find them effectively.

Jamie Burke
So obviously defined Hot now and you know most projects are claiming to either be a D five project or to be highly relevant. But, you know, clearly fache is very applicable to the world of defy in that it will allow for greater levels of complexity in terms of the kind of financial products or instruments that are going to exist in in the defy world. So, you guys have a couple of projects, metal x, which is a decentralised commodities market and atomics, which is allowing for atomic swaps. I know that, who am I and has led both of those initiatives for those that aren’t aware? Who am I and also has a very deep background in the commodities world, but could you could you talk through those two examples of how fetch is being deployed? I guess it’s in the defy context. When you could argue it’s actually bridging between decentralised finance and centralised finance. Right?

Toby Simpson
To a certain extent. Yeah. So and I mean, as you touched on, there’s a lot of people who are using the word defy right now and to a lesser extent, it kind of reminds me of the late 90s when people were slapping comm on the end of anything they wanted to and trying to be seen as an internet business. There’s there’s there’s a lot of that misinformation out there about what is and what isn’t potentially a key component part of constructing decentralised finance. But I think the one thing that everybody realises and knows and understands that actually this technology is able to build this stuff. And then when you’re able to combine all of these different components in interesting ways, then the kind of range of financial instruments that you can you can construct is suddenly a lot broader than it would otherwise be. The decentralised commodities exchange of metal x is actually really interesting because it’s designed to make things funny life a lot simpler to get rid of some of the complexity of hedging derivatives, which people find so complex and provide a method of getting some degree of protection against themselves. price moves. And there’s a whole lot of unique things that are in there. So they’ve got a dual token system for short tokens and long tokens. And we’ve been very pleased with what it is that we’ve been able to build with that and and the fact that that it is these technologies or technologies relating to fetch and technologies relating to modern decentralised bits and pieces generally, in cryptography, generally, that’s actually allowed us to do it. And it means that the costs are a lot lower. And it’s a lot easier for people to get involved. And some of the barriers to entry that were previously exist, aren’t there. And things like smart contracts and the generations of smart contracts to come. And we’re, we’re, we’re sort of touching the beginning of this, this right now. And, and that was one of the things that a theory in particular first gave us all a glimpse of this, these programmatic, smart contracts will be able to make these decisions independently of human intervention. And that would make these kinds of things transfers and applications very, very, very practical for the first time, but you know, you’ve still got costs involved in doing that, and something that we’ve seen recently with theory and costs. And some of the newer generations of technologies are able or built specifically to enable this kind of stuff. Now, certainly one of the things that we’ve been doing with fetch,

Jamie Burke
so obviously, I it’s natural that some of the first use cases that would happen on fetch would be defy related. But you know, the network, as I understand it now has over 140,000 different agents, active agents on it today. And you know, defy is just a subset of those. I know that your you alluded to this distribution delivery network. I know mobility is a big focus area for you. It has been over the last several years. You’ve had a number of different use cases trains, cars, parking, could you could you talk about why this is so relevant to mobility and bring to life perhaps how some how agent being fat agents have been used in in mobility systems.

Toby Simpson
Oh yeah. In fact, any any problem with a very, very, very large number of moving parts as a problem autonomous economic agents can probably help you with. And that includes things like supply chains, but mobility is the big one, because mobility is the one that affects us all on a day to day basis, and it’s a huge problem area as well. So feel free to the reason why we’ve been building this decentralised livery network, of course, is because the idea is that with the cryptographic technologies that we have now, there is no reason why you shouldn’t, in a decentralised way, be able to connect somebody who wants to get in a car and go somewhere to somebody who has that car. And that is effectively a decentralised, ride hailing network that can be built without any centralised entity on top and yet can still exist in a trusted environment. So things like verifiable credentials, and other methods of proving information that’s relevant. So As proof of location, smart contracts to govern how all of this works and deal with dispute resolution, a blockchain to provide history, trust and reputation. Depending on of course, what it is that you put on there to make that possible AI to analyse it and the ability to be able to conduct something a little bit grander. Because it’s sometimes it’s not necessarily that you want to deliver yourself home from from a bar, night. Sometimes you want to deliver yourself to another country for a four day conference and get back and conducting all of those component parts is considerably more complex. So we built this it works. So we have the we’ve done a couple of demos of this recently showing how the decentralised delivery network can act, allow people to be picked up and dropped off where they want to do and we’re gradually scaling that and building it into something grander for larger scale deployment. But also we’ve been looking at autonomous economic agents lighting up the world in A more meaningful way. So self driving cars is a big thing right now but actually getting to SAE level four or five for autonomous driving is a little bit much for current AI to do. Whereas if you effectively light up the world, and by bringing say, for example, individual signposts, junctions and whatnots, to life as agents, then you create this augmented reality where the load on traditional or current AI, for analysing the world around it is reduced because you’re providing more information. And that means they’ve been able to navigate around that 3d space becomes considerably easier as a start up. And that’s another one that we’ve actually built. We’ve got some great demos of that coming up showing how that influences our ability to make self true self driving cars autonomous driving actually work. And we built another system this year. relating to the UK railway network where we deployed individual agents for every single station and train live. And of course that can be deployed. There are other API’s that exist across the world. And suddenly, you’ve got this picture of this world, where you’ve got agents representing passengers representing taxis representing signs, junctions, representing stations, trains, and you’ve got this huge, effective mobility related population of agents, which of course means if you stand in that space anywhere and do a search, you can come up with all the things that are relevant to you. And from that you can then plan according to your personal preferences, but without leaking your personal details, which is really exciting. So you can suddenly get the best route to get from where you are to where you want to be, and have it change in real time and have the information that’s relevant to deliver to you, without you having to pay attention on a moment to moment basis in order to To ensure that you don’t drop something, and it’s that turning everything inside out that that is key to this, we’re trying to change this or we are changing this so that instead of you as a human being having to chain, check five different apps, and pay attention to all of this, this data, the data that’s relevant to you is delivered to you, including stuff that you never might have thought was there. And it’s that additional bit that makes it super interesting as well.

Jamie Burke
So, I mean, this is one of those things. And this is this is why I’ve always found fetch such a fascinating project is when you imagine a world where you have massive deployment of these trustworthy agent based systems. And you have an agent associated to you know, your person or your devices. The kind of the imagination runs wild like what’s so let’s assume that happened. Let’s assume fetches agents are now everywhere. These bottom markets are forming between agents. What fundamentally changes that what the human experience? How does that change? Because I could imagine that why would you need a search engine anymore? Why would you visit websites anymore? Why would you have to navigate around adverts? It becomes an entirely different web.

Toby Simpson
Yeah, good question. Why indeed. And and that’s, that’s one of the side effects of turning it inside out. The things that are relevant to you come to you and for you to have to wade through pages and pages and try and figure out what keywords to get the things that you want. That potentially all Gertz It’s about taking the complexity out of your life and I think we forget sometimes just how many of these silly little hoops that we have to jump through and you know, sigh going for those who still goes shopping, you know, you walk around the supermarket and you put everything in a trolley and you get to the till you take it all out, and then you put it all back in again. And it’s those kind of things as utterly insane. The little things that we do that are inconvenient. And I think we forget that when we travel anywhere, just how stressful it is of just how much we have to do to ensure that it works. And if it stops working for any reason, again, just how much of the responsibility for fixing it actually falls on our shoulders. And when you bring all of this to life in with with autonomous economic agents, you’ve got an agent that represents you, it knows what what you want, it knows your preferences. It’s out there in that space, doing concept, context sensitive searches, finding the other agents that can solve the problems for you in real time. And it’s far more human way of working. Because generally, we’re quite unpredictable in these things and we change things at a moment’s notice. And then some of the planning that we have to do, takes away that flexibility and this puts it all back And, and that’s certainly one aspect of, of how it changes our lives in a meaningful way. It makes them easier and, and solves problems actively rather than, than us having to pay attention to it. So hopefully,

Jamie Burke
by now, listeners will have an appreciation for how something like fetch.ai completes or even expands web three, in terms of what’s going to be possible with these technologies that are kind of centred on on users rather than platforms. And I just want to also acknowledge, you know, you guys, as a team, have been a pleasure to work with, but you’ve continued to ship, you know, despite the winter despite what happens in the secondary markets, and I think now the world is really starting to appreciate just what fache represent. So I just want to say well done to the team for, for writing that story. As well,

Toby Simpson
no, thank you very much. And you’re right about shipping. And this is this is the important thing, isn’t it. And we’ve got this grand version. And we know that this fits in with the economy at the moment. And we know that it scales in the future. And a lot about the future is going to be getting more out of what we’ve got, making better use out of the assets that we have not throwing away so much data, not shipping things around in largely empty containers. And generally reducing the pain of those involved in getting anything, whether it’s a person a package, a piece of data, or anything from one place to another. And that’s all about discoverability. And that’s all about making it painfully easy to do this. And that’s what we’re doing at fetch, you can build these agents to represent the assets that you already have, you can pop them into this world, at virtually no cost to yourselves and then they can be discovered as part of other people’s solutions and combined in new and interesting ways. And that’s what is really exciting. I think about this new world.

Jamie Burke
Well, there can’t be a bigger purpose than In this imperative of removing inefficiencies in the world and, and clearly that’s kind of the, the big intent behind fetch. So, Toby, look, it’s been a real pleasure to have you on. I’m looking forward to getting a glass or two of wine in with you at some point when the world becomes a bit more sensible.

Toby Simpson
I look forward to that too. Thank you very much.

Jamie Burke
If you enjoyed today’s podcast, please make sure you subscribe, rate and share your feedback to help us reach as many people as possible with the important mission of web three.