Fetch AI — Cryptoverse AMA Recap
20th Nov 2019
Guest: Toby Simpson — CTO
Q:Please briefly introduce yourself and #Fetch ai to our community?
A: I’m Toby Simpson, co-founder of Fetch.ai. I come from a computer games background originally, doing a lot of work with advanced biologically inspired AI before developing agent-based MMOG engines that could scale to hundreds of thousands of objects with full dynamics. I’d hoped to create the Matrix in my living room
I went on to be Head of Software Design at DeepMind as well as building a bunch of simulations to help with corporate training. I’ve always been a believer in technology that could make truly massive worlds, and populate them with vast numbers of autonomous entities.
Blockchain technology was key to this, as we realized that now, we could create a world of *any size we chose to*, regardless of whether we trusted any individual computer involved. This led us to the idea that we could break the economy down into its individual pieces, bring them to life, and create an environment that learnt how to connect them together so that they could get useful economic work done: an economic Internet, where AI and ML created a dynamic world that adapted in real time.
For us, this was the ultimate dating agency for value providers: we could connect something that wanted something to something that had it, and do it in real time, with no prior knowledge.
In tasks with lots of moving parts (transportation, supply chains, energy, etc.) this is potentially transformational: it makes better use of the excess in the economy, increases utilization and makes better use of what we have.
Q:What are the current technological limitations in intelligent data sharing, machine learning, and AI, and how does Fetch AI plan to address and overcome these limitations?
A:that is a broad question.
There are two sides to any limitations: the power we have to put towards things, and our ability to use technology right. So just throwing power at something doesn’t necessarily solve all the problems, if we are speeding up something that is already fundamentally broken.
Deep Learning, for example, will *never* lead to AGI. Not even close.
Most of what we see as AI and ML out there is really just statistical analysis of large data sets, and people can get some extra coolness out by throwing more data and more computing power at it, but that isn’t anything new, it’s just putting bigger engines on existing feed-forward neural networks, etc.
Having said that, ML and other parts of AI, when used right, can be incredible. We use them in Fetch to intelligently connect autonomous economic agents together. We use it to learn more about who connects to who, when, and why, and to provide trust and reputation information.
Also we *enable* decentralised AI with our amazing synergetic contracts: you can do real ML *on chain* in smart contracts. This is really cool when it comes to solving large-scale optimisation problems, such as route planning, network optimisation, and more, as everyone can contribute parts of the solution to a grander problem and be rewarded. It’s also important for disintermediating the economy — we provide the tools, and the platform, to do things that would not have been possible before.I could talk about this one thing for hours, but there are other questions to answer, so we’d better move on (but good question! we use AI, we enable AI, and we enable it in new ways)
Q:Does FET have plans to implement privacy in Fetch? Does IoT currently require a lot of hardware
privacy and does applying AI — Agents to control the device cause the device to be deactivated?
But as you know, privacy is a broad church and means different things to different users of the network. Not all of what one might put under the privacy umbrella is part of the Fetch architecture, we also know that we will interperate with other services and technologies to help.
We live now in a post-GDPR world (the EU data privacy rules) where individuals have more control over their data. And they should do, too. Part of what blockchain does is to move power and control from the centre of the network to individuals on the outside and this is a GOOD THING. So we have had to build and architect on this basis.
Some miscellaneous points:
1) Agents do not put their data on the Fetch network, they put special representations of what they have or want.
2) When they transact, that is encrypted
3) There are many ways for agents to protect the details of what they do
4) Agents attach to the network, what they do and how they do it is up to them (as it should be)
I would just add that accountability is a key part of securing IoT. IoT and pretty much all connected devices are a security nightmare. If you plug your fridge, toaster or TV into the Internet, go home and unplug them immediately.
There is no incentive for those devices to be configured right, or secured, or updated, or maintained in any way and they won’t be.
One thing we are doing is putting in place an architecture where devices that DON’T do these basic things are effectively removed from view from other agents in the environment, putting in place an incentive to update and secure.
(also, of course, there is privacy preserving computing, zk-SNARKS, homomorphic encryption and lots of other wonderous technologies which Fetch supports to increase the security and privacy of the network as well as creating opportunities to perform work on data or computing activities on code without being exposed to the data or the code. Really cool in decentralised environments!)
Q:1. what bought the concept of http://fetch.ai and where did you realize that this is something which can change the world. What changes can Fetch bring to the real world. What problems will fetch be solving in near future?
A:1. The world is big and complex, and it’s getting worse. Our utilisation of its assets, particularly data, is woefully bad. We wanted to turn it inside out, to make dumb assets smart, and to allow them to autonomously deliver their value or have the things they want arrive at their doorstep without having to think about it. Fetch enables this: and that brings more of our data assets and hardware into play, because you no longer have to know it’s there to be able to use it, it’ll come to you!
2. As well as utilisation, and better solutions, it’s about disintermediating the economy: connecting value providers to those that have that value without large centralised entities that take a large slice of the cake for simply connecting two things together and providing trust. With modern crypto technologies, this can be done without the need for such entities. It means businesses can be combined, in new, and interesting ways, and more of what you have is owned by you. We’re out to roll the red carpet out to everyone: to make a complex life easier.
(and as I’m fond of saying, Douglas Adams once said that people don’t want technology, they want to get things done. And if you have a technology that gets things done, then you’re in business. That’s what Fetch is about. Getting useful things done, immediately, and without you having to juggle 100 flaming chainsaws to do it)
Q:First of all congratulations, As we heard that Fetch Mainnet is coming in the next month, What new feature we will get in that edition?
Does Fetch achieved 30K TPS or it is just a theory? How much Node does Fetch individually have for securing fast transaction?
A:Thank-you. We’re amazingly proud of what we have been able to deliver, and that we have kept to our schedule this year. We’ve created an enormous amount of code, all in lovingly crafted C++, all there for people to see and use! Mainnet is a refinement of where we are with beta: everything is tuned, improved, more secure, more stable and ready to go. We will have one more interim release in just over a week and after that it’s mainnet. And then we turn our attention to next year: building on what we have in consensus, ai-powered smart contracts and decentralised search, as well as delivering and supporting new applications.
So let’s talk 30K TPS.
Yes, we have the technology that can deliver that in a real-world environment. Will that all be in place for the first day of mainnet next month? No. But we have already provided the code, papers and benchmarking that allows anyone out there to verify that what we say is the case. And it’s important, too. We’re NOT just playing bingo here — we need this TPS for all these agents to work and get stuff done.
Q:Collaboration with famous and reputable businesses will create opportunities for Fetch.AI to expand its application. So, with which partners did Fetch work? And who are potential partners in the future?
A:Agreed! And we’ve done tons this year; for example —
Fetch.ai partners with Turkish steelmakers to develop first AI-powered decentralized metals…
A group of leading Turkish steel makers and traders have partnered with Fetch.ai to develop the world’s first tokenized…
Fetch.ai and Grey Swan Digital join forces to reinvent the $500 trillion global derivatives market
The partnership between Fetch.ai and Grey Swan Digital will enable anyone to trade price, make markets, and offer…
and our other work with t-labs and many others. And much more to come!
Q:Lots of people in the community are talking about staking. Do you have an update for us?
Will have a lot of competitions from other blockchain in no long a time, what measures are you putting in place to couple with them?
A:Staking’s out there and working. Go to
The Fetch.ai blockchain uses Proof-of-Stake with Unpermissioned Delegation (PoS-uD) consensus. By staking FET tokens…
We’re on our second auction now, demonstrating the underlying model for our unique proof-of-stake with unpermissioned delegation. This is great because it removes some of the problems that occur with delegated proof of stake (which tends to descend into a cartel type scenario which is good for nobody).
The Future of Consensus: Proof-of-Stake with Unpermissioned Delegation
We outline how our innovative consensus protocol, underpinned by sound economic principles, will transform the staking…
You can take part in our staking now, using the ERC-20 token, via the above site. Next auction in about 24 days or so, I believe. And with all those tokens out there locked, it shows enormous commitment to the Fetch space — locking up out of ability to trade shows great faith in the systems and applications we’re building and future utility of the token increasing (it is, after all, the fundamental fuel of autonomous economic agents and ai-powered smart contracts!))
Q:What is building a blockchain from the bottom up? Is it a creation, innovation or a difference from competitors?
A:Unique problems require unique solutions. We needed certain levels of performance, we needed solutions to consensus issues, and we needed high-performance, resource friendly ai capable smart contracts. We also needed to be able to create the environment for economic agents and enable the future embedding of even more ML to make other things work as time goes by. Getting the efficiency, robustness, performance, low resource use, etc., that we needed required us to build our own technology. And you can see this, of course, on our github. We are very proud of the quality of the code that we have built.
And to answer the other part: it’s creation, innovation and it’s different, for all the reasons I mention above.
Q:You will agree with me that ADOPTION and UTILITY is very key to project survival, what’s being done to drive global adoption of FETCH AMIDST CRYPTO REGULATIONS AND LEGISLATION
WHO ARE YOUR MAJOR COMPETITORS, LIKE CORTEX AND HOW DOES YOU PLAN TO OUTPLAY THEM AND BECOME THE LEADING AI BLOCKCHAIN?
A:I agree with you completely.
We constantly engage with commercial partners, our community and developers in order to build actual use cases and deliver them. I popped a couple of recent partnerships above as they are great examples of things that we are able to do with our ai-powered smart contracts that would not be possible elsewhere, and it is this new activity that we work on. We’ve also looked at mobility: building systems that show how ride sharing can be completely replaced with our smart contracts and autonomous agents, as well as showing with other partners earlier this year that we could shave 30% of average journey time by optimising electric cars’ visits to charging stations. This is no small feat: it is real, it works, and it has led to further exciting work that we’re building that combines some of this to do things like optimising congestion out of the road networks.
Fetch both uses and enables AI. It uses it to connect agents, provide trust and reputation, and it enables it with its ai-powered smart contracts and how it connects agents together (as those agents can provide predictions, insights and further knowledge to a network that is learning how to deliver those capabilities to those that need it)
Q:Fetch_AI can your technology be used for developing decentralized exchange platform? If yes will your current token be used as a gas, as you need your own block chain to operate?
A:Yes. It can, it does, and we have lots to talk about on this subject in the coming weeks and months.
Stuff that happens on main-net will use the native FET tokens, not the ERC-20 tokens. So FET will become the fuel for all those operations: it’s the fuel for agents to get things done, and the fuel for running ML and other AI applications in contracts as well as smart contracts themselves (some cool federated learning stuff too!)
Q:What problems and uncertainties do you face when you look at your project? What changes will occur to the processes and security features after the release of the Fetch.ai home network? What advantages does the smart contract language Etch offer?
A:Life is a collection of uncertainties that unfold on a daily basis. I can barely predict what’ll happen in 5 minutes, let alone tomorrow, next week or next year. I guess that’s what makes it fun 🙂 But we are creating adaptive technologies that can learn how to change themselves to meet the current needs of the network’s users. That helps a LOT when it comes to delivering value. We’re also working on putting in place the underlying mechanisms that allow others to build and deliver, a sort of autonomous economic internet of value.
With regards to Etch, it’s designed to work on our sharded ledger and work with our ML features. You can do CNNs, for example, and other ANNs on-chain, and our ML interface is pretty close to Tensorflow 2. It’s worth looking at our docs at https://docs.fetch.ai (which we’re constantly improving, btw., new stuff on that in a week or two) and also our work at https://build.fetch.ai where we have a solidity to etch conversion tool! This is to help others get their existing dApps going on Fetch quickly, and we’re working on other new stuff there too.
Q:Nowadays many project’s tokens does not have a clear valuation framework. How does Fetch.AI excels in that segment and is better than other projects?
A:Yes. On-chain AI.
And for us: it’s attach agents to your data, IoT devices, services, people, infrastructure, and let the network help you deliver their value. For large scale networks (transportation, telecommunications, decentralised finance, supply chains, healthcare, etc.) where it’s harder and harder to manage them from the top down, Fetch offers an opportunity to let them manage themselves, from the bottom up. This is why we talk to the rail industry, for example, and telecoms companies, and anything involving supply chains: we get better utilisation and more robustness as a large population of small, autonomous things, than we do with a smaller population of monolithic centralised systems. Those centralised things don’t move fast, can’t adapt fast, are slow and brittle.
Q:What growth potential does the AI industry have? What are the difficulties of AI industry? Why did you choose AI technology to develop?
A:AI is everywhere. You are probably carrying numerous neural networks around with you right now, and probably interact with many, many more in the course of your daily life, mostly without realising it. We’re providing a platform for decentralising that, making it available to everyone, and allowing the value that is derived to be delivered to those providing it effectively. AI’s biggest problems were computing power and ability to move data around, both of which are pretty refined now. Decentralistion allows for more computing power and interesting ways of combining data without disclosing it — this is great in industry where you may wish to contribute your amazing data to a model, but not disclose it to others because it’s confidential.
so we’re providing a platform that enables new approaches. We’re providing a platform that delivers a decentralised world where agents can be smart, can see, explore, interact with, that rewards intelligence in new ways. It’s very exciting!
Q:What is building a blockchain from the bottom up? Is it a creation, innovation or a difference from competitors?
A:Briefly; this is a philosophy of using a large population of simple things out of which more complex behaviour emerges. It’s like ants, or bees, or cells. The failure of any individual part makes no difference, and the system can adapt to changing circumstances as there is no single point of failure. It’s like the way the Internet works, but applied to decentralised technology, with the addition of an economic framework. Also…
A Decentralised Virtual World of the Future
Teleportation, multiple dimensions, chemical navigation: just some of the many ways we optimise a digital world for the…
Q:How will Fetch plan on funding the future cost of running the foundation and paying wages to the large team? Will you set up any sustaining treasury, if not then how?
A:Running low on time, so here’s a post I wrote earlier in the year you might find helpful on the utility side of things;
BUIDLing on the Fetch.AI Economic Utility of the Future
Build, deploy, play, explore: how the Fetch.AI ERC-20 token enables access to the test network and all the things you…
On the other side of things, that comes down to good governance and delivering commercial opportunities that generate revenue. Additionally, of course, any increased utility value in the token makes a difference there, too, as the foundation is charged with the development of the core architecture and helping grow the community as a whole.
Q:What are the benefits of combining two pioneering technologies between blockchain and AI and how they can help each other overcome some of their explicit limitations?
A:AI and blockchain are perfect bed partners. They are ideal examples of technologies that where combined, are way more than the sum of their parts. Blockchain should not be treated as a thing in isolation: it’s an enabling technology that when combined with other things can transform the world we live in.
Q:Big and modern cities around the world are participating in the Blockchain race. Most recently China launched a blockchain-based smart city identification system, and only major Blockchain projects have led governments to trust them. So does Fetch position itself as one of the top projects with Governments?
A:Smart cities are perfect examples of a complex network with lots of moving parts.
When you put autonomous agents in those parts, energy, transport and many more things can be optimised from the bottom up and not in isolation from each other. That makes Fetch technology ideal for being part of Smart Cities, so it won’t surprise you to learn that we are working in this space.
Q:With 30k TPS you are really ready for the tasks? What advantage does this TPS have on the project? Is it just to boast a higher tps than rivals?
A:No. We are not playing “ours is bigger than yours”. And you should ask this question to everyone who has a high TPS: Why? What’s the point? It’s like saying “My bicycle goes at 1000 kmph”. There’s no point in that, the wheels would catch fire, the rider would die in no time… it’s just not for any reason. Fast TPS has to be there because of something, and in Fetch, it’s billions of autonomous agents transacting with each other to build value for passing on to others. This requires high transaction rates, and we needed to build technology that was capable of doing that.
Q:Is there any similarity between fetchai & cortex projects?
A:We are doing different things, but we do get to use some of the same words and subject areas 🙂 Wish I had time to write a touch more about this, but it’s worth looking at the two — what we use AI for vs. what we enable others to do with their AI (which are two things, of course) and the autonomous agents side of things as a method of getting the economy to start doing all the work.
Fetch.AI delivers a groundbreaking economic internet that enables emergent solutions to complex problems. It does this by enabling the deployment of multi-agent systems over a decentralised network and provides tools to enable the construction of intelligent agents. Fetch.AI delivers a unique, decentralised digital world that adapts in real-time to enable effective, friction-free value exchange. Powered by innovations such as the smart ledger, Fetch.AI has digital intelligence at its heart: delivering actionable predictions, instant trust information, enabling the construction of powerful collaborative models, improving efficiencies and streamlining processes.