Nethermind Workshop: Blockchain & Robotics - ETHGlobal San Francisco 2024

[Music] so yeah really really excited uh this is like a relatively fun idea for us which is blending Robotics and blockchains and we think there's some values that blockchain share that are incredibly useful for building robotics systems so yeah why why robotics well um yeah we envision a future where robotics uh are going to be operating in unconstrained environments so today they're typically deployed in factories these are highly constrained environments and of course the rules inside the factories are very different to the rules for robots operating in open systems so Concepts such as safety become critical um I'm sure we've all been in a wh Mo I mean if you put your hand up if you came here in AO yeah quite a few of us already so um we're all pretty familiar with uh just you know how dangerous these devices could be a Vil function the second piece is uh we think robots are going to be operating in uh collaborative environments and this means that humans are going to be working alongside robots and it means we are going to be placing a great deal of trust in the robotic systems that we're going to be uh interacting with and the third piece is uh we think these uh interacting is going to be routine they're going to extend into our homes they're going to extend into open environments around us and in 10 or 15 years time it's not going to be unreasonable to expect that we will probably be interacting with robots on a day-to-day basis I mean in San Francisco people already do they take taxis Robo taxis and there's no human operators no human drivers so this is very much a reality and it's going to become Global it's going to become ubiquitous going to become widespread and yeah we see this there's some very exciting Hardware being developed today um these are all single operator hardware systems uh you can see there's uas these are zipline uh DRS that can autonomously navigate in remote environments we have humanoid robots just a few here is the figure robot which is uh the top Center um we have the Chinese variant of a figure as well um and we have the 1X system which is designed to operate in homes it's kind of more of a safety considerate uh robot and then of course we have the self-driving taxis which are potentially the most lethal systems of all of these just because they have a lot of mass and so obviously when we think about multi-operator robotics where we have multiple different robot taxi operators for example they might want to securely share information with each other and they might not also uh be uh you able to communicate with each other they might not be connected to network there's potential attacks in so having some form of like localized form of trust is is going to be very very important and yeah as I as I spoke about this uh before the penalty for failure is extremely high for robotic systems you know if a software system breaks you know of a a web server who on my website or even you know say what of a validat hosting a blockchain fail um you know in AWS or in a server or data center or maybe my bank servers fail really the ramifications dependency for fail is uh maybe I can't pay someone maybe somebody loses a little bit of money um it's not great and uh there are some environments where software systems are critical like a medical but when we talk about physical systems and failures of physical systems which are under depend by software then the ramifications or the peny for failure really are catastrophic uh you know if a car crashes uh it could result in loss of human life or fatalities so these problems are immensely important they're immensely valuable and somewhat underappreciated actually in my opinion I think people get very excited about uh this kind of unconstrained open Tech but they don't really seem to make a lot of considerations about safety and as always failure is to be expected so uh unfortunately you know devices might work you know 99.99999% of the time but it's that edge case that is still going to result in the system failing and I kind of like the kind of trolley problem analogy um it's nice to know uh what kind of decisions a robot made um and you know uh this is maybe more of an Ethics question but having some kind of immutable record where you could uh store information about Rob robotic decision making um and other robots around it and other humans could verify and audit that robotic decision making that they could prove that the decisions made by the robot were made by a model that's been validated um and they can prove that the Hardware was never tampered with is going to become really really important because when we have lots of these robots we have very very complex uh scenarios which might arise and these robots uh need to make autonomous decisions um and as you can see here this is maybe very kind of a very jovial representation uh but yeah we have lots of trolley problems in very complex environments and so what are the kind of guarantees that blockchain can provide us with well really uh there's tamper resistance so we can publish signatures uh about physical Hardware to prove that the hardware uh you is running a certain type of software um that the hardware itself hasn't been tampered with and we could publish that to an immutable ledger so that's very very valuable the second piece is we can authenticate messages so we can uh authenticate The Source or the origin of a message and tie it back to a physical device and this really brings in Concepts such as uh device identity and we're really really interested in uh really how we can create large networks of device identity similar to what world is working on you know world is working on a human identity platform uh proof of humanity and it's not unreasonable to expect that we could create uh platforms or basically Primitives for robots to be able to communicate with each other um on blockchains and Sh share information with signatures securely with each other and the last piece is really State conflict resolution and um I think this is really really cool uh there's things Concepts such as uh UTM which are un unman Traffic Systems and you know when you have very large uh networks of operators um you may want to create like an economy around the provision of airspace for drones or an economy around the provision of uh you terrestrial space of uh notion of traffic uh for autonomous cars you know how do we prioritize uh prioritize the actions of autonomous vehicles and something like an ambulance for example might take president over myself uh commuting to work and uh you know there's many other examples here where we can apply these kind of economic systems and economic Concepts from blockchains to um to Robotics and robots operating open environments so this is super super exciting stuff and of course uh this is really scratching the surface of all of the potential applications and intersections between uh between blockchain technology and distributed Ledges and permissionless open distributed Ledges such as ethereum and all of the scalable rollups and uh and physical hardware and I think we're at a really interesting intersection a really interesting point now where we're now reaching beyond the kind of virtual trust boundary and exploring the kind of Physical Realm of bits and atoms and this is really really exciting uh points in time where you know L lot of these robotic projects and actual applied demonstrations of uh of robotic systems such as wayo taxis and yeah I think this is going to be very very exciting over the next couple years so with all of this in mind with kind of the problem space I've Spilled Out we have given some ideas we purposefully left this quite vague uh I wouldn't say vague but maybe open actually just because we want to give people the space to explore really any idea that's potentially related to Robotics and uh you know you don't need to scratch uh you don't need to dig really into uh physical hardware systems um but even ideas that touch ml uh machine learning and the models that might run on robots are also highly relevant uh to to this uh this hackathon so the first one was robotic identity now uh robotic identity I think is going to be really really important uh in a multi-operator system say we have large distributed autonomous taxi networks there are a lot of uh kind of Concepts that are required uh around ownership um and control of the robots um that are really going to be underpinned by identity and this is things like you know who owns the robo taxi where's the value crew to if a taxi crashes you know for any situation um where does the liability uh you know uh who who Who's responsible for liability for for an accident and so having uh a kind of uh immutable verifiable onchain credential system that is also tamper proof um which means that you know I couldn't extract the chips from my taxi or my friend's taxi and put them in my taxi um and you know try and spoof the system these are immensely valuable and um we think about these as kind of a primitive uh where blockchain is highly relevant uh it's really really really good use case and it's why we think this kind of uh these proof of humanity uh identity systems uh can really be extended into you know uh proof of machine or uh proof of AI kind of uh model systems the second piece is uh secure Communications so blockchains are very good for securely uh sharing messages with a a network and um you know we think uh that open peer here uh vehicle to everything communication networks are going to become really really important now there are attacks today which are highly relevant here um you know we a lot of the systems that we use uh depend on these kind of centralized or semi- centralized uh communication networks to share information and these include things like 5G and 4G and they also include uh GPS actually to receive passively receive location information so uh in a scenario where a sophisticated attacker might use a Jammer a lowc cost Jammer to jam out you know uh 5G radio waves or 5G frequencies or jam up the GPS frequencies these robots really can't operate at all so if there's methods for robots in their local vicinity to securely share information with each other and also proof of that information is is trustworthy that was perhaps derived from real sensors rather than synthetic um and wasn't sped uh again that's would be a very very valuable use case of blockchain it takes advantage of you know immutable Ledges um it could also take advantage of an identity system and some of the other ideas that extend into this are really around decision making you know you might have a swarm of robots so you have like a certain rescue swarm of of uas um how can they efficiently allocate you know tasks to each other over you know what kind of areas they want to search um and uh the last piece is really shared learning and uh this is be applicable in closed environments but also very applicable in open environments uh the world we live in is not stationary uh and this is particularly important in kind of if anyone who's here has ever worked with geographical systems um you know Maps change over the time the world around us changes all the time buildings new buildings are built you know Road works are put up uh you know it's the world is a very Dynamic place and it's very important that these systems are able to share uh securely information about this uh the dynamic world around us and again we think that uh using a blockchain using a mle ledger an mle record is a very very good use case uh for this uh one of the uh ones that I I really am a massive fan of is secure hardware and verifiable Computing and um yeah this is really about providing guarantees and assurances uh to end users about uh the Integrity of process on a piece of Hardware the uh correctness of a software it's running and there's some really really exciting Tech um this is really at the intersection of uh I'd say uh Hardware manufacturing and also cryptography and um here we can verify that you know if you stepped into a taxi a rer taxi for example this is one example you could verify that the model the taxi was running was like a canonical version of a safe uh you formally verified and uh pent tested and and uh uh you know uh correct self-driving model or the model itself hasn't been tampered with um you know it's it's not uh beyond the realm of the imagination to think that uh some uh you know nefarious actors could potentially interfere with Robo taxis um and cause a great deal of harm so being able to verify the Integrity of the hardware the software systems would be really really useful and the last piece is actually supply chain uh as we start to kind of uh you're more and more interested in trusted execution environments the root of trust is uh something that's really really important uh for tees to verify that the observability of the private keys or the um puffs or we call them physically unclonable functions uh which is a physical uh measurement which is made of a device is secure that hasn't been tampered with that some you know nefarious party isn't aware of this a very important uh piece element of the hardware which is used to derive a secure cryptographic key um this is really really important for trusted execution environments it's still kind of like an unsolved problem and yeah we'd be very interested uh this is more a theoretical one in hearing people have like supply chain applied blockchain Solutions and the last piece is data problems and this is uh really really poignant I'm sure some of you may have seen the world conference yesterday there was a very cool kind of demonstration of proof humanity and how it could be applied to verify uh someone using a deep face model on Zoom uh or'll verify if they were using a deface model if they were real and um yeah this is going to become increasingly relevant uh in the future I mean these kind of models that uh can change your face are really really lightweight I can go on GitHub and find about hundred of these and um it's really really easy now to generate synthetic data and it's quite scary there are some real real uh political kind of ramifications I mean we're approaching election season in the US um for this technology and so being able to verify that information was collected by a physical device by a real device by a camera um is really really really interesting concept for us and uh we think for robots this also extends into the kind of autonomous agents realm robots might want to share sensor information with each other and be able to verify that that information was untampered uh that it was collected by a real sensor a real device uh is really really important um to make sure you know in this big uncontr open environment that the robots are all using real data they're not you know trying to trick each other or trying to manipulate each other and um yeah so this is the hackerone uh we're really really open here you really just need to incorporate uh some use case of a of a blockchain um it can be ecdsa but really uh yeah we can be even more open than that and the submission doesn't require you to actually build any hardware we appreciate 2 days is not a lot of time and um really we want you to uh you know uh create an or come up with an exciting interesting use case uh of blockchain that is relevant to Robotics and um yeah we' just love to hear what what uh what people come up with uh genuine original uh kind of innovative ideas are uh going to be the winners and of course there is a bonus if you can create a working prototype um it could be something that runs on a phone it could be something that you know maybe runs on a piece of Hardware it would be super impressive if if uh any of you have brought Hardware with you um like you know a drone or a small robot and yeah uh really really interested to to hear what uh people have have to build so great does anyone have any questions how do you apply this robotics explain a little bit more on that yeah there's um quite a few kind of open libraries uh there's RS robotic operating system and rs is currently used for sharing secure messages um inside of a robotic environment it's basically a Linux operating system and um there are a couple of libraries I can share them uh which can be used to publish these messages on chain um so that's kind of like quite a good starting point um but again as I I said this is quite like an open book uh quite open book uh hagon there's also some prior art around uh auth uh data authenticity and Providence you might have heard of the content authenticity initiative which is backed by some very big names such as Adobe and open Ai and that's kind of an open standard for generating these authenticity or problem signatures so that's also a very good place to start but again it's not kind of blockchain native um it's for a very narrow set of use cases and so we're reallying looking for people to take uh either original new ideas or existing Technologies and extend their use case uh by applying blockchain yeah so here uh proving decision- making process um you know proving that a model was uh you know behaving as it should the output is correct um this is something that you can kind of already do with llms but uh it's also something that could be extended into uh robotic system as well sorry of course yes I I'll share these slides afterwards um if you come up and yeah so yes yes so the Apple uses trusted execution environments um and what they allow you what the the guarantees they provide to users is that users can share uh basically encrypted information with the trusted execution environment um and The Trusted execution environment can securely decrypt the information perform a computational process and then reencrypt the output and this is really important because CK uh is not very good for very computation expensive processes and fully homomorphic encryption Hardware just doesn't really exist yet for these kind of complex processes so it provides a guarantee that the process running side of a physical Hardware is nonobservable to an external user and this might be really really useful uh for uh robotics use case [Music] yeah yeah they use something called physical or physically uncleanable function yeah yeah I think um you know those functions so device ID to serial could be a very good basis for a robotic identity system exactly um thank you thank you everyone I think we're uh on time but great thank you very much and look forward to to seeing what you will come up with if you choose to to work on this hack Fun [Applause] [Music]