April 01, 2024
Solving Blockchains Verifier's Dilemma using Game Theory with Truebit | Episode #81
Join us in Episode 81 of Great Things with Great Tech, featuring Blane Sims, Head of Product at Truebit. Dive into Truebit's journey from improving Ethereum's computational limitations to establishing a new standard for offchain verification. Understand how Truebit's protocol, based on game theory, assures computational accuracy and integrity, crucial for the advancement …
Apple Podcasts
Spotify
Castro
RSS Feed
Breaker
PocketCasts
YouTube Channel
Join us in Episode 81 of Great Things with Great Tech, featuring Blane Sims, Head of Product at Truebit. Dive into Truebit's journey from improving Ethereum's computational limitations to establishing a new standard for offchain verification. Understand how Truebit's protocol, based on game theory, assures computational accuracy and integrity, crucial for the advancement of decentralized applications.
Don’t just trust, verify! Truebit Verify lets Web3 applications integrate safely with any data source, move data securely across ledgers, and trustlessly call outside functions.
Truebit revolutionizes blockchain computation, pushing the boundaries of web 3.0. In Episode 81 of Great Things with Great Tech, Blane Sims, Head of Product at Truebit, shares their journey from addressing Ethereum's limitations to pioneering offchain verification. Truebit enhances blockchain's scalability, enabling complex, efficient computations. Founded in 2017, their innovative approach, rooted in game theory, assures computational accuracy and integrity, driving the evolution of decentralized applications.
Topics Covered:
- Truebit's Genesis and Ethereum's Evolution
- Offchain Verification: Scaling Blockchain's Horizons
- Truebit's Role in Web 3.0 and Traditional Tech Integration
- Game Theory: Securing Decentralized Computation
- Truebit OS and Verify: Powering Transparent, Reliable Applications
☑️ Technology and Topics Mentioned: Truebit, Blockchain, Ethereum, Offchain Verification, Web 3.0, Decentralized Applications, Game Theory
☑️ Web: https://truebit.io ☑️ Crunchbase: https://www.crunchbase.com/organization/truebit
☑️ Support the Channel by buying a coffee? - https://ko-fi.com/gtwgt ☑️ Interested in being on #GTwGT? Contact via Twitter @GTwGTPodcast or visit https://www.gtwgt.com ☑️ Subscribe to YouTube: https://www.youtube.com/@GTwGTPodcast?sub_confirmation=1 ☑️ Subscribe to Spotify: https://open.spotify.com/show/5Y1Fgl4DgGpFd5Z4dHulVX • Web - https://gtwgt.com • Twitter - https://twitter.com/GTwGTPodcast • Apple Podcasts - https://podcasts.apple.com/us/podcast/id1519439787?mt=2&ls=1 ☑️ Music: https://www.bensound.com
Transcript
RAW TRANSCRIPT
just before we start episode 81 of great things with great Tech I just want to highlight that the following is not investment advice it's for informational and educational purposes only hello and welcome to episode 81 of great things with great Tech the podcast highlighting compan is doing great things with great technology my name's Anthony Span in this episode episode we're talking to a company that revolves around overcoming the limitations of blockchain computation ensuring integrity and accuracy in a
decentralized environment thus facilitating more complex and reliable applications in web 3 and beyond that company is truebit I'm talking to bla Sims head of product at truebit welcome to the show thanks Anthony so happy to be here today yeah awesome and before um I gladly dive back into the world of blockchain and web after a bit of a Hiatus on the show just as a reminder if you like great things with great Tech would like to feature in future episodes click on the link on the show notes or go to jtw j.com catch all episodes on
all the good podcasting platforms Apple Google Spotify hosted and distributed by Spotify podcast finally go to jtw JT podcast at YouTube link subscribe follow all that kind of stuff so you get all the goodness of the future episodes and by the way all previous 81 episodes as well well right bla now that we've got that out of the way uh let's talk about true let's talk about verification in the world of web 3 but also I think very interestingly how you know web 3 and more traditional web 2 world are coming
together and this is kind of like this happy medium that we now found ourselves in in 2024 absolutely um well I think the the first place to start is really uh how trouet got its own start right so uh trouet goes back uh really to the earliest days of the of the blockchain world in our case ethereum um uh our founder Jason toy uh is a computer scientist a researcher uh who um was really really close with the ethereum community its early days and um essentially started pointing out things that were uh you know I guess issues in
the in the design of ethereum that if they weren't fixed would be uh you know big problems right so uh these are uh things like you know the fact that the the what are called validators the the nodes that make up the ethereum network uh weren't necessarily incentivized uh to perform uh the really complex types of compute uh that everybody was imagining would be possible within within the ethereum blockchain uh fast forward uh you know to 2017 uh which is really the founding of true bid as a company uh we invented
something called offchain verification something that became uh really widely referenced as a uh you know as a foundational technology for scaling blockchains in general um meanwhile our research uh kind of kept going in that direction of great yes blockchains can scale but what about all the compute that doesn't actually fit on a blockchain what about the types of compute that are really you know emblematic of what happens in the real world these are kind of the the more you know everything from mundane uh kind of
uh backend tasks that power applications to uh you know API calls all of these things that really make up the way applications work uh and that's where we're at today so uh we're currently in in Early Access of a what we think is a pretty revolutionary product which is taking the same concept which is verification how do you make uh just as in a blockchain how do you make a lot of different computers that you don't know or don't trust uh you know trustworthy by checking their work um and how do you apply that then uh not
just to the transactions that's uh where where blockchains have specialized but to this much broader world of data integration uh of uh you know more uh you in some cases more routine types of compute tasks but in some other cases really really you know kind of large scale high performance computer tasks as well that that again aren't things that fit in blockchains yeah and I think you know you talked about you know solving a problem of ethereum scalability and in the early days well not even the early
days in the in the hype days of blockchain and crypto back you know a couple of years ago a lot of people were talking about ethereum as as the internet computer that's kind of what it was was well known as um um just explain that concept a little bit and then how um true bit helps to you know effectively you know from a is is it more like the concept of a side chain than anything because side chain to eums are like what like polygon and a few other ones like that but this isn't that is it this is something different but
maybe no this is yeah very yeah very different so yeah good question so truebit is not a blockchain uh we ourselves are not blockchain um we do certainly uh leverage blockchains in a really interesting way that I can get to in in a few minutes uh that that help us secure what we do but but yeah to your point uh you know the original vision and I think that you know that I've heard that the term used a lot of of ethereum being a world computer um you know was part of the you know kind of the ideology of the early days of the
the web 3 space um and I think what um you know what happened and what you see uh if you look at uh you know web three applications that are in our definition are applications that one way or another are based on blockchain transactions or leverage blockchain for for for recording transactions uh what you see is uh you know the majority of the code in those applications that you know that you'd actually go use isn't blockchain code it's it's Web Two code these are uh things that are running in platforms
like AWS Lambda or Microsoft Azure um and uh you know taking advantage of all of the robust capabilities that are needed to to build something usable uh to your point around um you know how did we get there from this this Vision um you know the the idea of of ethereum in particularly being a world computer was the uh you know here were all of these you know tens of thousands of of different uh you know servers you know called nodes uh that were hosted by anyone that wanted to plug in you know a server to the internet uh and uh no one
was in charge you know the the ethereum protocol itself was uh was running uh you know the uh all the transactions through those nodes Etc and then you would be able to use something called smart contracts uh to write uh programs uh that uh you know that execute on that on that Network well all that's true um you know there are tens of thousands of ethereum nodes there are smart contracts that execute on ethereum um but I think the the thing that wasn't clear at the beginning that is clear now is what are
the types of uh code that are appropriate to put in smart contracts well as it turns out the things you want in smart contracts are uh essentially uh the uh the code that's actually going to execute to to make a transaction happen it's the transaction layer it's not any of the other layers of an application that you might want to build so yes in the the blockchain world there's ethereum as what's you know now called a layer one or a foundational blockchain uh there are a number of other blockchains like salana uh is a
you know really popular one Avalanche there's new ones like sui and SE that all kind of play in this this kind of transactional space uh specific to the inum community there as you mentioned our our side chains which have been called you come known as layer 2 uh again a lot of that early research that trit did uh fed into the development of layer 2 uh Technologies you see uh kind of references to that uh in in the white papers uh of those platforms um and those layer twos are you know great additions to that that ability to scale
and perform lots and lots of transactions but you still are left with this question of how do I build a useful application uh on top of uh this this transaction layer that blockchains are providing and that's where tribit comes in to to to date if you uh you know what you end up doing is uh uh as a web 3 developer one of the key things you're generally looking for in your application is transparency and the other thing you might be looking for is something called autonomy so I want an application that I
can show to the world uh that all of the data in the application is correct and true that none of it was modified I didn't uh for example um you know in an in an automated insurance contract that pays out when um you know when a particular event happens like uh you know I've actually heard of a uh you can get hurricane insurance in Florida that will automatically pay out it's a blockchain based contract it will automatically pay out if the uh if there's the wind in your ZIP code goes above a certain speed well you don't
want someone to be able to go and just say hey I had a really windy day here and here you know how you know how do you prove that the wind was above a certain speed how do you actually make that input into that smart contract uh provably correct that's the kind of space that we're that you know that we're uh we're working in and that same question applies um you know maybe not to every single piece of code you're going to write in an app I mean do do you need the uh you know absolute provability when someone's dragging a
mouse across their screen and clicking on a button maybe not but where where data becomes part of the transaction where it's being sourced or or where there's a destination for data in a transaction uh and where there's critical things happening to transform that data those are things that need to be transparent and verified otherwise kind of all the things that you're doing to put it on a blockchain become rather suspect you you're like well great I have this record on the blockchain but
where did it come from is it true and that's where trit comes in yeah so in terms of that I mean you've got a couple of products in the lineup you got the tribit OS and the tribit verify so verification is is Central to a lot of your messaging right um just explain why there is the need for verification I mean OB it's it's it's obvious to certain people but in this world why is verification so critical yeah that's a great question uh the the first thing to think about is what is verification what do we mean
when we say that because I think that that word can mean a lot of potentially a lot of different things so what we're saying in our case is is verified computation uh this is where you're asking uh you know that server that computer that is executing code to be able to prove that it actually executed the code you intended it to uh to be able to prove that the inputs that you gave into that code weren't changed and that the outputs from that code weren't changed uh when they before they were returned back to you so the verification
exists to to provide that proof um the the way things work under the hood uh at least with trit is uh we uh actually uh kind of open up the hood of the car uh and look at a machine code level uh at the uh the compute instructions that are going uh you know going on to help that computer that's running the code uh you know prove one way or another that it's correct uh the good news is that we don't have to um you know every time uh stop and step through a program at a machine code level we we've employed uh
some really kind of great techniques from game theory that allow us to do this just when there's uh lack of clarity uh around whether a computer's provided the correct result uh but the the idea that you can be kind of called to task as a computer and say basically show me the op codes uh that came to this result step by step as I step through a program uh is what provides that backdrop that provides the verification and then what we do is take that forward into a direct uh essentially piece of evidence we call a
transcript you can you know take this transcript from your uh you know from your API call or from your uh serverless function call take a transcript and provide it to someone else to say hey I can show you uh that that this code that I executed is correct and I can show you why that's the cas because I have a transcript that steps through every single uh you know event that happened to get to that result okay that sounds like timec consuming but maybe talk about the efficiency afterward but I want to step back to Game Theory because
I'm quite interested in that because I was listening to another podcast that was actually more talking about um history and and wars and whatnot and the concept of game theory around there that people you know that were scrutinizing whether or not to you know launched a nuclear bomb back in the Cold War they were big fans of Game Theory so how how has Game Theory come to play in in this world well so game theory in in our world is all about um setting up the right uh set of incentives and the right uh what we call protocol right for uh
for the compute pass task to take place so uh if if you are uh in trits case one of our decentralized compute nodes and you do provide an incorrect result you're you're financially penalized for for providing an incorrect result and you have to stake in uh and uh you know into our Network and have and essentially put money on the line uh to uh you know to to ensure that you're you know you're in the game honestly um so what happens you know because there are incentives when you provide a correct
result disincentives to provide an incorrect result and a mechanism to judge whether uh you know what you're doing is true or not um you end up you know know creating an environment where uh you know the vast majority you know of the time you know over you know 99 you know a few nines percent of the time uh you know the players in the network are going to behave you know rationally and not get themselves penalized uh and then when you do have you know what we call erratic Behavior Uh you know we have the mechanisms in place to to
penalize that and a uh and and that you know that that transparency it comes from knowing that all the players are acting rationally extends not only to you know the decentralized nodes of uh you know that are performing the compute task but also to any of the components that we as trouet have to provide to do things like connect you to the internet you know connect a task to an internet or connect to an API things like that so this uh this kind of all surrounding uh you know kind of Spotlight that's uh that's going on all
the players uh in this uh in this network are uh you know kind of what what we're talking about there I I get it get I get it let's go back to that efficiency and scalability because obviously you know what trt's mission is one of them is to address the challenges of scalability and computation and blockchain Technology like you mentioned about making ethereum better um and cheaper as well more efficient but then you know what you described earlier it sounded like you were putting you know more steps in for the verification
because effectively to verify something you've got to take up you know a cycle of of whatever it might be so how do how do those how are those things working together so how are you guys you know verifying adding what same like more steps but then lading to more scalability yeah um the way we do it is we um we verify in parallel right so um you as you run a you know kind of a transaction through our Network um you know first step you write uh you know serverless function code again think AWS Lambda or Microsoft Azure typically uh
this is code that's written in JavaScript or python or or rust um you know uh you you write your code we're going to uh you know provide a uh you know a an interface into that code a web services interface to call the code so when the code gets called We're executing it in parallel across multiple nodes in our Network um and uh again in that that 99 and several n% of the time uh we'll get a result back instantly from all of the nodes uh and they'll the results all agree um and at that point
we give you an initial check mark that hey uh if you're doing something that's time constrained on in and our our Benchmark for this is usually uh somewhere around half a second uh you get a result back that uh that is uh you know appears to be valid uh you know we'll move on if you doing something that is very highly secure there's a second layer of steps that we do uh that might take another second to complete so um you know we're we're giving results back you know half a second to a second
uh you know after you run uh kind of your your request uh I think over time we'll see those timelines kind of decrease even further but certainly in line with what you'd expect from uh you know from uh kind of other serverless platforms yeah so you gave an example of of that Insurance app um or the insurance platform as well that was kind of validating and making sure that the wind speed was as it was and it wasn't being manipulated for financial games so that's kind of a good example of your
target audience So you you're kind of bridging web 3 and traditional application Platforms in developing the development world right that's kind of the idea here that's I guess where where we've settled down you know obviously crypto this year has gone up a notch again it's back I guess you could say um and now when the newspapers are running stories about blockchain and crypto again you know that it's got the hype but from your point of view where have you seen it settle more so not in the
last sort of six months but more in that last sort of down period where you know the heat went out of the market but you guys were still doing your thing because you believed in this technology yeah I actually think the um the period particularly last year where um you know there I would call it if you believe in the the concept of the hype cycle you know that trough of disappointment in terms of what uh you know what web 3 might mean what blockchains might mean and certainly it seemed to coincide with you know some of
the um you know some of the financial shenanigans that everybody was tracking at at that time as well um you you know that that trit disappointment really uh allowed us and I think allowed the the projects that we work with on trit uh to kind of emerge as what's real what's actually you know meaningful in web 3 it's not uh you know it's not about the the hype around you know the latest coin or something like that that might be you know uh where the you know the most of the news comes from it's really around
uh the applications and the and the disruption that uh you know the decent calization in particular as a as a mode of of compute uh you know can bring to you know to different uh different markets so for for us uh you know getting through that um that trough you know pointed us at the uh you know really interesting things that we're seeing like we're seeing uh you know projects that are actually uh interestingly even connected in with the Enterprise around compliance use cases uh I've got a uh you know one of the
partners working with that that works in the Pharma uh uh space the projects working with in the Pharma space that's that's essentially uh we're helping them prove that private documents haven't been altered uh without revealing the document so how can you say you how can you show that uh something that uh was supposed to be locked and never changed wasn't changed without actually showing someone that document uh we can provide the transparency around compute to make that happen um we see kind of projects
uh that are in supply chain supply Chain's been one of the probably longest running kind of uh Industries that's you know looking at blockchain for transformation because there are existing so many different independent players that don't have uh you know really well defined ways of working together in in a reliable way uh so we've seen you know one of our supply chain Partners bring to Market a brand new uh kind of product that actually connects data that's on disparate blockchains together um you know we've
seen interestingly and I think for where we're looking moving forward um a lot of movement in the AI space um so we've got uh you know one of the key projects we're working on in our Early Access program for trit verify right now uh is an AI based project that's uh you know using Tria to show that their automated decision- making is playing by the rules uh that they've set up and that uh you know they're not again changing any of the data or changing any of the results as they uh you know move data into their
model and and uh and call inferences on that model yeah I want to I want to touch on that a little bit later maybe finish with that because I think obviously it's it's the molding of two different worlds coming together you the two different hyp cycle worlds of the last you know 5 years but I want to talk just before we get to that point about how you've talked about it but the way that you plug into traditional apis and even storage platforms as well um and databases you know what's what's the key
way that that happens because I think it's it's interesting to understand how the web 3 world can cross over to the traditional I guess web 2 world if you want to call it that and you know what are the what's an good example of hitting an API doing some validation and then maybe even from my perspective getting a bit of data um you know having it transfer or created and stored somewhere maybe pulling it out looking at it verifying it and putting it back how can that happen with trit yeah um so great example of this uh would be uh you
know one of our uh one of our projects is working uh within Amazon's uh Dynamo DB as their uh their data persistence layer and um what we've done is essentially wrap their uh you know their calls to Dynamo DB uh into what we call truebit tasks um so when as an API call is wrapped as a truebit task essentially uh trit is making the trit network is making that call on on your behalf um we're doing it in a verifiable way uh we uh we use a uh uh a form of in this case multi-party compute secure multi party
compute to actually have multiple um you know nodes attached uh safely to uh to an API server uh to make that call and then we uh we provide a uh a transcript that shows essentially you know kind of everything from the you know the the TLs headers that were you know authenticating the server uh the you know that this was Amazon's AWS server that we authenticated to uh shows the uh you know essentially a snapshot of the entire uh transaction you know the request and response from that API server uh and uh
so yeah you get the you know the the task that you performed in Dynamo DB happens and then you get that's essentially uh kind of sidecar uh from uh from trit that is proof of what of what took place in that uh and that proof then itself uh that transcript gets uh you know publicly referenced on um you know on a blockchain that trueid is um is is uh kind of rolling transactions up to uh as a point of attestation for your uh your transaction okay yeah that that makes sense so it's kind of like logging and
sort of like what you would see if you were I guess debugging that sort of application you you'd see all those headers you'd see the calls um but what but what you're doing is you're taking that and effectively validating that to ensure that it was done correctly without any potential for a man in the- Middle attack or any sort of malicious intent or anything like that and it's it's it's there and once it's been committed through the truebit system that's the source of Truth effectively
yeah I think that's a really good analogy it's it's essentially like having a uh a public log but in this case the you know the log is not providing any of the private details of your transaction but it's it's publicly referenceable I can actually point you know through that um that transcript uh hash that we're rolling up into uh you know onto a blockchain that's the public reference to this this individual Atomic log record that documents everything and shows exactly what happened yeah and
what about the storage play what about integration into storage um it's very similar right so the you know again I guess our requirement here is uh you know for uh whether it's storage or you know that the case database API or you know any any sort of API uh you know we're looking at um uh you know integrating with platforms that are HTTP enabled so as long as we've got uh you know that uh the ability to represent storage something like S3 right as as an HTP transaction uh you know you can even
like a drop boox here depending on your use cases um you know we're um we're able to reference in this case offchain storage or or more centralized storage uh objects um in the exact same way I just described for that Dynamo DB transaction get it get it okay let's talk about um decentralization as a concept I know you got some thoughts about where where we are where we should be and obviously you know true bits you know methodology of pointing out that a decentralized way to attack verification is the right way to do it so give us
some thoughts on decentralization as it pertains to web 3 and then maybe how that extends into this more traditional Web Two world yeah um so decentralization is a essentially a base assumption in the web 3 World um this this entire universe of web 3 um which is you know depending on how you frame it you know is either kind of trying to change the way uh compute Works which is our perspective and and then there are certainly folks that that look at as this as a vehicle to you know kind of disrupt the uh you know the over
centralization and of of Technology uh you know in in just a handful of few Cloud providers uh you know wherever you you sit on that that Spectrum um you know what what is interesting about decentralization is this notion that um there are vast amounts of of compute resource yes many of them are owned by a small number of players but many of them exist uh elsewhere um you know there are compute resources within uh within universities within corporations within uh you know I have 15 plus devices in my own home uh everybody does at this point
that all have uh compute resources that are uh in many cases idle and can be you know brought to bear uh if uh if only there were a good way to do it so you know this this idea of of decentralization really comes down to how do you take uh a you know a resource uh like all of that idle compute and turn it around and make it something that's actually available to uh you know to anyone who needs it um and uh and when you start to look at that question you immediately come into well that sounds like a great idea but
but you know what about what about security what about I don't necessarily want my data you know going you know any which way you know what about what about what about and that's where the technology uh of folks like trueid and other folks in the the web few world is really you know answering hard questions um you know these are questions that are built around you know cryptography they're built around um you know thinking about how where do we go next from uh the standpoint of uh uh serverless compute um you know we're
thinking about uh as we mentioned before about you know how do you take uh you know abstract ideas from Game Theory and use those to to create new you kind of new security paradigms um so so for me you know decentralization I actually think is a bit of an inevitability right uh we will uh you know increasingly abstract the compute that we're doing to a point where uh you know literally any server anywhere could could perform it how do we secure and and feel safe in in giving a compute job to any server anywhere that's really interesting
because I just came off yesterday on my way uh back home from a trip I was in the car and I was listening to Alex freedman's latest interview with Sam mman uh from open Ai and his whole one of the things I were talking about was the need for compute just generally in the world for everything that's going on with AI um and Beyond right so you know there's only so many chips there's only so much fabrication there's only so much whatever that they they are specialized and using at the moment um and every
company is the same if if you take even the AI out of it look at um you know Amazon look at Microsoft with the amount of compute that they're pumping in for their Cloud platforms there's only a limited amount of compute in the world and at the moment we need compute to power the future and power especially Ai and then blockchain uses that computer as well in a slightly different way but the concept of then and what I what I just had having listened to that and listening to you talk was that maybe the
solution is to actually decentralize it so here's here's a theory right like in the future a lot of the AI stuff that that you do when you ask your model something in a chat bot it's actually going to use a compute that's potentially on someone else's phone somewhere in the world because that's that's the available compute that's out there at the moment right like I think that potential for decentralization is actually really really interesting given what I've just heard and giv The Thirst
for compute that this world actually has at the moment I I agree wholeheartedly which is should not be a surprise um look I think when you're you know when you're looking at a range of Alternatives you know which include things like you know Sam Alman from open AI trying to raise a trillion dollar uh kind of uh uh of cash yeah um and to go buy the you know I think the new price point for the the latest Nvidia chip is $30,000 a chip you know um sure a lot of good work is going to be done in that in that compute model but even that's not
enough you know the the Nvidia can't make enough chips and you know I don't know how many trillion dollar Investments the the world can sustain in order to drive you know an completely centralized form of uh you know in this case AI uh and you know let's just generalize it and say you know high high performance compute in general yes um so you got to look at the other end of the spectrum you know you can go big uh like you can go you know in the old Paradigm you could go the Mainframe route which
is a little bit what that's starting to look like uh or you can go okay how do we how do we break up uh you know these compute tasks make them smaller and make them fit in something like web assembly which is what trouet is built on right uh and so that any device anywhere can perform them meaningfully and how do we you know increase the capability of you know the the the smaller devices to to be able to do meaningful compute you know maybe in a different way and at a different scale but but maybe not to
very different uh kind of end result from from you know the power of the masses in a decentralized way uh I would say you know potentially outperforming the uh um you know what's possible in the uh you know the trillion dollar uh data center world that uh that we're looking at does it happen overnight no is that where things are at today uh not really uh but there are a lot lot of really interesting on roads uh you know into that into that Universe uh that are practical use cases like you know mentioned you know the fact that uh you
know universities are trying to figure out how to actually stay in the AI game when they don't have the resources you know to to to do all this compute well you know joining forces is one way to do it decentralization is a really great uh framework in which to uh to pull your your compute resources so I think AI uh is probably going to be you know if not the killer use case one of the killer use cases for web 3 and for decentralized Compu technology I think it's just starting to become a parent how all of that's going to kind of work
um and I'm you know I'm bullish on both but uh I think uh you know whether we get there because we're being idealistic about the way things work or just out of necessity uh we're going to need to tap into all that compute power for sure yes is there is there any sort of working or sort of tangible example of AI and large language model being transactional on a decentralized network or the blockchain at the moment uh I think you're starting to see right now um just the the beginnings of it so you you know some of
the um you know some of the the the precursor steps like preparing data and uh and kind of parsing through and getting data into the right formats are you know starting to to happen on on decentralized networks particularly the academic Community um and I think uh you know again with in trouet we do see um uh you know model training happening uh you know as truebit tasks uh those those very observable tasks we see the kind of the calls the inference calls to to R trained models happening as as true Tas so I I think we're starting to see it it
play out um are the likes of uh you know open Ai and uh anthropic and you know Google using decentralized networks to to train their models uh not now uh they don't need to and they they they have a closed model uh but again that's we're looking for The Innovation we're looking for The Innovation and the you the folks that don't have that that set of assets and resources I had a thought about a potential use case which is quite interesting again off off the conversation of that podcast yesterday
which was great um Lex Freeman and Sam mman round two the talk about the context the context awareness um and the tokenization length of a model and how that's at the moment pretty small um and what it's going to be like in the future but the idea about um referencing a from a contextual point of view referencing every single conversation that you've had with an llm so that you don't have to start again or so that every conversation helps helps it understand you and gets better so the way that I'm thinking about that is
where where would you store that in a way that you can't um you know have that um used for evil or you can't have that manipulated so how can you get your what you've got in here the convers that you have with an LM which knows you or might know you or you might be able to input all all aspects of your life you don't want that to change you need a validated where should you store it a good place for that would be a ledger that cannot be tampered with right um have I just come up with something great for the
future I don't know but I don't know I don't know you need to go talk to maybe talk to a VC about that idea it's interesting no I think I follow what you're saying um yeah yeah you you so so that's really interesting um kind of I guess call back to one of the you know the the key philosophical themes of web 3 which is privacy how do you uh take something that ultimately becomes extremely personal and and you know extremely sensitive as a result which is your you know your chat history with you
know over time uh you know if you thought search histories were you know revealing wait till chat history's become you know you know a thing as well how do you take that uh and keep it useful but also secure it uh where where would you possibly store that um other than on your own you know your own device in a way that can be used I think these are all questions that are very fair to ask of of web 3 um I think you know again verification verified compute plays a role in executing you know the the code from that um I think some of
the you know the things that we watch around uh secure multi-party compute uh in particular and uh and things like Z knowledge proofs provide you know key ingredients to help kind of keep the Privacy uh that you're you're thinking about in intact so yeah I I'd like to think that that uh you know we see that uh that emerge again um for that to be the you know the way things work it's hard to imagine that if all of the you know the AI is controlled by two or three companies that they'll be open to
that kind of model where they don't also control your your your history right so that's another reason why I think we're we're hoping to see uh something something else emerge not very open AI there you go hey a couple of minutes left maybe just give a little bit of a just a forward looking statement into you know where Tre bits going uh what's the next six to 12 months hold for you guys you know and what are you looking to effectively release and and get to Market yeah no that's that's that's
fantastic we are hyperfocused right now um so we are uh progressing really well in Early Access on truebit verify our uh our uh you serverless ver ified compute platform that we've been talking about um we will see you know uh by the end of this year we'll see that move into General availability um and then you know as we bring more uh developers more projects into Early Access and certainly as we reach GA you know I say we do two things one we we watch and we listen we're watching because uh a lot of the
things that we build on are truly bleeding edge um Concepts and uh whether it's cryptography Game Theory servess compute Etc um we're always looking at and giving ourselves space to uh you know to spike on on promising improvements and some of the the core technology we're implementing uh but the listing is probably even more important um you know we are looking for what makes a difference how do we actually uh help developers create transparency um make their applications uh as um you know as open as uh as they want them to
be uh so that there's never a Shadow of Doubt over what they're doing awesome good stuff I'm looking forward to keeping track of what trit are doing and it's it's a really cool world and I think you're doing the right work at the right time to make sure that blockchain is legitimized and is more than just thought of as a way for people to make money in Ponzi gam so you know I really love the work that you guys are doing absolutely thanks hey thanks again no worries so just as final WP if again as
a reminder if you like great things with great Tech would like to feature in future episodes click on the link on the show notes so head to jtw gt.com this has been episode 81 thanks to bla thanks to truebit and we will see you next time on great things with great Tech
Join the Email List
Listen On
Support On
