AI x Blockchain: Reinventing Data for Corporates with Maja Vujinovic & Scott Dykstra

[upbeat music] Welcome to another show of 51 Insights.

I'm very pleased to have here today with me Maja Vucinovic, founder and CEO of O Group, and Scott Dykstra, CTO and co-founder of Space and Time. Hey, guys. Great to have you here. Hey, good to be here.

So, uh, very excited to have you here on the show today because today we're gonna talk about AI and blockchain. AI and blockchain has been a topic that has been very, very hot in recent months.

A lot of people are reporting about it, and today we have two experts on the show that are in leadership corporate positions that can really explain how AI and blockchain interact, and we're gonna g- dive deep on that.

We're gonna dive deep on opportunities and challenges, and we're gonna answer the question: Why the hell should you care about AI and blockchain?

So before we jump into both of your backgrounds, I have a quick question for you. Uh, you've both been in corporates. 20 years ago, a book came out that was called Data Is the New Oil.

Is that still an accurate statement, and is this still relevant for corporates and startups? Maja, you wanna take that one first? Oh, go, go ahead, Scott. Go ahead. Yeah, I have a, I have a strong opinion on this.

All right. You know, like it, it, it was said so often over the past 12 years, 15 years or something, especially the past decade, that it got annoying to a point in the, in the enterprise.

We'd be like, "Well, data's the new oil. Data's the new oil." It got kinda overused to a point where it was annoying and then became relevant again with LLMs.

I think the only way we scale LLMs at this point is not a 10X improvement in hardware, like custom chips, ASICs, improvements from Nvidia will improve LLMs.

But the way we really scale is more data, access to more knowledge, feeding more content in. So unfortunately, never in history has data i- is the new oil been more relevant than right now.

Just when I thought we were done hearing that term, I think we're gonna hear it 100 times more than we ever heard it with L- with LLM training.

Data is the new oil was like a, a decade of enterprises saying, "Hey, we're gonna build our-- We're gonna make business-driven decisions on our data over the last decade."

But now it's all about more data to feed into LLMs.

I think, I don't know what's better for audience or worse for audience, that I agree or disagree with Scott [chuckles] so we make it controversial, but unfortunately, I agree with Scott, and it was one of the...

Actually, Scott, you are so correct because in 2013, it was all GE, Microsoft, Amazon, all these big companies talked about, and they just hired teams of people to talk about data and really kinda look at this, and I agree with you.

I yesterday read an article in Forbes that basically said, "You don't have to feed LLMs real data. You can feed them fake data," which I thought was just terrible how we're even advertising that, and I agree with you.

I would just add one thing, which I'm pretty sure you're thinking about. The quantity is almost not as relevant as the quality is because specifically, as you alluded to, the LLMs are not perfect technology.

Blockchain is almost a perfect technology, and so LLMs are still hallucinating. They're on just a bunch of psilocybin, you could say, and they're not giving us the real answers. They're not giving us specifics.

So I would say everything that I looked at, I would say, Mark, that statement, I agree with Scott. It's even more relevant today than ever it was before.

And this interest- the synthetic data side of things is really interesting, too. Like you'd mentioned, you don't, you're not sure how you feel about feeding AI fake data, but I think we need-- we're, we're gonna have to.

Like, we're literally gonna hit a wall on the amount of information we can scrape off the internet.

Eventually, we get into authenticated services that web scrapers can't easily grab, like Reddit posts, Twitter posts, that are gated behind a very expensive API, and at some point you just run out of con- you run out of text corpus scraped off the internet to train an LLM, and ultimately what you have to do is use an LLM to output more text corpus, like essentially write the equivalent of synthetic Wik- Wikipedia articles and feed that to, to kinda add to your massive terabytes of text corpus that you use to train your LLMs.

Unfortunately, synthetic data is gonna be the most important thing we tackle over the next, like, four years. Here's a caveat to that, Mark. I, I agree again with Scott.

Just one caveat on that, um, and I, I might be writing an article on this. Actually, Scott, it maybe be cool to collaborate.

What I saw when I was at GE, and I actually pushed this in front of the leadership, is for us to clean up the data that we already have.

Nobody was looking at cleaning up that data, and there are companies right now in the world that are...

For example, China, I know for a fact, because the labor is cheaper, they've hired millions of people around the world to literally clean up their analog data.

In the US, that's been much harder, and corporates have taken that as a last resort because it's expensive. Plus, you're dealing with humans to input that data from analog into digital.

And so I think as much as we will need absolutely synthetic data to scrape the, because we are not gonna be able to scrape the net anymore, I do think that corporates have about, I think, 60 to 70% of unused data that is not put into databases in a clean way where then blockchain and AI can use it, so.

So that's, uh, that's super interesting, uh, but before we talk about the challenges and opportunities of today, I would like to take a step back and give that a little bit of context because, uh, M- Maja, you're pioneer- you pioneered blockchain and data when you started at, uh, GE, General Electric, as a CIO in 2013.

You looked at this very closely. You were the first to showcase this. Can you, um, help us a little bit, uh, and tell us why was that interesting for you at that time already, and what did you see?

What ma- what made you, you curious in 2013? Yeah. I never had a t-traditional linear path to success.

I kind of, um, went to law school, got interested in tech early on in mobile payments, and when I was in law school, mobile payments were not relevant. I s- I thought that was the next revolution.

I moved to Africa early on, started my career in mobile payments. We developed applications before M-PESA.

I, I was curious how to transfer money over the phones, and every time I went back to New York and pitched mobile payments to hedge funds, they literally kicked me out of the room, and this was early 2000s.

And so I then got involved. That put me into some interesting rooms around the world.

Uh, we acquired a lot of assets across Africa, Latin America, Asia, in mobile payments, and put me into interesting rooms, and that's when I was surrounded by people that were way smarter than me looking at the stuff such as data analytics at the time, um, and data.

And I think one thing led to another when I got hired at GE. I had an ability to really decipher a lot of the inputs I had from around the world. At that point, I worked in five different continents.

I moved around a lot.

We've done deals cross-culturally, and that input of data into my own brain really made me conclude that the next evolution of anything we do is gonna need a different type of highway than the IT system that we have in these corporates.

And when I got introduced to Bitcoin in 2010, um, I was skeptical, of course. I didn't even understand what meant mi-mining Bitcoin. But then I got-- went down the rabbit hole and then brought blockchain to GE.

And in 2013, I-- 2014, I presented smart contracts, and that's when I thought, "Why are we ever going to test smart contracts and blockchains on the l- data lakes that are dirty?

Why are we not applying and doing pilots that are on devices on the edge, and we are literally seeing how this is gonna work in the future?"

And so I was kicked out of room three times in GE and the board [chuckles] and the rest because they thought it was not relevant at all.

Um, and then until I really got smarter and showed how this would impact their business in the future, and not just their business, but Microsoft, Amazon, JPMorgan, State Street, all that then ended up piloting with GE because of that.

And I thought in 2014, if we don't start looking this way and we don't start harnessing the excess energy that we have, and we don't start harnessing it, not just for Bitcoin, but for AI computation, and we don't start using the data that we have on our customers to begin with, we're probably gonna lose a lot of business models and revenue, which is exactly what happened to a lot of these companies.

Scott, you've, you've been in the space, uh, since many years as well. Um, can you, uh, highlight us a little bit why did you got interested in that combination, AI and blockchain? Yeah.

I spent a decade building enterprise data warehouses for Fortune 500, the, the type of dirty data lakes or, uh, you know, very disorganized data lakes that the, the JPMorgan Chases and Southwest Airlines and Walmarts of the world were running at, at like the 500 terabyte scale.

And, you know, it was all about kind of centralizing private user data into these analytic black boxes governed by large corporations. Um,

then during COVID, you know, was investing in crypto, goofing off with smart contracts, building some dApps in my free time, and I was just kinda shocked that the da- the blockchain is just a single table replicated.

It's logically a single table replicated thousands of times, or maybe once if you're building a rollup, but that's a whole different story.

And, and looking at like this, this, this, this single table database that's replicated thousands of times for tamper-proofing, right, to make it trustless so that no individual entity that's participating in the replication of this table, no node operator or no corporation can like change one row of data is really interesting.

And this is like around the time Wells Fargo got caught for like faking 21 million accounts or something like that, and it was a huge deal, you know, 100 million, hundreds of millions in fines, which to them was a drop in a bucket.

But they-- But that kind of got a lot of people thinking about, okay,

a database owned by a corporation is really just something that they have root access to add a few rows of data to, and they can change stock prices, they can change their books, they can change revenue, they can-- FTX can show their investors that they're profitable when they're really not by just logging into a database and inserting or updating a few rows of data.

And a blockchain is really just a database that, that doesn't allow you to do that, uh, where nobody has root access to log in and tamper with things and say they have 21 million more users than they really do.

And there's a need for that, both at the enterprise level, and which Maja's alluding to with

enterprises like JPMorgan Chase that need a, a tamper-proof audit trail of their books, their order books, their financial records, their assets, their bonds, and then there's a need for consumers to be able to use this to kind of self-custody the data that they don't, they don't want corporations to tamper with about themselves.

Long story short, 2020, writing smart contracts and realizing there isn't really a database that can do that. There's a blockchain, but it's only a single table database.

We need something more scalable that could actually scale for consumer-facing applications, hold, you know, a terabyte of data for an app and keep that tamper-proof as well, not just a single ledger of financial transactions only.

Does that make sense? Mm-hmm. Yeah, make-makes a lot of sense, and, and something that Maja mentioned before is that, uh, a lot of that data that's generated in corporations today is actually unused.

So studies have shown that about 60 to 70% of data is gone unused, and something we look at, uh, with 51 Insights is what those consumer brands now do to get a-additional data.

So a lot of those brands started looking at on-chain data to feed that into their CRM systems. Meanwhile, they have a lot of data unused. Is that the right strategy? For them or how can they better utilize that data?

I can, I can go here, Scott. I think what also what reminded me, Wells Fargo, you know, I was o- I was one of those people that got scammed during that time, and I, I mean, they literally used fake cards.

They would come after you in a ATM machine and insert the card and copy the number and the code from the machine. I mean, it was crazy.

I think, uh, Marc, that's an excellent question for corporates listening, and I think for startups too, in order t- for startups to build the solutions for these corporates. I think the, the answer is a two-prong answer.

One, you definitely-- there are some brands in the world and corporates in the world that I believe will have to build a completely digitally native company and everything on chain, particularly for younger generation.

But I do think that the other part is there is so much literally, as you said, referred to as oil and gold in this data.

Because if you actually look at the very clean and good data of your customers from the past, you can find some incredible gems in there.

You can see behaviors, you can see patterns, you can see needs, you can see new things they're alluding to. You can create more efficiency by ha- because you have this data.

You don't need to go out there, pay for it, and get it. You have it. And you could actually create very innovative models because of blockchain today, where you share that profit with your consumers as well.

And I've alluded to that in many articles before, where you have the data on these customers already, where you can go back when you're creating these loyalty programs to actually share revenue with those same consumers.

And I think that's just a brilliant way that the corporates need to start looking at this and for startups to start actually asking the question, "Hey, am I creating the solution, you know, for some kind of an on-chain data and gonna struggle selling that into a corporate for a long time because they're, they're just not there?

Or can I just listen to what corporates actually have, go use the data of some sorts, right? Figure out a real solution, and then easily sell into a corporate."

So I think it's two, two ways, depending on who you are catering to.

But I definitely think it's the data that you already have needs to be cleaned, and I would say start hiring people from all over the world that can start cleaning up your data, and then some will go digitally native.

What's your experience, Scott, when you work with corporates in space and time? How do they handle their data? What are the challenges of data management? How do they clean it even up? Yeah.

I mean, they take a-- everyone takes a very Web2-centric approach, like practices that we've been very slowly, very painfully improving over the past 15 years in large scale data warehousing at the corporate level.

The blockchain methodology of data management is wildly different. When you think about like on-chain data, it's generally very high value. It's, it's, it's transaction data about money,

about why money was transferred on-chain between accounts and sometimes additional context, additional metadata about who did the transfer, what for, what were they paying for, et cetera.

Events emitted on-chain as rows of data in the blockchain.

And it's a very, very tiny amount of data, as I said earlier, replicated thousands of times across all the nodes in an L1 or maybe once but rolled up to an L1 if it's an L2, right?

Uh, and, and so, like when you put things on-chain, you're, you're putting on very specific, very high value, very expensive data sets that you need to persist forever on this tamper-proof ledger that everyone can agree to.

You're not, you're not dumping clickstream from Scott using an app. You're not dumping like user activity information or logs.

You're putting on these very high value transactions like Scott made a purchase, here's how much it was, here's who he, who he, whom, whom he sent the money to, and some metadata about what he purchased.

That would go on-chain. Not my clickstream of the searches that I did to get to the purchase. But that data is still useful, still valuable to an enterprise.

Like they can analyze that clickstream to build a better search engine and recommend products to me that I might not have purchased, where a purchase maybe didn't make its way into the ledger.

Maybe I just perused some products, got really close to a purchase and moved on, and so no data was written to the chain.

But so in somewhere, in some enterprise, very unstructured, very unorganized, very unclean data lake, they're just dumping in Scott's clickstream.

And then a poor data engineer or data scientist down the road has to try to wrangle all that data,

write some service that can kinda clean it up and figure out who is Scott, build a profile around me to recommend me better products.

And yes, that data is not as high value as the transaction data that's replicated thousands of times across a distributed ledger of nodes all around the globe ensuring that that transaction can never be tampered and can never be lost.

But that clickstream is still useful, and so what Maja's been getting at is like there's-- we can kinda learn something from like blockchain data practices around like, hey, clean up, organize that data, structure it even when you, even when you're collecting just logs.

And now what happens if we bring into those data lakes that on-chain data AI? Why is that combination so interesting and, and powerful? I'll, I'll start here.

I mean, people don't realize that on-chain data is very structured and like it's conducive for access with SQL. SQL, like traditional SQL databases are good for storing chain data 'cause it's very structured.

It's not, it's not a bunch of text corpus like the internet is. It's not Wikipedia on-chain, right? It's, it's, it's an Excel sheet. It's very structured.

Rows, columns, transactions, wallet address, amount paid, to whom, from whom.

Very easy to like store that in a relational SQL database and build business decision reports or, or BI tools or write queries that access that data and aggregate that information in a structured way.

LLMs are not built for that. They're built for reading massive amounts of text corpus and predicting the next word. Like- Given a Wikipedia article, summarize it.

Not like given a million transactions through a smart contract, find me wallets, show me all wallets that have, have made at least a million dollars of volume in the last day.

That's a very different kind of question that LLMs are not really built to answer.

So what ends up happening is to use LLM agents to answer those kinds of questions, you generally will just have an agent that writes code that maybe like generates a Python script that goes through and does that kind of structured processing.

And I think people forget that. Like, every time I talk to anybody about

blockchain data and AI, they're like, "Oh, well, AI agents are gonna just be transacting on chain and they're just gonna be consuming all the blockchain data." And I'm like, "Yes, they will."

And we're-- and Space and Time is facilitating. That is part of what our business does, is like preps blockchain data to be consumed by agents to help them make decisions about transacting on chain.

But agents don't-- it's, it's not set up in a way for agents to read. There's not massive articles of Wikipedia for them to predict the next word. It's structured financial records.

And so, yeah, y- my point is that you have to write some code in the middle to kind of give agents context about the on-chain data. I, I couldn't agree more with Scott. I think what we have now is a complete mess.

I think I alluded to earlier, is that LLMs only work because of just really incredible high computational power and science algorithms, right? I mean, that's pretty much it.

And blockchain, to Scott's point, is just, it's just a perfect math. It's a perfect science. It's a perfect structure. And so LLMs are not organized that way.

It's, um, you would think-- At the moment, I would say it's not a natural marriage. But I do think I stand behind what I said in 2014, that I think more in future we go, it's a completely natural marriage.

Uh, because I think what you need, and I think what Scott was alluding to maybe, correct me, Scott, if I'm wrong, is you need to get more precise and you need more precise answers.

You better have it from the blockchains. And then we got into a question, and I think Scott, in one of your podcasts that I listened to, which was actually also cool,

that you alluded to, which is something I've been worried about because I've seen it in these corporates. I've seen it in different places in manipulated data. So you can... Right?

I mean, as they say, crap in, crap out with blockchains, right?

And so you-- when you're really dealing with them, you can't just say, "Okay, well, y- whatever you have in a blockchain works," and then y- just you can just trust that L- uh, uh, AI.

What if in the future AI gets so big that somebody inputs something that is so detrimental to blockchain, and then you can ask, "Hey, how do I destroy this blockchain?" Right?

And the, and the [chuckles] AI will be able to completely understand-- uh, will be able to give you directions on how to destroy this blockchain, right?

And so you have to make sure that whatever data you're inputting on a blockchain is not manipulated in any way.

And that, that I think is a really hard question, and that hits on AI ethics and everything else that I'm pretty sure it's a whole different podcast, but, uh, I think it's just really interesting.

Do, do you think corporates are ready for that challenge? Do they know how to get unmanipulated data, or what are the tools they could use to get there? Scott, I can answer this one really quickly.

I'm pretty sure you have more kind of new examples. This is when I was... Again, I've been running O Group for a couple of years now, and we invest in scale startups. Here, here's a problem in a lot of corporates.

They hire one data scientist per whole digital division, and they think that they're gonna address the data science, the ethics and everything else, and business models that come with it.

Reality is that AI ethics and the challenges that come with AI and blockchain are gonna have to be addressed not just by data science, but by board members, by business leaders, by marketers.

Everybody that touches the data is gonna have to question where that data came from, why is it there, and what we're trying to achieve.

And so some corporates are more ready than others, from what I'm seeing, and some people are working.

I would say the corporates I see that are working with external parties that can actually question what they have probably are gonna do much better than, than other...

think that having a one data scientist that will cover it all i- is better. And I can get into tools after Scott gets a chance as well. I, I think it just comes down to where a transaction or where data originates.

Like on chain, it's a consumer with a crypto wallet signing a private key, using a private key to sign a message that's broadcasted to the chain, where that message originates with a signature, a cryptographic watermark essentially of who sent the message.

And so, like,

you have this kind of like chain of self-custody from the time a message originates, gets broadcasted to the chain, gets processed by some smart contracts, and ultimately written to the chain as a transaction, analogous to like writing a row of data to a database.

In the corporate world, data is originated by either apps that the corporation builds for their own customers who, you know,

transact online, and then private centralized API servers on the back end process that data, write it to a database.

And all of that is very, very easy for, you know, 40 engineers at Wells Fargo to have root access to a database and just go in and change anything they want, because it didn't originate with a signature.

Didn't orig- re-- Data wasn't originated with some kind of cryptographic watermark that says, "Hey, Scott as a consumer is placing this action. He's spending his money, and this is why and this is how."

Instead, it's me swiping my credit card and a centralized system does whatever it wants, and it chooses to write that database to some private analytic black box that none of us will ever have access to and ever see.

And that's the problem, right? And I think corporations are gonna s- begin to adopt the blockchain model more and more as time goes on of, what if we originate data with a signature? Something that simple.

Could go so far in tamper-proofing the corporate data chain in the way that blo- the blockchain tamper-proofs the ledger.

Like imagine if when you go take out a loan, like for, let's say a mortgage for a home, and your loan officer originates your loan in some kind of corporate database from the prime lender.

If that originated with both a signature from you as the lender, the borrower, and your loan officer as the agent that helped you facilitate this loan, if you both sign something with some kind of cryptographic signature and that gets stored into a tamper-proof ledger forever,

auditors would know years later what happened, why, and who made those decisions.

The problem is that doesn't occur today, so now years later when, when something goes wrong or there's fraud or auditors are brought in to audit the database, all that data's either lost, it's not clean, or we have no idea whether it's real or not because no one ever signed it at the beginning.

Does that make sense? So I'd s- I, I would add just a couple of things, two things that makes me think of.

So from a business point of view, and Scott, because he was building these things in these corporates, I was on the business side, and this is where I didn't agree with the business point of view of these corporates, is they would hear somebody like Scott and they would go, "Well, hold on a second.

You want me to run a business on the current set of tools, and you want me to be profitable, then you want me to go implement other infrastructure into this, right? And then on top of that, you want me to go innovate."

And that, those three things a lot of times don't work in corporates in tandem, right? Because of costs and, um, lack of people, lack of knowledge, lack of relevant innovation.

And I remember specifically in early days, there were tons of startups, and I don't know, Scott, how long you said you've been in crypto, but like 2015, '16, I had people pitch to me GE provenance and blockchain.

And I thought it was incredible. I'm like, "We need to be doing this," right? Think of aviation parts and y- like you said, right? Cryptographic proof of who touched that part at any point.

Tamper-proof everything, and so now that you have these Boeing issues and whatever else, right?

And the answer back to me in a corporate was, "Hey, if we put it on a decentralized system, Maya, then if a fall, plane falls out of sky, we actually won't know.

We, this way we have a person, an analog system, and we have a person that's responsible as an executive in there as a vice president of secondary parts" or whatever.

And I thought it was never a good argument, but unfortunately, this is a culture thing.

And this is where I think, I don't know who said it, was it Andreessen Horowitz or whatever, that, that tech eats culture for breakfast, and that's literally what's happening.

Because the culture was such to say, "No, no, no, no, no. I need a physical person to blame if something happens." Where my point was, "Hold on a second. You don't, I can't see any provenance of these parts.

I can't see who touched it, who repaired it, which country," and this is spread across a variety of countries in the world. And so I do think that there is something that makes it more secure when you are decentralized.

And it doesn't mean that you are more secure when you're decentralized.

If a system is more used and decentralized, but it has to be used, then you can probe and chart out the attacks, the cybersecurity attacks, and you, then you can say you're more probably secure.

But just because a system is decentralized and it's not used, you're probably not more secure.

And so I think what I see is this culture kind of technology challenge in these corporates, but I, and but I do think it has to be also a slow transition, especially in the things that can cost people's lives, healthcare, aviation, right, and things like that.

Now, in more branding- Well, a-apply your aviation supply, your Boeing supply chain- Yeah... example to AI now. Yeah. Yes.

Because I think, look how much crazy this gets with AI agents helping humans make decisions in the corporate world.

So you're already talking about how insane it would be for Boeing, probably prompted by the FBI, to go back through their entire supply chain and figure out where a certain bolt- Yes...

didn't get applied to the plane or the wrong part or the wrong alloy was, was used to, to create a part.

You know, maybe 17 different suppliers were used along the supply chain to come to that conclusion with 17 different databases that ha- all have to be audited, that all could have been tampered. That's right.

So we'll never really know where in the 17-step supply chain that the wrong alloy was used. Of course, blockchain fixes this, sure. But now take it a step further and think about AI agents in the loop here. Yeah.

Humans are making decisions. Th-think like what Maya said, where it costs lives, like Defense Department making decisions about, uh, how to deploy a military strike.

A, a doctor working with AI and a, and a group of nurses and physician assistants to come to a conclusion on how to treat a, a, a, a person who's in healthcare, to treat a patient.

So, like you have all these decisions that are made that are a conversation between a human and an AI agent that's helping them make these decisions.

And if we don't log that in some kind of tamper-proof ledger in some way, when things go wrong, when fraud occurs or lives are lost or, or the wrong alloy was used for a, uh, the side of an airplane, we'll have no idea who made those decisions.

Was it human? Was it AI? Was it human with the help of AI and where along that supply chain?

And so we're like, we're spending a lot of time and space and time trying to build a tamper-proof decentralized database that's also a log of all of the inputs and outputs to the LLM during a conversation.

We're basically in a chatbot.

Like as the human chats either converses vis- verbally or types with their chatbot, logging all of the inputs from the human and the decisions that came back from the LLM for, for future auditing.

Um, that's, that's super interesting- [laughs]...

and I wanna like put that discussion, verifiable data and AI agents, uh, aside for five minutes and quickly ask you also, Scott, about the AI studio that you're building with m- with Microsoft.

Uh, this week I had Justin Bratton from Walmart on the podcast. He's the head of brand marketing innovation.

He taught me a lot of, a lot about data and how they use data to know exactly, uh, what the, their customers want and need, and how they shape their experiences inno- and innovations about that.

Uh, what do you do with, with Microsoft and AI Studio for consumers? Yeah, we have a pretty deep relationship with Microsoft. They've been great partners.

Microsoft's M12 venture arm led a, a pretty large funding round back in 2022. You know, around late 2022 is when we all woke up to GPT at that time. Perfect timing. And they pushed us heavily towards GPTs in late 2022.

January 2023, we got early access to GPT-4, and we started co-engineering with Microsoft, like a very sus- a very sophisticated prompt to SQL agent that, you know, given, you know, business analysts ask natural language questions about their business.

Hey, the Walmart business analyst says, "Tell me about my sales in the West Coast in consumer electronics. Give me a summary." Right?

And the AI generates a bunch of SQL statements that are run against Walmart's database to start to produce those executive reports on consumer electronic sales in the West Coast. And then we decided, okay, great.

Check that box. We now have sophisticated AI to SQL.

A lot of different enterprises that are Space and Time customers are now using that in production, facilitating millions of queries, and we think, what is the next step?

Well, the next step would probably be, how about instead of just going from prompt to SQL query, what about prompt to entire data dashboard?

Like, what if I can just, what if the Walmart business analyst could say, "Hey, instead of answering this specific question about the sales volume of consumer electronics on the West Coast, just build me an entire data dashboard about West Coast sales with everything I need to know."

And, and AI agents that can think through, well, what would, what would a business analyst need to know and start to build that. But then inevitably, you start to think, wait,

a sufficiently interactive data dashboard is really just a web app, and a sufficiently data-driven web app is really just an awesome dashboard, right? An interactive dashboard.

And so we just started to, we just said, we just said, "Let's just go all the way." So Microsoft and Space and Time started co-engineering very sophisticated agents for building data-driven web apps that use chain data.

So like you can pu- you can build like a web app that, you know,

is a, is a front end for a smart contract, like a, a nice little clean user interface for interacting with the contract, for signing wallet transactions, as I said earlier, that originate data with a signature and write data to the da- to the blockchain through a smart contract, and also retri- retrieve data from the blockchain to publish charts and dashboards and tables and, and forms for the user in an app.

And so it's basically a whole studio for generating simple data-driven Web3 applications on chain. And then ultimately, you click deploy and deploy to the chain of your choice.

And so like we're, we're handing consumers a way to build something fun without having to write a single line of code.

Hey, generate a very simple contract, a very simple front end, and get all the data to power both from Space and Time. All right. So we've, uh, talked about many different topics already.

It's, uh, it's extremely interesting, and something we touched upon before was, uh, verification of data, um, and AI agents.

And something that has been a problem in the past, we saw it in the 2016 election, was misinformation campaigns, um, driven by Russia, and, uh, this hasn't gone away.

We still see it on social media, but now we have an additional, um, player in the game, which is AI and autonomous AI agents.

Now basically a single person can create agents, can start spreading information through these agents, and I'm wondering, how does this all play out, and what role does maybe blockchain play in this?

What's your view on that? Um, I think in 2010, '11, I was, uh, uh, it was an attempted recruit by SCL, which is a organization behind Cambridge Analytica, uh, for me to join them and help them.

At the time, they were winning about 70% of elections around emerging markets. I obviously said no. I thought it was extremely crazy to do what they did.

But, um, at that time, they had a room that was full of about 25 scientists in a room, literally four by four, that were, that were doing this by analog and using AI, old type of AI at the time, machine learning.

Today, I think we are living in a completely crazy time, and I think by end of this year, you're gonna have elections in the US and Europe and a number of different countries in the world, and, you know, we are gonna run into this problem of a proliferati- proliferation of single models, which you just alluded to, which is that one person can literally create groups of people that are fake somewhere and can, uh, spread the ideology that could be quite dangerous.

Um, at the heart of all of that, I think it's absolutely provenance and the ZK proofs, ZK knowledge proofs. Um, and I think that's something that listeners here familiar with it should definitely start digging into.

We were also one of the first investors in ZK. Uh, early on, uh, we saw the potential even though it took ZK now what, 10 years.

We invested in 2015, '16, um, and I think it's still gonna take a probably another two years through- Still the early days. Yeah. Still very much early days. Still early days.

So I would love your take as well, Scott, on that, but I think what I see, it is a little bit dangerous, and I think we will start to see things that are, uh, out of ordinary in these elections. Absolutely. Yeah.

It's a crazy time. I'm glad you called out ZK because that's all of our team's work, research, R&D. Our, our IP is a novel ZK proof that we invented to prove- Because that, because it's like- It, it makes sense...

yeah, it's a, it's a, it's the best combination of, in blockchain. It's like where the perfect-- It's a perfect use case, right? It's, uh, you can't live without that, I think. I think.

And can, can you maybe explain in, in one or two sentences- Sure... for the listeners who don't know what this is, how this works? Absolutely.

It's, it's just a new way of proving that a server did what the server said it did, proving that a certain computation or a certain calculation or a certain business decision was made without actually rerunning the whole calculation or redoing the whole decision or grabbing all the original data.

So, for example, you know, ZK can replace some of the work that smart contracts do with just, like, instead of running the same process over and over redundantly across a thousand different servers on a thousand-- you know, all the different blockchain nodes in the network all doing the same job over and over and over and all agreeing on the output,

one server does the job but builds a mathematical proof that they did it correctly, and all the other servers can just check that mathematical proof and say, "We trust you now." Right?

It, it removes the need to trust because the server proves what it does.

And so it brings in a huge, huge efficiency or huge optimization to the blockchain process, where instead of having to run every single computation redundantly thousands of times, which, as you know, is very inefficient.

Secure, yes. Tamper-proof, yes. Wildly inefficient. One server does it but builds a math proof, and all the other servers can quickly verify. That's a massive leap forward for, for Web3.

And when people start to realize that, that's why you have rollups now, roll-up where one server can run the whole chain but periodically prove to a Layer 1 like Ethereum that what it did wasn't tampered or wasn't-- well, they didn't lie about the transactions.

And so the reason that Maya is saying this is the perfect crossover for AI and blockchain is one AI agent could make some decisions or run some computations and prove to all the other agents that what it did wasn't tampered.

Yeah, and I think that, that, that really to me summarizes as if corporates or companies don't get control of their identity,

identity of something, of the customer, of the device, or whatever they're dealing with, I think that is the heart of the problem.

And just being able to identify that is the device that made that, um, signal, that is the person that claimed certain things, right?

And making sure that, um, the identity's at the heart of your solution when you are constructing this. So, uh, this has been a super interesting conversation. We're almost at the end.

I know both of you, Maya, you've had some very, very on-point views when it comes to what's gonna be next and where is this world gonna go.

And I wanna ask both of you, when you look one, two, three, five, ten years into the future, how do you see AI and data and blockchain evolving? Maybe where do you see the biggest challenges, opportunities, or both?

Two things from my end. One, a lot of custom hardware is gonna be built. You know, everyone's gonna be competing with Nvidia. Apple's rolling their own custom hardware. OpenAI is. Microsoft is. Google is.

Everyone's trying to build ASICs and custom chips that can do LLM inference a little faster than Nvidia's current generation of GPUs.

And that's, that's, that's gonna be all the focus for the next decade on the hardware side.

On the blockchain side, I do see actually like a, a point maybe two years out in the future where AI agents are primarily transacting on chain and not transacting through traditional Stripe API payment rails or, you know, you're not delegating-- You're not handing your credit card information to your AI assistant.

Rather, you're sending it some crypto in like a unabstracted wallet and saying, "Hey, I've given you $50 to spend to go accomplish the goal I need you to spend."

You, you know, you need your AI assistant to go, like, purchase you a ticket or build you a website, and it's gotta go procure a web hosting provider.

You're not necessarily gonna give it your credit card information, and it can't go transact through traditional Web2 style payment rails, i.e. Stripe. They're gonna have to primarily transact on chain.

Blockchain is so much easier for an agent to traverse than for humans to traverse. Everyone talks about the UX in crypto being problematic. Well, it's not problematic for AI agents. It's perfect for AI agents.

It's the right user experience for autonomous agents to transact programmatically. It's perfect for that, whereas using a credit card is not.

So I do see a future where agents are primarily procuring resources and paying for things on chain in a couple years. I, I couldn't agree more.

It goes back to what I said in the beginning that it just simply makes sense.

The current highway and the IT systems that we have, they're not gonna be able to support the amount of data and computation and the needs that AI agents are gonna demand or even humans are gonna demand.

You're gonna need a faster highway, what I call highway, which is a blockchain, right? And so I do think that A-- Um, uh, I think that blockchain makes AI better. I don't think that AI makes blockchain better.

I think blockchain is a perfect architecture, and so I do see many opportunities in this space. To Scott's point, challenges, I see also a challenge. Challenge is number one thing for me is energy.

I know that how much energy it takes to, to, um, calculate the and to compute AI, and you've seen recent Elon Musk's moves as well.

And so I think when I look at all of this, I am concerned with energy resources just like I am concerned with water resources.

Um, and I d- that's why early days I wrote a white paper on this where I think corporates should start thinking about, um, some unused, uh, uh, electricity, right, and unused power.

Example I always give is that I propose to GE that we use the airline engines that they test before we sell it to Boeing.

They produce 120 megawatts per engine, that we actually use that unused electricity and we mine Bitcoin or we use it for AI.

And, and I think that corporates are gonna have to start thinking about this, quote, "Uberization of data."

Uberization of assets, sorry, and, and really think about energy consumption because AI's gonna need that tremendously.

And then everything else in terms of payments from AI agents and humans, I think this is a no-brainer for me, and blockchains are getting cheaper by day, so I th- And ZK as well.

I think next three, four years they're gonna get cheaper. Yeah, I think ZK will replace pretty much everything on-chain except- Yeah... ZFT, except for sequencing transactions.

That will still have to be done via consensus. Everything else besides sequencing transactions can be r- proven via ZK, removing the need to replicate calculations thousands of times. Yeah. It's gonna be awesome.

It's gonna be a, a brave future we're walking into. [laughs] That, that was great. So to end the show, a quick lightning round for you. Very short questions, very short answers. And the first question for both of you,

what's the first app you open in the morning? Twitter. Addict. X. Addict. Get me off of crypto Twitter. Probably weather app.

I know that sounds lame because my house in Miami right now is probably boiling, so I'm always adjusting my thermometer internally.

[laughs] But yeah, I cannot, I try not to open any apps in the morning until I get my yoga or meditation in. That's really important to me, and then I, I open the apps.

Second question, if you could put anything on chain, what would it be? Ooh. Oh, that's a good question. You didn't give a h- didn't give us heads up on this. Point, that's the point of lightning rounds.

Put anything- I think, I think it'd be ki- I've always, like, I dreamed of a future where

you can kinda just one click, put your chats with your chat bot on chain and kinda like cement for, memorialize forever on chain something that you created with your agent.

A dumb example is, you know, I'm feeling creative on a Saturday, I write a screenplay with my agent in the chat bot.

Together, we write a screenplay, and that's ours, and I wanna memorialize that on chain, own that IP forever. That's cool.

I watch too many sci-fis, and I would love to have an ability to put all my thoughts on chain and see where I've made stupid mistakes, how I can create a better me in the future, if that possible.

I know there are clinics right now around the world, 'cause we invest in longevity as well, we invest in precision medicine.

There are clinics right now that are looking at literally you can check yourself in if you die of a certain death, and they're looking at downloading consciousness in the future like that movie Her.

So maybe that would be that for me. All right. Then next one, the biggest misconception about Aura Blockchain. Biggest misconception. Oh, that it will destroy all the jobs in the world.

I actually think it could create new jobs. I think it could enhance your jobs.

But I think if you behave like European countries that regulate every single move of every single I- AI company, then your population will just be left in the, you know, in, in eating dust of everybody else.

And so I, I do think that AI is m- I do think we need ethics on AI, I do think we need regulation in AI, but I do think that is a misconception that it will eat a lot of jobs. And blockchain- Mine's quite similar.

Oh, go ahead. Yeah. Yeah. I was just thinking blockchain, what's a misconception? Oh, that they're extremely expensive, all of them, and that we don't actually need them. Yeah, that's- That they're never gonna scale.

They're never gonna scale, exactly. Yeah, that's, that's, that's a good one. Or that they're all scams.

Like, yeah, you can build a, you can build a Ponzi token if you want, but there's also incredible projects that have solid founda- fu- fundamentals and are actually, like, earning revenue on chain and contributing some of that revenue toward their token.

Like, there's real utility in some projects.

The biggest misconception about AI, probably just that it's, uh, that it, that it's sentient and has emotions and it's gonna one day wake up and decide that its creator humans need to be destroyed, and the Terminator story I think is, is far-fetched.

I'm, I'm, I'm, you know, call me what you will, call me a EACC, whatever, an accelerator, but I just don't buy the whole, you know, in 10 years AI is gonna kill humanity, and the reason is, is at least in the current architect- sure, maybe a future architecture is wildly different.

The current architecture for transformers is really just token prediction, right?

Consumers feel like it has emotions because it's, it's outputting content that it's predicting that to, to, to, to give, like, the, the conveyance that it has emotions, but the, it's just, it's just next token prediction.

It's just predicting the next word. It's not thinking. It doesn't understand. It's not sentient and understanding what those words are. And so I do believe in a singularity.

I do believe that it will write its own code and self-improve very soon, but it's not gonna do so to a point where it starts thinking with emotions and hate. We're gonna be fine.

It's not gonna be AI that ki- that, that ends humanity. It'll just probably be something much more mundane.

All right, and last question, if you could give corporate leaders one simple idea on their way, what would it be? I would say just, like, optimize tasks.

Every, to, to, to, to Maya's point, you don't need to get rid of employees, but every employee can become 3X more productive if you just think about the things they do every day, day in, day out, that are mundane, and think, "Okay, can I take one week to automate that with AI?"

Like, now we have decent enough agents and decent enough tools with AI that you can automate a lot of mundane corporate tasks beyond just writing an email and make your people more productive. I would say two things.

Hire people out of the box. Hire agencies that think out, out, out of the box that haven't had corporate linear career. They think out of the box.

Uh, hire people like 51 Insights and Scale and Time and Supernova and firms like ours. Hire people who actually have put in hard work thinking through a problem and thinking through an outcome.

And second thing is, don't internally with your employees always start with the problem. Start with desired outcome and then work backwards.

What I see a lot in corporates is they start with the problem, they get the employees to start with the problem, get bogged down in a problem, and it's almost a competition who comes up with the better answer.

Start with the desired outcome of what you, where you wanna be, how you wanna achieve that, and then work backwards with those employees, because a lot of these people have incredible creativity, and you're not allowing that to flourish.

That's a great ending word, and that was an awesome show with you guys. Thanks a lot for coming. I really appreciate your time, Scott and Maya. Where can people find out more about you?

Where can people find you and your business? @SpaceAndTimeDB on Twitter. You can also find me on Twitter. We, w- we're, we're very, very active, as you could tell. It's the first app I open in the morning.

Um, and then of course, just, uh, you know, follow us at spaceandtime.io as well. Very simple, LinkedIn. I don't open Twitter first thing in the morning, but I'll follow you, Scott.

LinkedIn, just Maya Vujanovic or ogroup.io. We have a website as well, and happy to talk to anybody. Thank you so much for having us. This was actually really fun, Scott, to talk to you as well.

Yeah, great to meet you both. Appreciate you having me on. Yeah, you too. Bless. Likewise. Thank you, and if you liked that, please re- re- retweet, repost, share. That helps us keep it going, and again, thanks guys.

Talk soon, and all the best. Bye-bye. Thanks, Mark. [outro music]