Synopsis: In this Techstrong AI Leadership Insights video, Ed Watal, founder and principal for Intellibus, explains the role blockchain platforms will play in ensuring only quality is used to train artificial intelligence (AI) models.

Mike Vizard: Hello, and welcome to the latest edition of the Techstrong AI Leadership series. I’m your host, Mike Vizard. Today, we’re with Ed Watal, who is a founder for an outfit called Intellibus, and they are a consulting and integration firm. And we’re talking about how blockchain is going to be used to help organizations negotiate with AI vendors, and I’m going to let Ed explain how that all comes together. But Ed, welcome to the show.

Ed Watal: Oh, thank you. I really appreciate it.

Mike Vizard: I think there’s clearly a need for some sense of immutability when we deal with AI and data. So what is the intersection between blockchain and AI platforms exactly?

Ed Watal: So if you think of blockchain and how it came about in the first place was really, it was a way for people to know that there was a validity of data and no one person was the guardian or custodian of the data. So kind of think of the days of music where Napster was a torrent and you kind of stream data from anywhere. Blockchain is in some similar sense, a distributed ledger. So the data is stored vastly on the computers of different people who are part of the blockchain network. So it’s kind of stored in a distributed fashion of the network, much like the internet. And because no one person is the custodian of the data, how do you know what part of the data is accurate? So every time a new transaction happens, you’re committed to a chain. And so you essentially go back and you look at, okay, is this transaction valid?
You have a concept of consensus and validation, so different parties to the transaction who are validators validate a transaction, and that’s how you sort of know whether a transaction is valid or not. And in essence, you’re able to do that because you’re committing transactions, like you mentioned, to a immutable blockchain where we now have a source of truth. And in the AI world, that is a challenge. AI today is being fed data off the internet. As we know, 90, 98% of the data on the internet could be duplicative or false or could have bias within it. So there is no way to really do source or truth attribution to the data that’s been fed to AI, and that’s the challenge. And so the intersection that I see is we need some sort of a data custodian model, essentially, that needs to evolve, and that’s the missing piece.

Mike Vizard: Do you think this whole shift towards AI is forcing people to come to terms with some data management issues that they’ve kind of ignored for a long time and all this stuff is now coming home to roost?

Ed Watal: Truer words were not spoken. I think data management is a problem which has been ignored vastly through the growth of the internet. And I wouldn’t say it’s been ignored, it’s just the rate of growth was so high and it was no one person’s job to really organize the internet. And so we’ve created a very, very ineffective and disorganized internet in some sense, because there’s a lot of duplication of data. Imagine it’d be just very unhealthy if you did that in your own hard drive or your own business, it’d be very ineffective to operate your business. The internet is kind of like that, and simplifying that is possible. It’s not impossible, but it takes a lot of effort. For example, think of just simply simplifying the encyclopedia information takes a army of people like Wikipedia, a massive community, and invariably then the question becomes if you do it from a nonprofit perspective, then it’s sustainable to a certain degree, but then how do you do it at scale? What Wikipedia has done for encyclopedias, can that be done for every last piece of data on the internet? That is a question.

Mike Vizard: It seems to me we had, shall we say, a lot of irrational exuberance around all things generative AI last year and now this year, everybody’s kind of wrapping their heads around it a little bit. But are we encountering challenges operationalizing this now because we’re trying to figure out how to embed generative AI into various processes?

Ed Watal: I think like any other technology, early technology always has its challenges in adoption because the tools of today are fairly effective and the tools of tomorrow that you’re trying to deliver may or may not be as effective in the short term. And obviously, change management is a big aspect to it. There’s be a lot of push and resistance internally within large enterprise and organizations. That said, the vendors and investors are doing a great job of pushing a lot of dollars behind this early technology with the promise behind it. So we do see, at least in the industries and markets and customers that we talk to, that significant shift is happening. For example, in an economy, in a market sometimes when dollars are scarce and people don’t want to spend a lot of dollars, they’re still spending dollars on AI.

Mike Vizard: How much of this do you think is a technical challenge versus a cultural challenge? Because a lot of roles are going to change. So where should people be focused first, on the technology or more of the people side of the equation?

Ed Watal: I think both. You can’t ignore either. You have to take both along, side by side. People in education leads technology because technology, you can’t really bring in technology. That’s just change management 101. So I think you have to take both along. You have to step in with people first, get people comfortable, understand what the guardrails should be, and then you bring in technology. A lot of organizations which sort of go headlong and try to catch up or be first in the race and try to just go really fast with technology invariably fail or falter because they didn’t put enough guardrails or people aren’t bought into the idea right from the beginning.

Mike Vizard: I think one of the ironies of all this is that while we talk about whether or not AI is going to replace people or not, it seems like we don’t have enough people with the right skills to implement AI in the first place. So are we looking at something that feels like an AI skill shortage right now?

Ed Watal: So given the current environment and the nature of things, so let’s unpack that question a bit. AI skills shortage in the sense of finding PhDs who know how to create a better LLM, obviously, there’s a shortage, because everybody wants to build the next best LLM and there aren’t enough PhDs around. So if you go and look at universities with AI programs, most PhDs are already working or being stolen away by companies to work on programs rather than continuing their research, with the promise that they can continue to do their research at the company because they’re there for a reason. So yes, there is a skill shortage when it comes to creating the next best LLM or whatever’s going to replace the next best LLM. But generally within the enterprise, it’s just a question of more retooling and rescaling and understanding how these tools work as you bring them in. And if you have really good engineers, so the companies with good engineering pedigree will certainly be leading the pack in whatever industry you’re in.

Mike Vizard: A lot of times we hear stories about organizations going out and hiring data scientists, but the data scientists don’t really know all that much about the business. So how do we kind of bring them in a way that delivers value? Because there’s a joke running around that says data science team comes up with a report that says we’ve noticed that sales dips every seven days, and everybody in the business side goes, “Of course, we’re closed on Sundays.” So how do we get to something where the data scientists are creating something that the business can actually appreciate and get some value out of it?

Ed Watal: I think the piece that, it’s interesting, businesses wanted to create better algorithms, and so they needed really smart people, PhDs, who would not necessarily want to come and work in an enterprise unless there was something really cool to do. So a very cool name was given into that work called data science, and we created something called data scientists, for a long period of time, used to wonder what that is actually, because in my mind, science is really talking about physics or chemistry or biology or doing advanced mathematics.
And in some sense, you could argue that data science is really about advanced math, statistics and algorithms. And so you’re right where you say that those individuals who have these advanced math capabilities aren’t necessarily from your industry. And if you were trying to get the best math and statistician, how do you couple them with the person who has the biggest domain expertise? And this has been the challenge in any change management of any organization. There’s always that handful of people that really understand your business, and then you need them to run your business for today. And that run the business versus change the business separation is what most businesses struggle with. And so unless you can really carve out the run the business versus change the business and really have, at the corporate strategy level, done this right in an operating model perspective, you can’t really yield returns, whether it’s data scientists or AI or whatever you do.

Mike Vizard: One of the issues you hear a lot about is data management, as we discussed earlier. And there’s these folks called data engineers that are running around. Is there enough of those folks? I feel like in some ways, they are the unsung heroes of this whole AI movement, because they’re the ones that actually get the data prepared in a way that’s consumable.

Ed Watal: Data engineering is an interesting topic. And if you think about it, should we need data engineering in the first place, if your data was very organized and structured in the first place? The reason we need a lot of data engineering today is because data is locked in silos. And if you think of silos, you can think of apps where data is bound within the boundaries of an app or websites where the data is bound within the boundary of a website and so forth. So if you were to rethink the internet or if you were to rethink an enterprise, you would think of a common ontology, common dataset, common data model, a simplified data model where this data was housed centrally and available in a central data model. And I’m not talking about centrally in terms of physical storage because there’s this question of reliability and scalability, but really from a ontology perspective.
And the word that I think that needs to sort of come more to the fore is the idea of a simplified ontology that connects and binds all this data. And that’s, I think, a big missing piece. We talk about policy, we talk about regulation, we talk about standards and ethics, but at the world stage, there is no conversation about ontology. It happens in bits and pieces, but in organizations, but there is no word stage for ontology. And so one of the things that I’m trying to do as part of a not-for-profit effort called, sort of bring that awareness, that policy standards, regulation, ethics, these are important, but then you tie them all together with the idea of ontology.

Mike Vizard: So what’s that one thing you see organizations do as they start out on this journey that just makes you shake your head a little bit and go, folks, we can be smarter than this?

Ed Watal: I think it’s really rethinking your strategy around your data. In essence, most organizations are still trying to build point to point connections between systems. We live in a event driven world. When is the last time you actually made phone calls? If you’re a teenager, never. You just send text messages. So the world as we see the next generations, the Gen Zs and others as are coming up, they’re more living in a message-driven world. And we can see all of social media is just messaging. And so systems and enterprises still talk through APIs, service-oriented architecture, which was supposed to be a very modern thing that came out after the mainframes, but that has already been superseded.
You do need APIs in some sense, interfaces, but those APIs are no longer rest APIs at scale. They’re becoming more and more event-driven APIs, message-based APIs. And so if enterprises are to rethink their future, they want to think of decoupling all their enterprise systems as just individual apps that communicate through what back in the day used to call it the enterprise message bus. So it’s sort of the revival or resurfacing of that concept and that idea, except with the unified event-driven models. So think of if you’re running a business and you don’t have a business event model to operate your business and you don’t understand what the business events are that are emitted by one system and consumed by the other system, and what is that simplified data layer that processes this, then you have a gap.

Mike Vizard: All right, folks. Well, you heard it here. If you want to master AI, you got to master data. And of course, the model is only as good as the data that was used to train it in the first place, so it all comes back to data management one way or another, and that’s going to involve all kinds of new stuff, including blockchain platforms. Hey, Ed. Thanks for being on the show.

Ed Watal: Thank you.

Mike Vizard: All right. And thank you for all watching the latest episode of the Techstrong AI Leadership series. You can find this episode and others on our website. We invite you to check them all out. Until then, we’ll see you next time.