Episode 37: AI and Crypto with Flipside Crypto CEO Dave Balter

Host Rob May interviewed Dave Balter, CEO at Flipside Crytpo and Venture Partner Emeritus at Boston Seed Capital. Tune in to this episode on AI and Crypto to hear Dave's insights as both an executive and an investor about things like how he is thinking about AI, the intersection of AI and crypto, what he does to distinguish who is legitimate and who is not, and much more. 

Subscribe to AI at Work on iTunesGoogle Play, Stitcher, SoundCloud or Spotify

Rob Circle Headshot

Rob May, CEO and Co-Founder,
Talla 

Dave Balter Headshot

Dave Balter, CEO,
Flipside Crypto


Episode Transcription   

Rob May: Hello, everybody, and welcome to the latest edition of AI at Work. I'm Rob May, the co-founder and CEO of Talla. I am your host today.I am here with Dave Balter who's the CEO at Flipside Crypto and a Venture Partner Emeritus at Boston Seed Capital. Dave, welcome. And first, why don't you give us a quick overview of what you do at Flipside and then what you did at Boston Seed Capital.

Dave Balter: Excellent. Thanks, Rob. So yeah, Flipside Crypto is in the business of helping understand the fundamental health of cryptocurrencies. If you've done any investing in crypto, most of it has been produced off of traditional price metrics, market cap, charts, everything that people would do to evaluate a company on price. We look at the fundamentals of a crypto project, so including customer activity, developer behavior. And we provide those insights through a scoring system called FCAS that lives on The Street, MarketWatch, CoinMarketCap, a bunch of other sites. We sell tools to crypto projects to help them understand their health and to investors to help them make good investment decisions. So that's Flipside.

Boston Seed, venture capital firm here in Boston, seed stage, still a part of all the funds. I’m there, as a partner, focused on investments and all the fun stuff that Venture Partners get to do.

RM: Let's take the conversation in a couple of different directions. Let's start digging. We have had a crypto guest on yet. As this is an AI-focused podcast, is everything that you're seeing at the intersection of AI and crypto that's interesting? As you live in the crypto world, with respect to Flipside, when does AI come up?

DB: There are a whole bunch of projects in the crypto space that do either have AI in their name or somewhere in their philosophy of what they do as a business. There are all different types. There's ones who are working on ad technology. If you think of any business service. Maybe the perfect scenario is you combine crypto with AI with some other form of machine learning, and suddenly you've got to say that everybody wants to pay attention to.

What I would say is, this is being a bit forward, almost all crypto projects are still extremely early in their scope and scale. Even the ones that are developing an AI solution or other things are still pretty early. By early, to us, are they growing customers? Are they retaining customers? People ask all the time, are there companies that are driving revenue? Are some of these customers? No doubt. But, I wouldn't yet say there is any perfect blend of AI and crypto, and this is the company that you should be paying attention to yet.

RM: Gotcha. When you look at trading in crypto and some of that stuff, are there people using AI algorithms the way they are on publicly traded markets? Or does that even matter or come into play at this point?

DB: No, that is certainly a spot. One of our tool sets provides investors with a spreadsheet to evaluate every crypto asset by its fundamental activities. Many of the folks who use that tool are from the traditional finance world and maybe spend a long time building quant models, AI models. We get called about many things they're working on in that world. So I would say if you probably parse the space, on the investor side, there are certainly people trying to apply some of the similar technologies they might have in a traditional finance world. On the project side, they're still figuring out how to put these two things together.

RM: You've done a lot of investing, both personally and through Boston Seed Capital. So you were also involved during this transition period, when people started saying they were AI companies, and then everybody said they were an AI company. And like, what were some of your observations there? As somebody who works in tech but doesn't come from a deeply technical background, what did you do to try to distinguish from who was legitimate and who wasn't?

DB: I think AI is one of those terms a lot of people partially understand. There's a few people who really understand, maybe like you guys. But there's a lot of people who think they have a view of it and they kind of have their hands on it. Is it Alexa? Is it self-driving cars? What is AI? Where does it go? I would say during that era, during the heavy investing era, most of what we would try to get to is is there some technology that it's going to make the business more effective at delivering the value it's trying to drive?

I probably would say I'd sit in meetings and hear the word AI and go, yeah, OK. But like, walk me through the business. And whether you call it AI or machine learning, are you confusing the two? It seems like you're using those interchangeably. I'm not really sure they're the same, maybe.

I would say, maybe where we are now, we've entered an era where there are actual application of AI that people can touch and feel to a point where they're like, oh, if that's AI, I get that. I get my Alexa. Like, I get what that does, and there must be something cool behind it that's making it work. But I still think there's just a massive world that has yet to be tapped that's going to open up.

RM: With respect to the tangible stuff that you're mentioning, so you and I are both angel investors in a company called LinkSquares. That was a company that started with no AI as sort of a sales contract workflow tool. And as they spoke to customers, they developed this view that like, wow, customers are trying to sell something. And AI, particularly natural language processing as a piece of AI, is how we solve it. Do you want to mention what they do, where they ended up?

DB: Vishal Sunak, who is the CEO who we both know, the company basically looks at large corporations, contracts, their legal agreements at scale, and is able to identify trends and patterns between them. You know, I want to know every contract that says it's got an indemnification clause or any that has a 30-day out. If you're in a large organization, with millions of pages of contract, that's hard to do. And there's all sorts of times you might want to have that data on hand. They figured out how to do it.

What I find fascinating about LinkSquares is in the beginning Vishal was like we got to look like we have AI because that's cool, and people will want it. And every client who showed up, they would sort of say, yeah, we can do this stuff. It's there. And the client would say, well, show me what you do. They would do it.

They didn't really have AI. They had a bunch of people doing a bunch of really good work, and that built the business. Over time, they've actually began to really develop AI tools. And now it's become almost the inverse. People are like, we just want to make sure you know what you're doing and you do it well. You're by far the best at it. And it's no longer, like, do you have an AI? It's like, can you produce? Right?

That might be the justification that AI is reaching, some focal point where people are like, we believe in it enough. That works? OK, great. I don't need you to tell me it's AI. I just need it to really work.

RM: That's interesting. We've seen that. I mean, we've seen that at Talla. I've seen that across the board in some of my other AI companies, which is some of the first markets where I started to make sense actually are now markets where people are looking at the performance of the AI from the perspective of the AI Spanish now, and it's just the performance metric of whatever the thing is they're trying to do that's AI driven. You know, how good is the prediction? Or, how much can you automate? Or whatever. Either personally, or through Boston Seed, are there any other AI companies that you've been in that you can talk about?

DB: Sure, so I'm an investor in Data Miner. Data Miner tracks social sentiment using Twitter feed and other tools, things like the Twitter feed. They use that to help financial institutions, governments, folks like that understand what's happening in the world. A lot of pattern recognition AI that's been coming through, I think, in making that happen.

Boston Seed is an investor in a company called 4JI. They're doing a lot of big data aggregation, and they use AI to understand patterns and trends. And they produce that as a simple way for people to understand the content about their organizations that's coming online. That's probably one of the worst explanations. Jen is going to kill me for that. But they're really cool.

RM: We'll make sure she doesn't listen to the podcast (laughs).

DM: Jen, if you're listening, I'm sorry. But that's actually, I mean, what they've been doing is pretty powerful. That is full AI. You know, back to that point, they're delivering a service that is working and people need. It's no longer because they have AI in their name.

I think there's a reason where AI, that's important. But like, maybe we've reached a trust curve almost. Like OK, that's cool. But, you don’t need to tell me your AI. It works. And we understand it works because AI's gotten to a point where it actually can deliver benefit to the market.

RM: You started a couple of companies, investing in a lot. You've lived through a couple cycles of new technology coming online. I guess sort of two-part question-- how do you learn about the new thing when it comes along? Do you do read about it? Do you dive in? Do you talk to smart people you know? What are some of your tools there for learning new things, particularly if you want to build a company in that space or adopt it? And then a second part of that-- was AI any different because it's so technical? Or is it still, from an applied level, a similar framework?

DB: I am non-technical. I'll go on record as saying that. You do not want me to work in your technology group. I still would say I could probably produce for you what I believe AI means. I'd probably be 20% right. Now, I don't know if it's the right 20% or the wrong 20%. But I'd probably be like, yeah, I get artificial intelligence. Here's how it might be used, et cetera.

I would say, I'm going to guess for many listeners, if you say AI, machine learning, big data. We munge those ideas into some format that looks right. Like even in my head, I'm thinking, one thing we do is we ingest just blockchain transactions at volume-- terabytes and terabytes. We apply data science models to understand patterns of behavior.

Now, I'm in my head I'm going, if I asked our head data scientist, whether we use AI, I don't even know if he'd say yes or no. We may be, for all I know. So, that's just the spot that's hard to pinpoint. Actually, what is AI?

RM: The definition that I use is I typically tell people that a system is intelligent if once you deploy it, it can learn and change on its own based on interactions with the world. If it can come back and give you a surprising insight or whatever, rather than just run a basic report, that's valuable. And there's a whole lot beyond machine learning. When you look at some of the next level, it will be machines that can do reasoning.

We haven't had a lot of guests on about this yet, but one of the big debates is whether or not deep learning in the current neural network models are going to get us to the next level of AI, or whether you're going to need-- a lot of people believe you're going to need this sort of cognitive architecture approach that they call it, which is like you have modules that perform specific functions in the brain. And this sort of ties back to a lot of research that says animal brains, human brains, they actually don't start, you know, tabula rasa, blank. It's actually there's some structures that have evolved to be in there innately and that you need to design a system with some of those.

DB: Wow, OK. That's deep. I would say we have a system where every blockchain wallet address performs transactions of some sort. They're into engaging in some transaction. The transaction could be a pattern that looks like it's touching an exchange, which is speculation. Or a pattern that looks like it's two users actually transacting between each other, which would be a utility. We set in motion some of those labeling purposes at the beginning. Then the system begins to self-reason and learn, and that's how we scale. You couldn't do that without doing that. Maybe I just learned we do something with AI, which is exciting.

RM: Your last company got acquired. You spent a couple of years at a very large company after that before you came out to do Flipside Crypto. In your role as sort of a senior executive at that point, were you starting to see transitions at work into AI? Like, were you guys talking about some of those tools, where you're seeing them deployed? And if so, where were you seeing the early adoption?

DB: The last two companies I've been in-- or last two startups-- were acquired by larger companies. BzzAgent was acquired by Tesco, and I spent four years in a division called Dunnhumby, which did data analytics for retailers. We talked about AI quite a bit there. Most of it was around things like where we have to scale the patterns of recognition. The business can't succeed without being able to apply what we do to large enough organizations with enough data-- every SKU level purchase at Kroger, for example, very difficult and complex without that.

One of my roles there was head of M&A and innovations. We struggled a lot with how do you get a tool that is strong enough to do those things to help the business move forward? One area we did work really hard to apply it was actually an ad tech. We did a few acquisitions in the ad tech space. And part of the discussion with those firms, I remember clearly, was understanding whether they had some ability to deliver AI to programmatic ad buying.

On the sale of Smarterer to Pluralsight-- so Pluralsight's model is developing technical learning modules online that people can use to grow their skills and their career. I would say we talked a lot about whether there was a way to understand what content people would need next or what learning path they might be on. Applied AI might be some of the patterns that could get us there.

I would say the theme I saw in those big companies was it was always on the table. It was always a conversation piece. I think my time at Dunnhumby was 2010 to 2014, and it was probably misused as a term. My time at Pluralsight was-- I'm going to get it wrong-- 2015 to 2016. Definitely used more effectively, if I think about understanding the general world. It would be interesting to go back today, 2019-- would they go, yeah, we understand what we're talking about now? So, yeah.

RM: Gotcha. As an investor in lots of startups do you have any AI theses? Or is there anything that you haven't seen that you'd like to see, somebody listening could pitch you on?

DB: My pattern recognition tends to be more about people than ideas. You know, probably the idea may not even be that good when I get to it. If the person is right, that seems interesting to me.

It's like a person, where you almost are here in the room, and you're like, I know I should want to leave, but I kind of don't want the meeting to end, type of thing.

So I tend not to say will somebody develop a blank-- a quantum computer that does x? If so, I'll invest in it. So I don't think there's anything like that. I would like to see-- maybe the question is, if I look around town here, the number of people who show up with an AI product that feels well-formed isn't a lot. Or it's not who calls. It might just be because I'm not technical enough. Not my investment strategy.

RM: Yeah, that makes sense. I mean, you and I both invested in LinkSquares when it was a different idea because we liked Vishal. We like the CEO.

DB: That's right.

RM: Definitely going to go somewhere. Good. Just as a sort of last question here, as someone who's not super technical and not dug in to sort of AI and where it's going on a technical level, do you worry about these debates, about killer robots taking over the world? Like, do you have a bunker somewhere in case Armageddon comes. Or are you just like not worried about it?

DB: I have lots of soup cans, and it's going to be great. I just read the thing, which may be hyperbole at this point, where some guys put a couple of marks on the road. And the car, I think it was a Tesla, it fooled the Tesla, and it drove into traffic. And so I'm pretty confident the world is going to end up in a place where AI can be used for good and evil. It's a little like the speed trap in a cop car and, you know, the radar detector in a passenger vehicle, where as the radar detectors get better and as the speed traps get better, they keep upping each other and figuring out how to solve the problem.

The problem with that Tesla with the road marks, that'll get solved. Then there'll be something else that could go wrong. And so I would say I worry less about the technical harm that comes from it because that's how the world has evolved. We achieve because technical innovations happen. The world in 50 years always looks like a scary place.

By the time you get there, you go, oh, there's all these things that make it less scary when you're in the middle of it. And it's because all these other things have happened to make it seem reasonable.

I would prefer to say, yeah, AI's going to do some crazy stuff for this world, stuff you couldn't imagine possible in this day and age. And yeah, some of it might seem scary today. But there will be other things that make it less scary by the time we get there.

RM: I tend to very much agree with that. So Dave Balter, thanks for being on. Those of you listening, thanks for listening. If you have guests you'd like us to interview or questions you'd like us to ask certain guests, please send those to podcasts@talla.com, and we'll see you next week.

Subscribe to AI at Work on iTunes, Google PlayStitcher, SoundCloud or Spotify and share with your network! If you have feedback or questions, we'd love to hear from you at podcast@talla.com or tweet at us @talllainc.