The Exchanges

Q And how do you view the integration of big data and AI? Is AI helping big data deliver the promise of a couple of years ago we thought would potentially be sooner?

A Yes. Um, so AI, uh, from my perspective is very much a child of, uh, big data. So the, the, uh, fascinating aspect of, uh, of AI is that, uh, you know, a lot of the current resurrection of AI is, um, around deep learning specifically. And, uh, and deep learning for the most part, uh, was really developed by, uh, some, you know, particularly, uh, forward-thinking researchers Uh, like, you know, you have a con, Jeff Hinton, that, that, that group, and it's really only up until recently when, um, those algorithms intersected with big data, so specifically the ability to, um, process very large amounts of data quickly, and also with all the latest developments, um, around processing powers, and especially all the Great work that NVIDIA did around GPUs. That when all of this intersected, then deep learning sort of was, was resurrected, and, um, and its worth became truly apparent to everyone. So, AI is a child of big data. At the same time, AI is really, um, enabling big data to truly deliver its promise. The, uh, great thing about big data is that you have plenty of data, but of course it doesn't do you any good. Unless you can really, uh, you know, do actual, um, analysis and actual predictions with it. And, uh, that's exactly what, what, what AI and, you know, these various forms, uh, and machine learning in general, um, enables. So there's a whole series of big data applications…

AI assessment note: “AI is really, um, enabling big data to truly deliver its promise.”

Q So it's the solution for those companies to cross the chasm and actually adopt the big data technologies that they haven't so far yet adopted. Is the solution a one-stop shop solution by IBM? Is that a potential?

A Well, I think that's what, uh, all those companies have been hoping for, um, and, uh, I think it will eventually happen, but it's unclear how quickly that will happen. It's actually a very daunting, um, proposition when you, when you, when you look at the reality of the big data, because the big data is really, uh, not one technology, but it's really an assembly line of different technologies and processes and culture. Just on the technology part, as I mentioned earlier, you have the repositories, you have the analytics, you have the application, and the, the concept of building a one-stop shop is, is tricky. So eventually the, uh, industry will presumably consolidate. Every industry, uh, at some point consolidates. Uh, but my sense is that it's not happening anytime soon. It's certainly not at scale. I think, uh, a lot of the Startups in the space actually doing quite well and experiencing, uh, rapid growth of revenues and, and those companies are going to want to go, uh, public or continue as independent companies. It would be extremely expensive for a large vendor to, uh, acquire enough companies to offer one-stop shop. So my, my sense and my recommendation, uh, to large companies that are still a little bit, um, on, on the sidelines of big data is that now is, A good time to start playing with those technologies a lot more. The ecosystem is mature. There's not going to be a…

AI assessment note: “There's not going to be a one, a one stop shop anytime soon.”

Q much about it, and that's big data. But before we do dive into this, I do want to say a disclaimer that Matt's research and expertise extends way beyond big data. And you can see this on his blog, which really is one of my favorites, mattturk.com. Um, but today we're on big data. So let's start then by covering what the buzzword actually means. So what is big data?

A So in terms of what it means, I think that's really two definitions, a narrow one, and then perhaps a broader one. Narrow definition is really that there's a growing group of companies that have absolutely massive data sets, whether they are fast growing digital companies or large multinationals. And for those companies, there's a whole arsenal of tools and enabling technologies that enable them to process data faster and cheaper than ever before. And that's really the world of, you know, Duke and Spark and all those other frameworks. Uh, but beyond that definition, there's, there's a broader, uh, definition that to me is almost more interesting, which is what I would call the spirit of big data. That's really that, uh, as every company is becoming a technology company, uh, there's an ever increasing amount of data that's available to them that can be analyzed. Um, so there's an opportunity to develop a very powerful data-driven culture and be much smarter about one's business. And There's an aspect that I find truly interesting, uh, when you have a combination of data tools and data culture, and that becomes part of the fabric of every modern business, not just the businesses that have those absolute massive data sets.

AI assessment note: “So in terms of what it means, I think that's really two definitions”

Q much about it, and that's big data. But before we do dive into this, I do want to say a disclaimer that Matt's research and expertise extends way beyond big data. And you can see this on his blog, which really is one of my favorites, mattturk.com. Um, but today we're on big data. So let's start then by covering what the buzzword actually means. So what is big data?

A So in terms of what it means, I think that's really two definitions, a narrow one, and then perhaps a broader one. Narrow definition is really that there's a growing group of companies that have absolutely massive data sets, whether they are fast growing digital companies or large multinationals. And for those companies, there's a whole arsenal of tools and enabling technologies that enable them to process data faster and cheaper than ever before. And that's really the world of, you know, Duke and Spark and all those other frameworks. Uh, but beyond that definition, there's, there's a broader, uh, definition that to me is almost more interesting, which is what I would call the spirit of big data. That's really that, uh, as every company is becoming a technology company, uh, there's an ever increasing amount of data that's available to them that can be analyzed. Um, so there's an opportunity to develop a very powerful data-driven culture and be much smarter about one's business. And There's an aspect that I find truly interesting, uh, when you have a combination of data tools and data culture, and that becomes part of the fabric of every modern business, not just the businesses that have those absolute massive data sets.

AI assessment note: “I think that's really two definitions, a narrow one, and then perhaps a broader one.”

Q And how do you view the integration of big data and AI? Is AI helping big data deliver the promise of a couple of years ago we thought would potentially be sooner?

A Yes. Um, so AI, uh, from my perspective is very much a child of, uh, big data. So the, the, uh, fascinating aspect of, uh, of AI is that, uh, you know, a lot of the current resurrection of AI is, um, around deep learning specifically. And, uh, and deep learning for the most part, uh, was really developed by, uh, some, you know, particularly, uh, forward-thinking researchers Uh, like, you know, you have a con, Jeff Hinton, that, that, that group, and it's really only up until recently when, um, those algorithms intersected with big data, so specifically the ability to, um, process very large amounts of data quickly, and also with all the latest developments, um, around processing powers, and especially all the Great work that NVIDIA did around GPUs. That when all of this intersected, then deep learning sort of was, was resurrected, and, um, and its worth became truly apparent to everyone. So, AI is a child of big data. At the same time, AI is really, um, enabling big data to truly deliver its promise. The, uh, great thing about big data is that you have plenty of data, but of course it doesn't do you any good. Unless you can really, uh, you know, do actual, um, analysis and actual predictions with it. And, uh, that's exactly what, what, what AI and, you know, these various forms, uh, and machine learning in general, um, enables. So there's a whole series of big data applications…

AI assessment note: “AI is really, um, enabling big data to truly deliver its promise.”

Q What's the most challenging aspect of your role as managing director at Firstmark?

A That's, that's a, that's a great question. I think, uh, so there's one, one, one, uh, one very technical thing about managing, uh, bandwidth. Uh, it's, um, there is this, this image, uh, of, of VCs that maybe had some, uh, basis, uh, a few years, you know, as, you know, people who sort of, uh, decide when they want to have a meeting and, uh, play golf and all the things. I think that The reality of it is, uh, at least from my perspective, much more in the trenches, and it's a combination of, um, uh, you know, generating and then analyzing and, uh, you know, processing a massive amount of deal flow if you're, if you're so lucky that, uh, you know, people, uh, want to, want to reach out. So that's one aspect. Uh, the second aspect is that, uh, we make a, a very strong, uh, commitment to our founders. Uh, so FirstMark is, uh, one of those firms that, uh, Doesn't make tons of new investments. You're, you know, very much of a sort of classic Series A type model, uh, but, uh, spends a lot of time with, uh, you know, with, with our companies and our founders. So, uh, as you build a portfolio, that tends to add up quite a bit. Uh, and the last part is that, um, being a partner at a firm, especially, uh, you know, not a firm that has a very large number of partners and, and, and, And billions of dollars in the management is that, uh, you essentially run something that in many ways feels…

AI assessment note: “one very technical thing about managing, uh, bandwidth.”

Q And I, I'm intrigued that a lot of people are saying, you know, we, we don't have a hundred billion dollar marketing Uh, tech company or MarTech company, uh, as some like to reference it. Um, so, so do you think big data will allow for this MarTech unicorn to be created?

A Yeah, I think, I think there is a whole generation of, uh, big data driven companies doing very interesting things in, um, in marketing tech. So, you know, MarTech is, is a, um, beautiful but tricky market because, um, on the one hand, there is a tremendous amount of, uh, Money available for the right technologies, and that's the whole train around the CMO becoming a larger producer of technology than the CIO. Uh, so that's a great part. The more tricky part is that, uh, as a result of the above, uh, there is a tremendous amount of companies, uh, in that space. So it's very competitive. It's, it's hard to break through, uh, the notes. So that's, that's in terms of general landscape. Now, specifically, I think, uh, I think marketing is a, uh, perfect area. For, uh, big data technology to be applied. The old dream of being able to, um, consolidate all the information that a company gets from multiple channels, whether that's, you know, online, mobile, call center, and all sorts of different sources, and, and crunching all of this, uh, cheaply enough, fast enough, enabling a real, The actual, uh, campaigns being built on top of the data crunching, all of this is, is really becoming a reality. We have, uh, a company in the portfolio called Action IQ that does, um, exactly, exactly that. And interestingly, what those guys are doing would simply not have been possible to do, uh, up u…

AI assessment note: “I think there is a whole generation of, uh, big data driven companies”

Q So it's the solution for those companies to cross the chasm and actually adopt the big data technologies that they haven't so far yet adopted. Is the solution a one-stop shop solution by IBM? Is that a potential?

A Well, I think that's what, uh, all those companies have been hoping for, um, and, uh, I think it will eventually happen, but it's unclear how quickly that will happen. It's actually a very daunting, um, proposition when you, when you, when you look at the reality of the big data, because the big data is really, uh, not one technology, but it's really an assembly line of different technologies and processes and culture. Just on the technology part, as I mentioned earlier, you have the repositories, you have the analytics, you have the application, and the, the concept of building a one-stop shop is, is tricky. So eventually the, uh, industry will presumably consolidate. Every industry, uh, at some point consolidates. Uh, but my sense is that it's not happening anytime soon. It's certainly not at scale. I think, uh, a lot of the Startups in the space actually doing quite well and experiencing, uh, rapid growth of revenues and, and those companies are going to want to go, uh, public or continue as independent companies. It would be extremely expensive for a large vendor to, uh, acquire enough companies to offer one-stop shop. So my, my sense and my recommendation, uh, to large companies that are still a little bit, um, on, on the sidelines of big data is that now is, A good time to start playing with those technologies a lot more. The ecosystem is mature. There's not going to be a…

AI assessment note: “There's not going to be a one, a one stop shop anytime soon.”

Q And I, I'm intrigued that a lot of people are saying, you know, we, we don't have a hundred billion dollar marketing Uh, tech company or MarTech company, uh, as some like to reference it. Um, so, so do you think big data will allow for this MarTech unicorn to be created?

A Yeah, I think, I think there is a whole generation of, uh, big data driven companies doing very interesting things in, um, in marketing tech. So, you know, MarTech is, is a, um, beautiful but tricky market because, um, on the one hand, there is a tremendous amount of, uh, Money available for the right technologies, and that's the whole train around the CMO becoming a larger producer of technology than the CIO. Uh, so that's a great part. The more tricky part is that, uh, as a result of the above, uh, there is a tremendous amount of companies, uh, in that space. So it's very competitive. It's, it's hard to break through, uh, the notes. So that's, that's in terms of general landscape. Now, specifically, I think, uh, I think marketing is a, uh, perfect area. For, uh, big data technology to be applied. The old dream of being able to, um, consolidate all the information that a company gets from multiple channels, whether that's, you know, online, mobile, call center, and all sorts of different sources, and, and crunching all of this, uh, cheaply enough, fast enough, enabling a real, The actual, uh, campaigns being built on top of the data crunching, all of this is, is really becoming a reality. We have, uh, a company in the portfolio called Action IQ that does, um, exactly, exactly that. And interestingly, what those guys are doing would simply not have been possible to do, uh, up u…

AI assessment note: “I think there is a whole generation of, uh, big data driven companies”

Q What blogs would you say we should be reading? What, what are your must reads when they come in?

A You know, in general, uh, uh, that's an interesting question. You know, so starting, starting with the classics, you know, Fred Wilson, Chris Dixon, Radfeld, and all those things, and I think that's, that's really, um, you know, part of, uh, what, what all of us should be reading on a, on a daily basis. There are, you know, a handful of blogs that I find particularly, uh, interesting and thoughtful, um, One that comes to mind is Stratechery. I also like broader blogs, not necessarily about tech, or that touchable tech, but not all the time. You know, I think, wait, but why is particularly interesting, and actually to your question about, about AI, there's two or three blog posts that he did there on AI that were a particularly good introduction to AI for anybody that's curious about the topic, but doesn't want to spend, you know, hours and hours reading it.

AI assessment note: “starting with the classics, you know, Fred Wilson, Chris Dixon, Radfeld”

Q From a single company's perspective and from a data set building perspective?

A Yes, so it's a, it's a, uh, it's a complex question as obviously it's somewhat, um, case specific. What I'm seeing is that large companies, um, so there's, there's really two types of, of companies right in the, that, that, that, that have been adopting those big data technologies. So one is the first group that was really a whole series of, uh, sort of internet native companies. And, uh, when, when you look at the history of big data, a lot of the tools were created by, uh, you know, the, the Googles and the Yahoo's and LinkedIn and Facebook. So those internet native companies that had those massive amounts of data that they needed to deal with and basically build their own technology, build technology for their own needs and became not only developers, but adopters at scale and in production of those technologies. So that's, that's one of them. What were those guys, um, had, uh, going for them is that as reasonably, it was sometimes very new companies that didn't have any legacy infrastructure to, that they needed to deal with. Now there's a separate group of, um, of, of companies, which is really the, the, pretty much of the rest of the world. Uh, and those are, those are companies that have a legacy infrastructure and, um, you know, in general, have, have, have an infrastructure that, that works. So for those companies, uh, you know, the, the, uh, it's, it's been, um, much …

AI assessment note: “it's been much more of a tricky exercise to adopt big data technologies”

Q What's the most challenging aspect of your role as managing director at Firstmark?

A That's, that's a, that's a great question. I think, uh, so there's one, one, one, uh, one very technical thing about managing, uh, bandwidth. Uh, it's, um, there is this, this image, uh, of, of VCs that maybe had some, uh, basis, uh, a few years, you know, as, you know, people who sort of, uh, decide when they want to have a meeting and, uh, play golf and all the things. I think that The reality of it is, uh, at least from my perspective, much more in the trenches, and it's a combination of, um, uh, you know, generating and then analyzing and, uh, you know, processing a massive amount of deal flow if you're, if you're so lucky that, uh, you know, people, uh, want to, want to reach out. So that's one aspect. Uh, the second aspect is that, uh, we make a, a very strong, uh, commitment to our founders. Uh, so FirstMark is, uh, one of those firms that, uh, Doesn't make tons of new investments. You're, you know, very much of a sort of classic Series A type model, uh, but, uh, spends a lot of time with, uh, you know, with, with our companies and our founders. So, uh, as you build a portfolio, that tends to add up quite a bit. Uh, and the last part is that, um, being a partner at a firm, especially, uh, you know, not a firm that has a very large number of partners and, and, and, And billions of dollars in the management is that, uh, you essentially run something that in many ways feels…

AI assessment note: “one very technical thing about managing, uh, bandwidth.”

Q From a single company's perspective and from a data set building perspective?

A Yes, so it's a, it's a, uh, it's a complex question as obviously it's somewhat, um, case specific. What I'm seeing is that large companies, um, so there's, there's really two types of, of companies right in the, that, that, that, that have been adopting those big data technologies. So one is the first group that was really a whole series of, uh, sort of internet native companies. And, uh, when, when you look at the history of big data, a lot of the tools were created by, uh, you know, the, the Googles and the Yahoo's and LinkedIn and Facebook. So those internet native companies that had those massive amounts of data that they needed to deal with and basically build their own technology, build technology for their own needs and became not only developers, but adopters at scale and in production of those technologies. So that's, that's one of them. What were those guys, um, had, uh, going for them is that as reasonably, it was sometimes very new companies that didn't have any legacy infrastructure to, that they needed to deal with. Now there's a separate group of, um, of, of companies, which is really the, the, pretty much of the rest of the world. Uh, and those are, those are companies that have a legacy infrastructure and, um, you know, in general, have, have, have an infrastructure that, that works. So for those companies, uh, you know, the, the, uh, it's, it's been, um, much …

AI assessment note: “there's really two types of, of companies right in the, that, that, that, that have been adopting”

Q You, you absolutely nailed it there, obviously, with as long as you have the data, and your final question before we dive into the quickfire. You're an investor, obviously, in, uh, Amy, the brilliant assistant, um, absolutely love that, but is there not an incumbency advantage, then, in this early AI startup space for startups without the available data sets that your Facebooks, Googles have?

A Yeah, so that, that's the, uh, that's a great question, and that's, uh, certainly something that I think about a lot. I think from, uh, a founder, you know, startup perspective, and from an investment perspective, I think, I think, uh, one needs to think about, um, applications, uh, where, uh, you have a chance to build something, uh, reasonably, uh, reasonably quickly, uh, and something that's not going to be Absolutely in collision course with something that's, uh, core to, uh, you know, the goals of the, of the world. I think there's going to be a whole range of interesting, uh, and compelling AI vertical applications, and the X.ai that you mentioned is certainly one of them, and certainly at the forefront of that trend is a rare opportunity to become a category-defining company, but I think one needs to To, uh, be very careful in, in how they think about it. So, uh, you know, if you, uh, do something like image processing, which is something that deep learning is particularly great at, uh, you know, you need to think of a business model and, and a way of penetrating the markets that, um, somehow does, doesn't put you in a direct collision course with, with Google, um, because, uh, you know, Google Or he does, and will want to do this at scale, because it's absolutely part of their, of their core mission.

AI assessment note: “doesn't put you in a direct collision course with, with Google”

Q I'm intrigued that I can, I can't stay away from that big data applications. Why are you most excited for big data applications to be exploited?

A So if you think, if you, if you look at, uh, the evolution of the big data landscape, there's, there's really been several phases. Um, and a little bit to the, to the earlier question as well. First few years were really about the, uh, creation and deployment of that core infrastructure. So, you know, we have all this data floating around and then we keep getting more and more and more data every day. Uh, where do we store it? And that's really the, uh, infrastructure phase that was, uh, Hadoop initially and all the latest framework and many of the NoSQL databases. Um, so that was sort of phase one. The phase that almost immediately followed was, okay, we have, we have all this data. We've been able to store it in great repositories. Uh, now what, what good does it do us? I mean, how can we analyze it? Uh, how can we make sense of it? And that was really a whole, um, generation of analytical tools that worked, um, on top of those big data repositories. And then the, the latest phase, which is really happening, uh, now is the emergence of all those, um, applications.

AI assessment note: “the latest phase, which is really happening, uh, now is the emergence”

Q I'm intrigued that I can, I can't stay away from that big data applications. Why are you most excited for big data applications to be exploited?

A So if you think, if you, if you look at, uh, the evolution of the big data landscape, there's, there's really been several phases. Um, and a little bit to the, to the earlier question as well. First few years were really about the, uh, creation and deployment of that core infrastructure. So, you know, we have all this data floating around and then we keep getting more and more and more data every day. Uh, where do we store it? And that's really the, uh, infrastructure phase that was, uh, Hadoop initially and all the latest framework and many of the NoSQL databases. Um, so that was sort of phase one. The phase that almost immediately followed was, okay, we have, we have all this data. We've been able to store it in great repositories. Uh, now what, what good does it do us? I mean, how can we analyze it? Uh, how can we make sense of it? And that was really a whole, um, generation of analytical tools that worked, um, on top of those big data repositories. And then the, the latest phase, which is really happening, uh, now is the emergence of all those, um, applications.

AI assessment note: “latest phase, which is really happening, uh, now is the emergence of all those, um, applications.”

Q Let's start with your favorite kind of academic text. Is there kind of a big data AI book that you most recommend for us to take away today? Mine would be Nick Bostrom's. I don't know if you've read Nick Bostrom's.

A Yeah, so there's a whole series of, um, there's a whole series of those. I find the difficulty of a lot of those books is that there's two types of books. There's one book, which is deeply technical, and for the non, for the business user or somebody who's curious about AI, then you end up feeling overwhelmed fairly, fairly quickly. And then there's a different type of book, which is, uh, talks about very fundamentally important issue around, you know, is AI dooming us as a machine? Um, arriving and, uh, taking control of our lives. And I think that's, you know, I'm not saying this facitiously, but it's a fundamentally important question. Um, so I, I actually find that I've learned a lot more around AI, uh, by just, just reading a lot of stuff online, typically in shorter formats. You know, there's all sorts of, uh, people to follow on Twitter and, and, and blogs to read and, and all those things. And I, I would recommend that as an approach as opposed to, um, to a specific approach.

AI assessment note: “I would recommend that as an approach as opposed to, um, to a specific approach.”

Q Let's start with your favorite kind of academic text. Is there kind of a big data AI book that you most recommend for us to take away today? Mine would be Nick Bostrom's. I don't know if you've read Nick Bostrom's.

A Yeah, so there's a whole series of, um, there's a whole series of those. I find the difficulty of a lot of those books is that there's two types of books. There's one book, which is deeply technical, and for the non, for the business user or somebody who's curious about AI, then you end up feeling overwhelmed fairly, fairly quickly. And then there's a different type of book, which is, uh, talks about very fundamentally important issue around, you know, is AI dooming us as a machine? Um, arriving and, uh, taking control of our lives. And I think that's, you know, I'm not saying this facitiously, but it's a fundamentally important question. Um, so I, I actually find that I've learned a lot more around AI, uh, by just, just reading a lot of stuff online, typically in shorter formats. You know, there's all sorts of, uh, people to follow on Twitter and, and, and blogs to read and, and all those things. And I, I would recommend that as an approach as opposed to, um, to a specific approach.

AI assessment note: “I would recommend that as an approach as opposed to, um, to a specific approach.”

Q You, you absolutely nailed it there, obviously, with as long as you have the data, and your final question before we dive into the quickfire. You're an investor, obviously, in, uh, Amy, the brilliant assistant, um, absolutely love that, but is there not an incumbency advantage, then, in this early AI startup space for startups without the available data sets that your Facebooks, Googles have?

A Yeah, so that, that's the, uh, that's a great question, and that's, uh, certainly something that I think about a lot. I think from, uh, a founder, you know, startup perspective, and from an investment perspective, I think, I think, uh, one needs to think about, um, applications, uh, where, uh, you have a chance to build something, uh, reasonably, uh, reasonably quickly, uh, and something that's not going to be Absolutely in collision course with something that's, uh, core to, uh, you know, the goals of the, of the world. I think there's going to be a whole range of interesting, uh, and compelling AI vertical applications, and the X.ai that you mentioned is certainly one of them, and certainly at the forefront of that trend is a rare opportunity to become a category-defining company, but I think one needs to To, uh, be very careful in, in how they think about it. So, uh, you know, if you, uh, do something like image processing, which is something that deep learning is particularly great at, uh, you know, you need to think of a business model and, and a way of penetrating the markets that, um, somehow does, doesn't put you in a direct collision course with, with Google, um, because, uh, you know, Google Or he does, and will want to do this at scale, because it's absolutely part of their, of their core mission.

AI assessment note: “doesn't put you in a direct collision course with, with Google”

Q So two more questions. Your deep dives are some of my favorites. So how do you approach the deep dive? Is there a kind of a content creation plan, a research schedule? How do you approach a big topic like big data or IOT?

A Yeah, so I think I do it as a thought exercise. Part of the idea is really to open source the effort and, uh, Uh, it's been tremendous how, uh, those deep dives just generate, um, all sorts of new conversations with people I hadn't had a chance to, to meet. That has a forcing function as well of, uh, just, uh, making me just constantly read and constantly meet new companies and, and whether, you know, small or large companies really stay on top of, uh, of an industry. So it's, um, you know, I find it, uh, interesting to do. I find that it pays all sorts of rewards. Um, it, Works very well with, um, all the things I do on the community building front and event front, so as, uh, perhaps you know, I run a couple of large events, one called Data Driven NYC, which has almost, uh, 11,000 members, and is, uh, the largest community of its kind, uh, in the country, and another one, uh, called Hardware NYC, which covers hardware and software startups, and, um, something that has, uh, I think something close to 5000 members. Um, so all of this, uh, you know, works very well together and, um, and forces me to basically, uh, stay on top of an industry that I'm passionate about. But at the same time, given what I was mentioning earlier about, uh, you know, constant bandwidth challenge, um, it's, it's, uh, yeah, so I would sort of kick my own butt if you want.

AI assessment note: “works very well with, um, all the things I do on the community building front”