Jason Mars: How to Build a System for the System, Tech, Productivity, and More - Ep #97

Yeah, it's delightful. Thank you so much for having me. Looking forward to chatting.

Yeah, it's right next to Venezuela. I was born in Jamaica. My mom was studying at the University of the West Indies when I was born, and I grew up in Guyana, but both my parents are Guyanese. That's in the Caribbean, right next to Venezuela.

Until I was seven or eight, and that's when my dad and mom emigrated to the United States from Guyana.

Oh, absolutely. It's a completely different kind of socio-economic situation, and there are a number of cultural differences. The GDP per capita is incredibly low compared to what we're used to in the Western world or Europe, so it was definitely a different life experience. There's a common narrative that the only way to prosperity is through education, at least that's what my parents ingrained in me ever since I could remember. We didn't have a lot; we scrimped and scrounged. When we came to the US, we landed in New York City for a while. I lived in Queens. My mom worked at CVS. My dad had been an academician, a professor at the University of Guyana, so he was pursuing academic positions in the US. He ultimately landed at Wayne State University in Detroit, and that's how I ended up moving to Michigan for the first time. I moved to Detroit, lived there for a few years from Queens, then to the suburbs of Southfield. It's fascinating because Detroit is interesting. When I arrived, it was like Detroit was a Black city, and you cross 12 Mile, 13 Mile, hit 14 Mile, and then it's completely white. It's a very strange thing. We moved to the nicer suburbs, and that's kind of where I grew up. My upbringing was very different from the typical journey. I was poor, but I was trained early on that prosperity only comes through achievement. In every waking moment, it's like that.

Actually, not at all. Their journey did not influence me. Let me tell you how this worked. My mom was a lawyer in Guyana, a very successful lawyer. When we came over, my dad got into Wayne State. They were academicians, which is very different from the engineering sciences. My journey never had an interest in academia. I was always a tinkerer. My dream as a kid was to code. I started coding when I was like 15, and my dream was to create novel code. In high school, I created this pseudo company called Extreme and Intelligent Code, XIC. I would watch 'Pirates of Silicon Valley' over and over. My parents threatened serious violence to me if I didn't go to college, so I went to college really to appease them. Along the way, I discovered academic research. I didn't really appreciate or understand what it was, but I discovered it and recognized that it's a place where you can tinker and create unencumbered by a boss. You can get a PhD, become a professor, and research whatever you think matters, work on it, publish it, convince other people it's cool, and that's how success works. So that's actually my pathway back. I'm sure in my subconscious it was a thing because they loved their jobs and praised it all the time, so maybe in my subconscious I thought being a professor was an awesome thing. But it definitely wasn't my plan. My pathway to it felt like my own discovery when I was in college.

Yeah, something I'm really proud of is early in my academic career as a professor, I was able to be quite productive. I published a lot of high-quality work. Even the institution of creating itself, or the institutional system of academic research, I was never much of a follower. There are a lot of standard approaches to how you operationalize your research program. My research program experimented with innovating the process of producing research. To do high-quality work, you have to really understand what the bar is for rigor in its execution. I systematically studied that, and the work is out there on execution. I also systematically studied how the market works. I would look at proceedings and journals and study what fractions of papers cover what topics over time to understand this market almost as an economist, a microeconomics view of academic research. It's almost kind of disappointing when you look at some of the idea classes that just aren't being as successful but are quite interesting. For example, approximate computing: the idea that we can allow our computers to get things wrong some of the time for efficiency. If you develop the right mechanisms, that can be an adequate and even preferable way to do computation. This approximation domain was harder to get papers published in for quite some time. But then there are other factors. You start to realize that when your papers get rejected, it doesn't only correspond to the quality of the ideas and the effectiveness of the work itself; it also has to do with the market temperature and the level of rigor at which you executed your work. I focused on making sure I could relate every idea to the market. I always try to encourage all my students to do what I call PLP at three: they should have three distinct research stories they're working on at the same time. I've seen many students get discouraged by working on one idea, getting rejected, and thinking they're a bad researcher. That's the number one thing. I developed a number of these things, and that accelerated my productivity in academic research. I also created my own tools for writing papers. Students in my lab know about this thing called FPSWM: Fancy Paper Writing System from Mars. We used that for all of the first six or seven years of my research. It's a tool I wrote in Python that automates almost everything about writing papers, like graph generation and organization. It's really interesting. One of these days, I'll release it for the world to see.

No, absolutely. There are topics where everyone would say it's impossible to publish a top-tier paper. For instance, there's something called loop tiling, an old compiler technique. There's a paper I have called Shapeshifter; it was initially called Universal Tiling. It got rejected when it was called Universal Tiling. This is an idea that most folks would say, 'Oh, this is a compiler, very niche compiler work. Nobody cares about loop tiling.' But it takes relating your ideas to that audience. The ideas don't change, but the way you execute your work and the way you frame it and relate it to that audience, if it's sensitized with the nature of that market, you actually can make more of an impact.

I'm somewhat conservative. I'm kind of very socially compassionate, but I'm somewhat conservative. I believe in capitalism, free markets, etc. Of course, it doesn't solve all problems, but it reflects human nature most accurately. Even though we aspire to something that looks more like a socialist or communist situation, I think our nature fundamentally aligns with a more free-market system. I just want to emphasize that. But I've been recently looking at some of Karl Marx's work and the dynamics of how that works from an economic viewpoint. Honestly, what we're experiencing today, you mentioned the nature of communication and how you frame things matter and what's okay to say or not okay to say. There's this weird dynamic when you take some of that theory and intermingle it with social media, the internet, and the connectivity of everyone being connected all the time, plus the pressures of COVID that accelerated our reliance on digital systems. We're observing the realization of this well of power that the masses have. I'm not going to attribute it to being good or bad; it just is. There's this well of power available to everyone. At the same time, there is the propensity for people to align themselves with the truths they prefer and an ability to stay in the realm of the truths they prefer in this new digital world. So you get to stay with the truths you like. You've got a lot of power. This thing manifested in history: the laborers are always dissatisfied with the folks that own stuff. The folks who convert their time into resources and money, there's always a tension with the folks who convert people into profit and own all the means of production. So that's there. Then you have cancel culture, justice, this power, and people able to encircle themselves in the truths they prefer. We live in a very unique time, and we have emerging technologies.

Yeah, it's been a long journey. These tools, conversational AI and AI in general, give us new kinds of tools to make our lives more efficient and our experiences more convenient. If you look back at it, there was Sirius. It's amazing. I don't know if Forbes is good at what they do, but I talked about this in detail. It was called Sirius, and Apple reached out with a letter saying it was too close to Siri. I was like, 'What? Sirius XM, Sirius Black, Sirius is a star system. Come on, give me a break.' I tried to push back with the university, but the university fought with me. There's a hilarious chapter about this in the book. At one point, the university said I couldn't do it because they were terrified of Apple's lawyers. That's what springboarded the entrepreneurial pursuit that became Clinc, which is to create a next-level conversational AI. These kinds of tools, people are understanding how to integrate them into their way of life in different ways. I think we're just at the beginning of this evolution, and we're still figuring many things out. It's weird because you've got different classes of things. This is one of those slow-moving, world-changing things. It's moving slower than many of us anticipated for a host of reasons, but it's not dead. We can juxtapose it to virtual reality, which hasn't died but hasn't gained as much traction. AR or VR, AR kind of had Pokémon Go, but you haven't heard much since. Conversational AI is a different beast because every year it's growing. People aren't doing it much better as time passes, at least from the experiences we have, but the market for it and the investment is growing. Tech is the barrier. I always used to say we've had Siri for decades now, probably over 10 years. The quality of these experiences hasn't grown as much as we'd expect. But I don't think that's because of the science; the science has progressed substantially. I think it's because of the realization of these technologies into products and the tuning or necessary research for how we can apply some of the science. With what I'm working on with Jaseki and ZeroShotBot and some of these other technologies, I think we're taking another leap forward. You really have to drive this kind of innovation by challenging the established means of how we build these things. It's going to be interesting to see how we utilize these technologies to address some of the societal problems, but we're so early that folks are just trying to save costs, monetize things, and create sexy features. It's still early to have that kind of impact.

As a professor, you do a lot of really quality work. The pinnacle of that work is getting top-tier papers, winning awards, being inducted into Hall of Fames. It's a little community. All these professors are trying to have impact, which means we go and evangelize our research to companies like Google, Amazon, or Facebook. We try to say, 'You should use our research. The results of our research should impact how you do things.' Professors go give talks at these institutions and say, 'Please use our stuff; it can make you better.' Some things get picked up, some don't. I wanted to have an impact and create great outcomes for the group. The students really realize the biggest impact. But there's a ceiling to how much of that you can do purely as an academic. You have to take your creations into industry if you want to have that impact. It's a lot like pharmaceuticals. If I come up with a cancer drug, it has to become a product by Pfizer or my pharmaceutical startup before people can take the drug and it heals them. Same with technology. Clinc was a rocket ship. We raised $250k, then $6.5 million, then $53 million, year after year. It grew incredibly fast. What matters even more is that the technology we created was able to penetrate the actual market. People were adopting it. We had 20-something million users of our tech, and we were beating out the incumbents in the space, like Google with their Dialogflow, as well as the others. The way I led that company, we had to do some against-the-grain stuff. We had to focus on a principle called 'show, don't tell.' While the first investors were encouraging us to sign letters of intent and show that customers want it before we built anything, I said, 'No, we're going to stall that and build something amazing.' We architected and built one of the most sophisticated virtual assistants yet, called Finney. It was a big splash in the market because it was obvious that it was very different. That was an incredible journey. We grew it from zero people, just me, then a couple of grad students, then 130 people. It was a jubilant creation in the market, and we demonstrated a shift in progress in technology on one of the hardest problems in the world. Apple, Google, Amazon, Microsoft have all been investing heavily in conversational AI, and we were able to do something special there and beat those kinds of folks out.

Yes. After architecting it, I brought on some of the best engineers to build it out. It was all novel tech. We have 30 patents from Clinc on that journey. It was research turned into product.

The book chronicles the whole zero to 100 of the company, peaking in the market. Chapter four talks about how to organizationally create an engine that is a research engine and how to build that culture where you can operationalize a research program so that the output is impact in the market.

This is a very fascinating topic area. Something you can tuck on top is that people are all working together to create goodness for the entire organism. It's a collective. There are hierarchies in that thing. With scale, there were so many learnings on this journey. This was my first company. I was fresh out of grad school into professorship, did that for a couple of years, and then started my first company. In just three years, the company had 130 people. There were a lot of perspectives, cultures, and philosophies. One philosophy was 'let the best ideas win.' It doesn't matter if it comes from the janitor; ignore the hierarchy. Everyone has a voice, everyone can share, and let's have a culture of respecting not who an idea comes from but what the idea is. That way, the organism collectively can let the best ideas bubble to the top. This was good at a certain size, but after about 20 people or so, it's difficult for that to scale because people will have a lot of ideas about everything. You've got engineers, salespeople, financial management, all these different groups in a company organism. Engineers have a lot of perspective on the product and the design of the code. Salespeople have a lot of perspective on what the market says about the product. Most people don't deeply appreciate the scope of their perspective. For instance, you'll go on the street and hear someone say, 'Trump was the worst president America ever had.' Most people don't have the perspective to make that judgment. They're usually certain of what they think, but if you ask someone like Obama about Trump, you may get something much more nuanced because Obama has a different perspective. He's seen different stuff. A layperson hasn't created policy for a small company, let alone the country, and may not have led more than three people, but they will be certain that Trump is not a good leader. That's a perspective thing. Folks will feel as though their perspective is the right one. When you create a culture where everyone has a voice, which sounds like a holy good thing, the problem is that people with less perspective will think they have the best idea, and people with more perspective will think they have the best idea. Salespeople will think engineers should be doing something different with limited perspective, and engineers will think salespeople should be doing something different with limited perspective. When you're small enough, it's easy to share perspectives. If you're 10 people, you just talk for a while and everybody can see the same picture. But when you're 20, 30, 40, 50 people, you need structure in that organization. You have to establish the roles and scopes of people's work, and then you have to manage that people are focused on that scope with the right evaluation criteria. Performance evaluation serves a very important role in the psychology of people because it lets people know what good looks like or the direction of success. You have to lay that out at a certain scale. It's intuitive when you're small; everybody knows what success looks like. But when you're bigger, you need to define what success looks like for each person. If you don't have that, you'll see things unraveling because people will assume they should spread their attention to everything going on. They'll also be at the disadvantage of less perspective, so their judgments are less durable, but they believe their judgments as much as the random person saying Trump was the worst president. Most people haven't been alive for most of the presidents of the United States. How do we know who the worst president is? That's not to say I'm a big... anyway. You need to create engines that are resilient to some of those natural things that happen in humans. Scale requires structure. As you get bigger, you have to have more structure. There are a ton of those kinds of mistakes that I think ultimately harmed Clinc. The naive mistakes that leadership was responsible for.

Yeah, that's absolutely right. There are a lot of dynamics to the way we work and think that come into play. There are hierarchies that have to be acknowledged and respected, both from the folks at the top and those lower down. The folks at the higher end of the hierarchy have to respect the hierarchy in a very sensitive way. I think a lot of the challenges at Clinc now, especially since the people in the company are very different now than when I was there, are not coming from a problem with the product or market fit. The problem comes from the way it's organized. Mistakes were made in not evolving quickly enough in tandem with size, not learning what changes are necessary at different phases of the growth journey. That's too bad because it's a beautiful creation, and it's not realizing the potential that was still on the table. It's crazy.

The market is evolving even more. It's crazy. The first thing I did after Clinc was the book because I thought there were a lot of special insights in that journey. But I can't understate the wealth of new perspective. As an academician doing top-notch research and understanding how novel science and ideas come to be, I did that for many years as a professor. The five years at Clinc was being in the field, in the market, in the world, and there's incredible perspective built from that, both on the real implications of AI in the world to help people. Now, I'm taking the wealth of all that perspective, both academic and industry or commercial, and I'm focused on building the most sophisticated AI engine ecosystem paradigm. There's a new programming language called Jaseki, a paradigm for computation, a computational model that's quite different than anything I've seen before. You can articulate solutions to sophisticated problems with this AI computational model plus language. There are three products built on that, each wrapped in its own company, like Alphabet. All three products are live today with users, and they represent a new generation of what's possible when it comes to improving our lives. They are incredibly disruptive products, currently invite-only in early access, built on Jaseki. If you go to jaseki.org, you can get a bird's-eye view into those three products. I believe this engine, this computational model, this new technology stack, is going to fuel a generation of new kinds of AI products and new kinds of computer science solutions to big problems. I'm really keen on that. I'm putting a lot of my energy into that these days. I'm motivated by creating something beautiful that could help people. I'm not motivated by money, success, or glory. I'm completely past that. I've learned so much about the world and life that that's not interesting anymore. What's interesting is creating something beautiful that helps people. So that's kind of the current journey.

It's been about a year and a half now.

Thank you, man. I'm still quite early in my career, so it's a journey. You develop these things. Actually, what was almost a fault, my wife is like, 'Dude, why do you care so much about all the people in the world? People are crazy.' I'm like, 'I don't know, it's just stitched into my DNA.' That's just what motivates me: creating beautiful things that can benefit others.

Yeah, LinkedIn is good. I'm happy to connect with anyone who is interested or curious or just wants to chat.

Oh, anytime. Thank you very much. It's been great chatting with you, and we'll chat again soon, I'm sure.