Welcome back, everybody. We are now joined by Dr. Jason Mars, where he will be talking about a new voice from a strange place in the AI revolution. Jason, so glad to have you to join us today. For those of us in the audience, Jason, Dr. Jason Mars was the founder of clink conversational AI company and associate and an associate professor at the University of Michigan. I'm so excited to be listening to this presentation, I will be in the background taking furious notes. Jason, welcome.
Awesome. Awesome. Thank you so much. Can you see my slides? Anyway, let me jump right in. And the way I wanted to start this conversation was with this slide. So if you look at this slide, this slide shows us a timeline of the human species, right. So this is 12,000 years of the human population. And if you if you look at the slide, it's it's interesting, right? Just around, you know, the 1700s, we had an inflection point in the population of the human species, from from millions of years through our entire existence, the scale of our population on this planet grew at a, you know, modest rate, right, a more regular rate, some stuff happened in the last three centuries, in the last few centuries, that caused us to skyrocket. And in this environment, where we're we're growing as a species, I think it's a responsibility for all of us to think about how we can create and benefit this world, how we can do things that create an environment for continuous flourishing of our species. And you know, there's something special going on with this anomaly in the universe, that is us. And this is a responsibility to engineers, to scientists, to entrepreneurs, to business people, to artists, to everyone that creates things, and everyone that puts hard work in engaging our neighbors each other. And that's what is an underlying driving mission and principle for all of the work I do. Now. I want to talk about a journey that I've been on in this environment in an environment where I believe strongly that we should do good in the on this planet. So so I'm a professor, thank you so much for the introduction. I'm a Professor of Computer Science and Engineering at University of Michigan. I've been on a crazy journey since I met my wife, Dr. Ling Tang, who's also a professor at Michigan, we've teamed up to take on hard problems and to cause a meaningful difference. Now, this journey started in grad school when we met each other and started doing research together. And then it went on into our professorship where we were working on the problems we believe to be the most meaningful problems on the planet for now. And in the next 10 years, how do we solve the problems that are preventing us from realizing human potential in the next decade, we've done some phenomenal work in the academic realm pre clink, right and up to 2015. We were funded by Facebook and Google and Microsoft, we spent time you're contributing technology to their ecosystems to increase scalability to reduce energy costs, to reduce the cost of doing the processing that humanity needs right now. We were funded by IBM Watson to create the next generation of AI, we were funded by Intel and the National National Science Foundation to create vision applications like Terminator, which is awesome, where you could identify everything in the vision. And we we on this journey, we created Sirius, which was an open source, conversational system with vision and voice capabilities. Now, we did a lot of work, right? We wrote a lot of research papers. Indeed, we're both inducted into the Hall of Fame for research productivity, of top tier research in the most prestigious venues for computer architecture, Computer Systems Research, reasoning about scalable systems for artificial intelligence, etc. But that's great, right? Like we were doing all this great work. And, you know, we stepped back after I don't know the 30th paper and said, well, are we changing humanity? Are we doing something meaningful for the world And we realized that you have to take your conscious intellectual contributions and put it into the lives of humans on the planet. And that's why we created clean we started clink, to disseminate advancement of technology to the masses in the hands of people. The best way to do that is to create a company to commercialize yourself, and not necessarily wait for someone to benefit from your intellectual ideas and then decide to commercialize it, it turns out that 99% of the research from engineering scientists, engineers, professors end up not ever seen the light of day in actually improving people's lives, only about 1% actually get out there. So we created clink, to make a difference. Now clink was an incredible journey. And it happened in a strange place. It happened in Ann Arbor, Michigan, not the East Coast or West Coast nots, Silicon Valley, San Francisco, or, or the East Coast area, New York, etc. And under our leadership, it grew incredibly, we raised three rounds in a short amount of time, right as CEO, we did a seed your seed round 250k lab, phenomenal local investment group series, a with Dr. Capital, who cares about this region, deeply 6.5 million, Series B with inside ventures, we broke records for tech companies in Ann Arbor when it comes to the speed and growth of our contribution to the world. And we took on a holy grail problem. This is the problem that everyone cares about. conversational AI really is an amalgamation of all kinds of interesting technical and scientific problems to create one experience. This is something that Apple, Google Amazon. There's another one a missing Apple, Google Amazon, what's the other one, Microsoft has been investing heavily in and has created a lot of momentum to productize however, there we observed a stagnation in the market that we believed could be transcended with the next generation of technology. So we grew the company to 120 people. You know, under my leadership, we achieved over $10 million of revenue Arr, around $9 million year over year growth averaged 3.2x. It was a phenomenal creation in this ecosystem in the nRb ecosystem. So this demonstrates that you do not have to be on one of the coasts to do phenomenal work right out the gate, incredible recognition, you know, awards from numerous places, you know, most innovative CEO Best in Show at Bank ai finovate 2016. Best of Show accolades from the press, you know, Clint creates the most events, artificial intelligence, etc. And what caused this right and it was one key insight. And this is how I would describe it to a potential customer, right, I would say, well look, there has been an inflection of the science of how to do conversational AI, and how to do NLP in general. And this, this inflection point, this track this new evolution, when it comes to this, the style of AI used is not is not generating the value in human lives that it should be, it wasn't being uptaken by essentially the market at the time, and that was to go from a parts of speech tagging, computational linguistics approach the approach where you identify the nouns, verbs and adjectives in a sentence and create rules and grammars as to what the system can understand to a deep learning approach where where words wash over a neural network much like my three year old, my five year old learns language, without having any no notion of what nouns are verbs are using a completely different approach. And this is how I would explain it in general, but just give one one slide of tackiness, I'm just gonna get just I promise, just one technical side, the key insight really, behind the technology is recognizing the the epic disruption that the new scientific approach of embeddings language models that that embed in the way you represent words as numbers embed in that intelligence, right so simply in the encoding of the text, that is the stream of words, you can embed so much high fidelity knowledge. For instance, you can create a numerical representation for say, King and a man and woman and Queen where you take the numerical representation representation of King, you subtract men from that numerical representation and add woman and it will produce mean as an embeddings, or a point in that embedded space that is very close to Queen, basically, taking words and encoding them deeply with intelligence was the first step of this insight, good intelligence into words encoded in a scientific approach to do that, then taking that encoding, and then training neural networks, transfer learning, taking that already intelligent ones and zeros. And then training new models with that gives you an ability to capture a much higher bandwidth of knowledge and code many more patterns into this neural network approach that has been disrupting disrupting our field. I know you've probably heard the news, Microsoft recently licensed GPT, three GPT. Three is one such encoding, right? It's actually a thought it's actually 1500 times bigger than what was recently possible of popular perks. Right, it's a massive model, it encodes the entire internet, essentially. And that encoding is worth a billion dollars to Microsoft, this is changing the world and, and that is precisely what clink has adapted and innovated upon to create it's amazing technology, of course, you have to do things with with the dialogue management so that you can have a conversation and the AI like remembers where you are, remembers what you said, is able to reference back knowledge, it's already procured, so it's flexible, and you can heal conversations. So these innovations, these scientific innovations, is what allowed for a clink to actually have be able to deliver on the promise of a completely different approach to AI that transcends what was believed to be possible, right. And so we did that a great scale, we worked with some of the most sophisticated enterprises on the planet, we we close some of the largest deals of startups our size, indeed, when we were raising money, you know, from venture capitalists, they would be in disbelief that a company of 30 people can close a deal with a company of 100,000 people that is more than the initial investment of the company in one deal. And so we were doing that because this was a holy grail problem for society. And people hadn't seen what was possible yet. With the absolute state of the art in technology, we wrote dozens of patents, we wrote numerous papers, research papers, from the innovation that was going on in the company. And so when we deploy this to actual human beings, you know, if you think about Alexa and Google Home, just anecdotally, I know that my friends, you know, when they started using Alexa and so forth, everybody was using it for a while, and then it fell off very rapidly. However, when the conversational experience works, and you can deliver value, what we observed with our experiences launch to millions of people, is that month over month over month, more and more users were choosing to use conversational AI to solve problems, as opposed to the mobile app itself. This is a virtual assistant embedded in a financial mobile app. And people prefer transferring money with voice Hey, send 100 bucks to my mom, rather than tapping 14 times to steal funds transfer, because it worked. And it not only did it just work, it works every time. And so we saw an increase in how many times folks were even using it when you know how many interactions were they having with the virtual assistant over time, and the completion rate was 92%. And we're seeing this continuously, right. So after about five years of creating this contribution to the world, I've stepped back there's a phenomenal set of leaders right now at qlink. That's continuing to take the journey into the stratosphere when it comes to contributions to the planet. However, I stepped back, I'm working on a new technology that I believe will have a massive impact. It turns out, I learned a lot about challenges when it comes to building highly sophisticated, multifaceted artificial intelligence applications and limitations and what is possible. And so I have I was debating whether I should share it. This is early, right so we're planning to announce early 2021. But I'll give you a little tidbit. Basically, what we've created is a new computational engine for AI that that has a new programming language. This program was created over tons of learnings from prior experience. And I think that the languages that we use to express solutions to sophisticated AI problems is actually a limiter where essentially in this assembly code language or punch card era of ways of expressing solutions to competition. Problems with AI. And so this language, the unique contribution is that allows for something I call collective intelligence. More on that later on this language on this computational engine. We're working on a product. And this product is a platform for human growth. And it's gonna be direct to consumers. And it's built upon this novel technology stack. That's all I'm gonna say for now. expect an announcement q1 2020. But really what I want to dive deep in is, how did this happen? Why did this happen? I mean, there's 1000 plus conversationally AI startups in the universe, I mean, in the, in the cut in the world, and many of them are struggling, most of them kind of peter out and go extinct. Actually, many AI companies go extinct, right? So how do you actually create a vehicle to disseminate this kind of contribution from nnr an unlikely place and Arbor, Michigan, and in a way that was able to, to essentially beat out the incumbents so fast? Right? Well, there's a couple key elements that I think are critical, especially when you're in a unlikely place, like in the middle of the country. And there's a couple key principles you have to follow a, one of them is first you have to have the right attitude, right, you have to have the attitude of taking a leap, I've seen many folks treat a vision as a side project. But you really have to engage in like a life's mission, right? You have to have the mindset that this is a mission that will be achieved, and a conviction around your vision that's gonna drive you there, the expectations you have for where you're going to take your contribution to the world, is what will define your potential. I'll say that one more time, the expectations you have will define the potential of what you can achieve, right. So if you expect to touch every human life, that becomes your potential, if you expect to have a lifestyle company and make I don't know, 50k, a year extra, that's what you're going to achieve right? their expectations will define her potential. And so you have to, you have to have that mindset if you want to really achieve greatness. Now another key insight and another key principle that is used and the culture that was fostered at clink was continuous evolution and adaptation, you have to chase harsh environments to challenge and condition yourself, your company and your ideas. You cannot stay in an echo chamber, you cannot be safe in a realm that that you won't hear the the hard truths about the idea of the creation, the thing that you aspire to achieve, because when you start, you're you, you may start like the creature on the upper left of this picture, right, you might start as a little rabbit thing with teeth in that picture, but you might end up if you adapt quickly enough, perhaps the market requires that whale and you have to get to the whale as fast as possible. Now, if you try to avoid the realities that will condition you to be to realize your potential, then the evolution is going to be slow. And in that you may go extinct, right? If you evolve too slow, so you have to stay outside of the echo chamber, you have to really embrace the harsh realities. And concrete example is often I'll see entrepreneurs, essentially reject harsh feedback from venture capitalists. The thing is, you have to take all of the feedback and assimilate it this is gonna seem to contradict what I'm gonna say very shortly. But you have to take you have to take the feedback and assimilate it into your understanding of truth and what where that gap is between where you are and where you need to be, so that you can close the gap, take that criticism, you know, allow hard truths, to force you to challenge yourself to do better. That's incredibly important, concretely on investors and I mentioned it's gonna seem to contradict, you'll see, investors will notice you no matter where you are, look, do not tell yourself or use as an excuse the fact that you're not in the valley as to why you are having hardship. Because that's not reality. And we can tell ourselves things to make us comfortable. But and that's nice short term, but you will not achieve your goals if you do that, right. The reality is investors only they will look anywhere to see the lower right of the screen to see up into the right, right so it doesn't matter where you are, you could be in Timbuktu, right I don't even know where to look to it, but you know what I'm saying like you can be anywhere on the planet and If you show demonstrable promise with that you will attract the interest of investors. But in order to engage this journey, especially with adaptation in mind, when you engage investors, and this is like super important. This comes from lots of learnings, like a lot of folks will say, I want an NDA with the investor, I want to keep my thing stealthy. I don't want people to steal my ideas, you got to get rid of that right away, and share so that you can get the feedback that allows you to adapt, right? And the best investor relationships never starts with a pitch, I like it money and never go like the best relationships don't go like that. The best relationships are more of the style. You get to know the investor, you, they coach you, you adapt a bit you change and you build a relationship over time, I might be running up all my time limit. So I'll just run through...
You're plenty good. plenty good.
Thank you so much. Thank you so much. And the mindset of the special remember that and this is the contradictory part. Remember that investors look for patterns, they look for hump of the curve patterns that will lead to company success. That's their job. That's their conditioning, they're conditioned to find the patterns for predictable success. However, for you to be a successful startup, here's a contradiction. You have to diverged from the masses, you have to divert from the pattern. So that's a contradiction. The thing is, you will get lots of advice from many sources, but you have to integrate that with what you know to be true, right? What you know, to be right. Even if it flies in the face of your investors, you have to pursue, I'll give one concrete example. My seed level investor, they really believe that I need to get letters of intent talk to customers, I went off and said, No, we're gonna go stealth mode and build something that no one has ever seen before, we spent nine months doing that. And that's what led to all of those awards that I showed her earlier. However, during that nine months, they were like, I can't control the CEO, like I, you know, we gave him the directive to go talk to customers, before you have a product before you have anything to show, you know, try to get excitement, show letters of intent, we want to see indicators that this is good. And I was like we need to focus on building something no one has ever seen before we did it, it was massively successful and raised a massive round after that success. We went with what we know I you know, we saw in the market, that there's too much noise, everybody's saying the same thing. We have to show it because nobody's showing your insights, let that drive you. Coming up to the end here, pitfalls and learnings. So there's some things we didn't do? Well, I'd click and there's some things I didn't do well, as a leader of click, right. First and foremost, there's qualitative and quantitative company building, there's a qualitative mindset and a quantitative mindset. Early on, it's very tempting to be qualitative. There's 10 people, you can help everyone with everything. And then, you know, you just jump in and help Oh, I can help coach you, etc. But metrics is going to allow you to scale so that you do so that you can operationalize your company properly. I can get into that more offline, if anyone's interested in chatting more about that evangelical versus scalable, recognizing what success is, you can have success, and it may not be scalable success, you won't have success, because it was an evangelical, kind of sale angelical relationship. So you're leading folks into a new market that they don't understand. However, you may not be able to get 20 sales people to lead people on a journey that they don't understand, understanding the difference of what's scalable, from your go to market product fit. And what's evangelical is very important. And I believe that we're in a very Ilana angelical phase, when you're working with massive Big Whale enterprises, but not so much when you're working with developers and the people that are actually trying to build stuff. Institutional organization, leadership and culture. I had a I left the company with a very bad mantra. I my motto was not work friends, real friends. Early on, this was great, right? Because everyone is a group of friends that are successful, so rapid, our scaling of the company was so rapid that I didn't keep up with that perspective, right? So I was like, No, we can all be real friends. throughout this entire journey, I believe in everybody. However, as you scale as a business, your numbers of people will scale and solo, the Delta, between the leadership, executive leadership and the rest of the company. If you're, if your buddies had four people with an engineer, and you're just great friends with an engineer, four people when you're 130 people and you're the leader of the company, your engineer is still an engineer, but you're something very different, right? And that creates an environment where you have to adapt. And you know, I used to have the belief that Well, you know what we have to do reduction in force. I've got to do demotions, I should do it myself. We can still be friends. But it's you have to adapt that perspective. And so those are important pitfalls to our slides. Also remember that the iPhone was not the first iPhone. The palm Treo was the first iPhone apps, icons App Store identical. Palm Treo, failed iPhone succeed. Your objective is to have the insight to know why. And I'm not gonna answer the question. I have a hypothesis but asked me offline. But understanding the why that insight is going to allow you to know timing and when you will succeed with an idea. And I want to wrap up with this slide. In just the last few years of our epic growth, we discovered electricity in like 1900s 1960, something whatever space in computers was in the 1960s, internet came 1980 to 19 to 2000 is when when Netscape was invented, that's where I put the arrow there when it became popular AI just recently, our progress as a species is accelerating, and at a very fast rate. So who knows what the next 50 years hold, and I'll just say, find your place in that next 50 years and contribute to the growth of humanity. Thank you very much.
Thank you so much, Dr. Mars that was absolutely incredible. I think everyone here and everyone listening, we're just in awe at the energy and transparency and insights you brought. That was phenomenal. I'm so excited to hear more about what your stealth company is. Next year. I'm so excited to stay in touch, learn more, and just hear more about, you know, your your fascinating approach to life. appreciate it so much. Thank you. Thank you.
Thank you so much. It was a pleasure.
Yeah, absolutely. Take care.
Yep. Bye Bye everybody.
Up next we have Edie shiner, lead copywriter at a very well known brand jeetu. If you're in the SAS software space, there's 100% chance you know what g two is? Eddie shiner. Up next, also co founder of what is it? What co founder what's Eddie's co founder? What is Eddie, the co founder of burger co founder of very good coffee. Sorry, Eddie. Up next, Eddie Shriner. See you soon.