Artificial intelligence guru Gil Syswerda is back to talk about angel investing and to give practical advice to founders. We closed with entertaining stories from Gil’s career.
Highlights:
AI Guru Gil Syswerda Is Back to with Tips for Angels & Founders
Why Founders Should Value Their Time at $10 per Minute
“...I stopped driving...I take an Uber or Lyft because it's not worth my time to be in a car... An Uber is way cheaper than $10 a minute.”
Advice to Founders: Find a Co-Founder with Critical Skills You Lack
Advice to Founders: “You need to be in the trenches, along with everybody else.”
“One thing you make a must-read is an article by Paul Graham. It's called Maker's Schedule, Manager's Schedule.”
“...The Hard Truth About Innovative Cultures, Gary Pisano just nails it, when he talks about how to create an innovative culture...”
“...get an executive assistant as soon as you can.”
“...you just cannot function at 100% if you don't get enough sleep.”
Exercise Is Essential: Intense Outdoor Exercise at Sunrise; plus do Bouts Like Taking the Stairs
“It's actually quite difficult for most people to make money as an angel investor.”
“...I believe that Jeff Arnold probably does pretty well on his investments but he has pretty grounded knowledge in the biotech space and he has a methodology for evaluating companies.” How to Make Money in Biotech with Jeff Arnold
Frank Ferguson Was a Spectacularly Successful Investor (Bose and Curriculum Associates) by Focusing on a Few Investments Intensely Practical Dreamer with Frank Ferguson
Howard Stevenson Got Warren-Buffett-like Returns on His Angel Portfolio [Wealth & Families with Howard Stevenson]
Sal Daher and His Biotech Screen for Angel Investing
“These people [academic founders] have tremendous skills. If we can help them turn those skills in the direction of something useful, I think they can create a lot of value.”
VistaPath Bio Using Off-the-Shelf Machine Vision to Create Value in Pathology
Gil Syswerda Stories
The HP Interview or Everything I Needed to Know I Learned in Third Grade
Growing Fish at BBN Labs
The Interview at Lawrence Livermore Research Laboratory
Gödel, Escher, Bach & All That
The AI Researcher and the Daughter of the Soviet Scientist
Simulating Professor Paul Scott
How to Have a Board without a Board
Gil Syswerda’s Perception of Where AI Is Today
Gil Syswerda Highlights People Who Were Influential in His Success
ANGEL INVEST BOSTON IS SPONSORED BY:
Transcript of “Angel Investing”
Guest: Gil Syswerda
Sal Daher: This podcast is brought to you by Purdue University Entrepreneurship and by Peter Fasse, patent attorney at Fish & Richardson.
AI Guru Gil Syswerda Is Back to with Tips for Angel Investors & Founders
Welcome to Angel Invest Boston, conversations with Boston's most interesting angels and founders. I'm Sal Daher, an angel investor, who is very curious to learn how to better build startup companies. Today, we are privileged to have repeat founder and angel investor, Gil Syswerda, on again, my Walnut buddy. Welcome, Gil.
Gil Syswerda: Thank you, Sal. It's a pleasure to be here.
Sal Daher: This is the third episode that we're doing with Gil. The first two episodes, we focused on the companies he founded. The second episode we focused on things going on and AI. In this episode, we thought we would focus more on angel investing and a little bit on advice for founders. Gil, if you listen to the previous episodes, I highly recommend it, he's a highly experienced founder in artificial intelligence and he has done that in the conjunction of artificial intelligence and the investment world. He has been working with world-class investors such as Paul Tudor Jones, and his work is groundbreaking and revolutionary and it caused a lot of interest.
I can tell you that Walnut turned out en masse to watch Gill talk about one of his ideas that he had for this genetic induction algorithm that he's worked on for a long time. Anyway, Gil, what's the advice you find yourself just over and over telling startup founders, listen, this is important?
Why Founders Should Value Their Time at $10 per Minute
Gil Syswerda: Sal, I'm going to talk about what I find important in a startup. This year, some hard lessons that I have learned over the years because we all go through a learning curve. I think the first thing I'm going to talk about is how to value your time. There's lots of different kinds of startup founders. We all have a goal of creating a successful startup and then eventually having an exit. There's the morass of details you got to take care of on a day-to-day basis and if you lose sight of really why you're there in the first place. It's actually relatively straightforward to give a value for the value on your time and that value is $10 a minute.
While you're doing a start-up, your time is worth roughly $10 a minute. Now, some of your listeners are going to think, that sounds like a lot. A few might say, really, only $10? It's easy to get to that number. You just do some basics, let's say three founders. You just set aside stock for employees. Now, you’re going to get some investor dilution. You'll see you have a nice success, sell the company for $50 million after five years; put in a lot of hours. You as a founder will get roughly $10 million at the end of all that before taxes, and you do the math, you will have spent roughly 1 million minutes achieving that $10 million.
[laughter]
Sal Daher: You got to get to a million, you are going to a million minutes, even I can do the arithmetic, it's $10 a minute.
Gil Syswerda: Exactly. Put it on just a little bit of a different perspective. If you're the type of person who can create a successful startup and have a nice exit, you're actually probably able to get a job that pays at least a million dollars a year. There's lots of examples of that in big tech and finance and so on. If you're working at a million dollars a year compensation, you're working more or less normal hours, you're already at $5 a minute and you're getting paid vacations. And a lot less stress, and a lot less risk. Going to $10 a minute isn't really that big of an ask.
If you're working hard and you are doing $10 a minute, those thousands of dollars a day should just round it up to $10,000. My advice to you is figure yourself is worth $10,000 a day and every day when you get up in the morning, your thought should be what can I do today so that in the future this day will be worth $10,000 to me. That just needs to be the mantra. That'll focus hopefully, your attention on the future so that the future brings that $10,000 to you for the day that you're about to do. If you internalize that, you'll find yourself making some different kinds of decisions.
“...I stopped driving...I take an Uber or Lyft because it's not worth my time to be in a car... An Uber is way cheaper than $10 a minute.”
For me, I stopped driving to go to many places. For years now I go into the office, I take an Uber or Lyft because it's not worth my time to be in a car and you have traffic and parking and so on. An Uber is way cheaper than $10 a minute. The second thing is, you need to really be unrelenting in your goal and your goal is success. There are so many other things you could do with your life. You're doing a startup is hard, feels stressful, has high risk, you just need to keep your eye on the ball if you decide to do this thing and you should just be unrelenting in that goal, make this company succeed.
Advice to Founders: Find a Co-Founder with Critical Skills You Lack
You're going to have to learn about what you can and can't do well, that's a hard one. Because if you're just starting out, you don't really know what you can do. You don't really know until you're in the thick of it, and have to do some of these things to know whether you're any good at it or not. This is where the question of co-founders come in. Because if you're not good at, let's say, sales or you don't think you can become good at sales, you need to have some backing on that with another co-founder.
If you're young, get lots of experience, do lots of lots of stuff, make the stuff hard, and try to try to get a job, somewhere along the way, we had to sell stuff, that's invaluable. I would recommend being very hands-on, I'm a very hands-on person. To give the Optimax example, I went on every single sales call for the first two and a half years. There wasn't a sales call that happened that I wasn't there. It's invaluable because you learn what customers want.
Advice to Founders: “You need to be in the trenches, along with everybody else.”
I have talked to plenty of founders and worked with some founders who really want to take a high road and think lofty thoughts and have other people do the work, I don't recommend doing that. You need to be in the trenches, along with everybody else. You probably want to read a business book or two, it's been helpful to me learning things. I'm not going to recommend a specific book, because there's lots of good ones out there. You really have to find a couple of books that are going to resonate with you, but I do think that's important. Maybe a management book too, especially one that focuses on the psychology of people inside companies.
“One thing you make a must-read is an article by Paul Graham. It's called Maker's Schedule, Manager's Schedule.”
One thing you make a must-read is an article by Paul Graham. It's called Maker's Schedule, Manager's Schedule. What Paul does there is he lays out the different personality types, for people who manage and people who make things, your technical people. You can make employees unhappy in ways that they don't understand if you don't understand this distinction. Also, if you're an extroverted type, you should read this thing every week, because my experience with extroverted people is they can't internally model what an extrovert life is like, and they forget. You need a constant reminder.
Sal Daher: The introvert’s life yes, the extrovert can't get into the introvert’s head.
“...The Hard Truth About Innovative Cultures, Gary Pisano just nails it, when he talks about how to create an innovative culture...”
Gil Syswerda: Yes. They can add their head and they can toe the line on a rules-based thing for a week or so and then they forget. I would also read an article in Harvard Business Review by Gary Paisano. He's at Harvard, this is called The Hard Truth About Innovative Cultures, Gary Pisano just nails it, when he talks about how to create an innovative culture, there's some do's and don'ts. What you really want is to be brutally candid inside your company, but also psychologically safe. He talks about how to make that happen. That article really resonated with me, because this is how I run my companies in the past, to create an innovative culture.
Hire people who are smarter than you. This comes along with innovative cultures, but also, some people are afraid to hire people that are smarter than themselves. What you really want is you want to have really smart people disagreeing with you to the extent that they're right, you accept their disagreement and change your mind, that's hugely important. Nobody is smart enough to know everything. You may also hire the smartest people you can possibly find.
“...get an executive assistant as soon as you can.”
Another is get an executive assistant as soon as you can. It's so important because there's so many details in the startup. If you don't have that person, an executive assistant isn't really quite the right term. You'll figure out what their title should be. For instance, in FeatureX, we had Emily Rogers, who was awesome. Emily basically ran our recruiting, often had a couple of temps reporting to her along the way, did all the travel arrangements for candidates, she put together the booths and ran the booths at conferences and at job fairs, and so on. She was basically our HR department.
Sal Daher: Yes, for small contracting for what they call the business manager. Instead of an operations person, a person who does the recruiting and hiring and all that stuff, when you're really small, and you don't have an HR department. You don't have a marketing department and you don't have all of that. Someone who's very high functioning on top of things, and can coordinate and keep lists.
Gil Syswerda: Can just handle many, many things. You get it.
Sal Daher: A highly capable multitasker.
Gil Syswerda: Right. If you don't have that person, then you end up doing all that stuff yourself. It becomes a big time sink. Just a few other things I know why you want to wrap this section up but use some discussion forum software because Slack is pretty good, that's the glue that can hold everything together because everybody is aware of everything if you have something like this. Also, you use some knowledge bases software like Influence or something like that, you start building up a repository of documents and decisions made, and task lists and completed items and so on. Run the business like you're about to be acquired, you can be really innovative and your products.
Sal Daher: This is really, really important because a lot of times, people, when they're about to get acquired, they discover, oh, my code audit. "I'm not going to pass the code audit because I have stuff in my code that's not properly documented or properly acquired. I don't have rights to it," and so on and so on.
Gil Syswerda: Yes, that's a big one, also, just make sure your finances are in good order. Use an accounting package and have a bookkeeper because their life becomes so much easier. When somebody wants to look at your finances all you have to do is run a report. Same thing, by the way, is keeping track of all your legal documents, all the contracts you've entered into and so on, that's also important to keep organized. Then the last couple of things, get enough sleep, you just cannot function at 100% if you don't get enough sleep.
“...you just cannot function at 100% if you don't get enough sleep.”
Sal Daher: Yes, human beings have natural limits that they cannot exceed, and your productivity goes down. You think you're producing, if you're shortchanging yourself with sleep, you think you're getting all this extra time to work, all your doing is doing lousy work.
Gil Syswerda: Yes, you're just becoming less, and less effective and there's a concept of sleep debt too. If you have a five-hour night one night and you're a seven-hour person, you've got to make up those two hours, they just don't disappear.
Sal Daher: Oh, absolutely.
Gil Syswerda: Then the last thing is get enough exercise, it's actually going to be hard to get enough sleep if you don't get enough exercise. You're just going to enter into this downward spiral.
Exercise Is Essential: Intense Outdoor Exercise at Sunrise; plus do Bouts Like Taking the Stairs
Sal Daher: Preferably, if you're in the Northern climates, in the morning, get out there with the sun in your face so that you're going to sleep better. Exposure to sun early in the morning leads to deeper sleep, better sleep at night. Get out, get some fresh air, 20 minutes, 30 minutes and some intensive exercise, the most intensive exercise you can do given your age, and that's going to clear your brain, it's going to do a lot for you.
Gil Syswerda: Even simple things like take the stairs instead of the elevator, go for a walk around the block while you're having a meeting with somebody, as opposed to just sitting in chairs, for example.
Sal Daher: Those are all helpful.
Gil Syswerda: That's my advice to start-up founders.
[laughter]
The basics.
Sal Daher: That was really fast, you're getting your workout today with that one. How about for angel investors, what would you like to get across to angel investors?
Gil Syswerda: First, I think if you're launched into this, I don't want to talk anybody out of becoming an angel investor, because there's going to be some negative things here. There's lots of reasons to be an angel investor, it's fun, you get to meet a lot of people, you're giving back to the community, and so on, but don't really do it expecting to make money. It's actually quite difficult for most people to make money as an angel investor.
“It's actually quite difficult for most people to make money as an angel investor.”
Now, I wrote an article on this in Medium, you can look it up, where I work through the math on that. That's been verified by several people, some hedge funds, and somebody who's invested in hundreds of startups, and basically said the math is right, this is the results you get. Just a quick little example, let's say you have $100,000 portfolio, and the way you invest, the typical company you invest in for 50% of the time, you lose all your money.
For 25% of the time, you get your money back but that's it, for another 20%, you get three times your money back and for 5% of the time you get 10 times your money back. Those are pretty good stats, nobody's ever pushed me back and said those are really pessimistic stats. With that kind of stats, on an average investment, you make you will lose money. The reason is, first of all, if you make a lot, a lot of investments, let's say you made a million of them, then your average return is going to be 35% per annum. That happens over 5 years 10 years, it takes a long time for startups to return something, on an annual basis, that's not a very good return.
If you make just one investment, you have a 50% chance of losing all your money, one's not good, a whole bunch isn't really good because you also just put your money in the bank. If you do 10 investments, there's only a 65% chance that you'll actually get your money back, the odds are really against you for making money. You can get lucky and hit a home run and invest in something that turns into a unicorn, you make a 100 extra in your investment, and that does happen, so you can get lucky.
It's hard to think about just ordinary angel investing and to make money, you can have a lot of fun along the way, and hopefully you get all your money back but you're not going to become wealthier necessarily unless you have an edge.
“...I believe that Jeff Arnold probably does pretty well on his investments but he has pretty grounded knowledge in the biotech space and he has a methodology for evaluating companies.”
You interviewed Jeff Arnold [How to Make Money in Biotech with Jeff Arnold] in the space of biotech. That was just a fabulous interview, and I believe that Jeff Arnold probably does pretty well on his investments but he has pretty grounded knowledge in the biotech space and he has a methodology for evaluating companies. Then he picks a few and then he tries to help them out because he has expertise. That's great. He's going to completely change how I do angel investing because I'm just going to invest in what he invests in now. I'm just calling it that.
[laughter]
Sal Daher: He's very guarded about it. I tried to invest in one of his companies. He wouldn't let me in.
Gil Syswerda: Yes. I've never actually met him, but I would like to.
Sal Daher: Oh, he's great. Yes. He is a great guy. This is a little bit like Frank Ferguson. Remember our Walnut colleague, Frank Ferguson [Practical Dreamer with Frank Ferguson] Who's been in the podcast once. He's passed away. An amazing guy. I don't think he invested in more than 15 companies in his whole life.
Gil Syswerda: It seemed to me like he always had his checkbook out, though.
Sal Daher: Yes.
Gil Syswerda: There's a real dichotomy there somehow.
Frank Ferguson Was a Spectacularly Successful Investor (Bose and Curriculum Associates) by Focusing on a Few Investments Intensely
Sal Daher: There were not a lot of companies, a lot of checks to the same company. For example, I was a co-investor with him in Senscio Systems, which is artificial intelligence for treating people with chronic illness, for monitoring people with chronic illness to keep them from being readmitted to the hospital. He wrote a lot of checks to Senscio. I wrote one check. [laughs] He bankrolled them for a while and they're still alive and kicking. He had a handful of things. Investments he made in Bose, originally. Investments that he made in Curriculum Associates, and then he jumped in to get it off the ground and so forth.
Howard Stevenson Got Warren-Buffett-like Returns on His Angel Portfolio
He made himself a huge fortune on these companies that he was very hands-on with. He had a little bit of the Elon Musk invest in a handful of things and be really hands-on, really involved. Instead of being a detached investor in hundreds of different ventures and he had success with it. Jeff is a little bit along that line. I also interviewed Howard Stevenson [Wealth & Families with Howard Stevenson] who, over the years, has invested in several dozen startups over 35 years, and he did a very careful analysis of his angel investments. Now, keep in mind, Howard Stevenson is a very prominent professor at Harvard Business School. He was one of the co-founders of Baupost.
He had a consulting practice and so forth. He had a very, a really excellent deal flow so to speak. Things that he came across were very interesting. His experience was that over 35 years, he generated an internal rate of return of about 17%, which is Warren Buffett territory, but he said that it was extremely lumpy. If he took out one investment that was a 400X from that, the return for the entire portfolio drops to 12%, internal rate of return, which is still really attractive, but it's not 17%. That one investment that he talked about was a 400X at the time because he was still invested. He sold some of it and he still invested.
Was a really very mundane business he thought it was a stupid business proposed to him by some really smart people. The company's Asurion and what they do is they ensure cell phones. [laughs]
Gil Syswerda: What happened?
Sal Daher: This is the most mundane business, but at the time, I don't think there were any competitors. They managed to establish, to run it brilliantly. It has been his most successful. It has made his entire portfolio. In a way, what angel investors are always looking for is that whale that makes the whole portfolio. There's going to be a handful of these complete outliers that will make the portfolio really shine. I think what they need to do is to, they need to avoid the obvious losers, the ones where you can tell that the people are not going to be able to raise money. That's a really easy criterion.
Don't put your money where people are not going to. It doesn't look like these guys are going to be able to get funded. My personal experience has been that I've invested in about 60 odd companies. I don't have enough backup that I can give a really exact calculation but the rough numbers that I have is that I have the software and non-life science companies. The non-life science companies and the life science companies. Some of them are software in life sciences. It's got a little complicated there.
Sal Daher and His Biotech Screen for Angel Investing
Generally, in the non-life science companies, I haven't had anything better than a two and a half, three X return. I've had a lot of losses of that the 67 companies, 68 companies I've invested about 29 of them were life science companies, and of those I've had six complete losses, six, maybe seven there's one that probably that they haven't said so, but [chuckles] probably dead. They don't want to give up, but the reality is that all the value of my portfolio is in the life science companies and why? Because they have patent protection that keeps out competitors.
If the technology works and there really is a market for it, then the chances are that you are going to have something that can build value over time. I'm completely focusing my investing and I've created a target for the type of company to screen for the type of company that I'm looking for. I'm looking for, number one, an academic founder who's really top technically, but who has the potential to become a founder and a CEO. Why? It's really important. These early life science companies are going to be under-resourced, very hard to bring in outside talent. You need an academic founder who can be more than just an academic founder, who can be some kind of a CEO.
Gil Syswerda: That can be a big ask from academic people.
Sal Daher: Believe me, that is a huge ask because a lot of academics are really impractical. I find frequently that immigrant founders, immigrant academics are a lot more practical [laughs] than people born here. They been through situations in their lives. They haven't had with a so typical in academics circles, which is a perfect career with those stumbles. This person is absolutely brilliant with top of her class in high school, in college, the graduate school, published a lot of papers, got to be never, never encountered any adversity, people who are immigrants tend to along the way they've had the deck reshuffled on them a couple of times.
They had to bounce back. Number one, for me, the founder's really important. Number two, the technology, it has to be a technology that works. This is where Jeff [Arnold] comes in, founders don't know what it means for technology to work, academic founders. Something that works in a lab is very far from working in industry or in a clinic. There are a lot of things that work in the lab that never up making in the clinic.
Then it has to be protected by patents in an area where there are strategic players that really care. Where it's going to be consequential for their bottom line. Otherwise, that startup is not going to be worth anything. Then the last criterion is that it can't take more than $4 to $6 million of invested capital to get that biotech startup, to the point where a strategic player is going to come in and care.
When it comes out of the lab is so wild and wooly that the strategic, "Oh, that's a science project." As Michael Marks likes to say, "That's a science project." You have to de-risk it to the point where yes, this thing really looks like it could work and, "Oh, wow, this could change our bottom line." Then they have a chance to do a collaboration with a strategic player, maybe get acquired, maybe license off a certain aspect of the technology and develop the rest, and so forth.
I had this model in my head that I developed along the way of being like I led the angel round in SQZ biotech, SQZ Biotech, which is now public. At one point, they were up 11 X, 12 X they're down now from where I acquire, it's still compared to what I paid for it, it's still several X up, but it is a public company. As a matter of fact, I think it's an interesting buy at the moment. I think they have like nine things in their pipeline. I don't want to give people advice, but it's just my opinion that I think that the stock at $12 a share is underpriced, because they have a lot of things that could be extremely valuable, but this opinion is not unanimous.
Among people that I know, there's some people who I respect very much who don't agree with me who think their technology there's nothing special there, blah, blah, blah. I'm not going to go into the details, but there are other people who do who there are also very smart. Anyway, I also helped develop this idea when I invested in Savran Technologies, which is a platform, and so the SQZ and Savran are both platforms.
They have a lot of different uses. If you are successful in developing one use case, then you can help fund the other use cases and that helps build value in the company. That's another aspect of the thing that I look for. I think with that screen, I can make money as an angel investor focused on life science companies, but I have to have the discipline to only invest in the companies that you can get somewhere with 4 to 6 million in funding because some big VCs are not coming in for these guys.
Gil Syswerda: Then you have your FDA approval. You have [unintelligible 00:25:33]. I guess in you're looking for an exit before those approvals are required or they already have acquired access to technology that's already been approved.
Sal Daher: Let's look at the portfolio I have right now. SQZ has a bunch of things that are in clinical trials right now. How did they get there? They got there because they got $125 million in non-dilutive funding, from Roche Pharmaceuticals, from a collaboration Roche. What they're doing for Roche is their manufacture the doses of these cell therapies for cancer that Roche developed, but which SQZ can manufacture faster. It used to take three weeks to produce the doses to treat a patient.
They take blood from the patient and then they process that. They engineer immune cells to help attack cancer. When Roche is doing the initial process, it took them to like 21 days, SQZ was able to do that in 24 hours. If you have someone who's in a cancer trial, they're usually terminal patients. That time is really valuable. This is like, for me, the model. They're not coming out with a therapy. I don't have enough capital to invest in therapies. I have capital to invest in a technology that something that can be printed up in a 3D printer which is somehow going to do something that's going to 10X somebody's bottom line so big pharma then they'll pay for it. That's the key.
The key is figure out where you can have this enabling technology that's high value to someone, that's the guess. Or a technology that's underdeveloped. Let's take QSM Diagnostics, for example. These guys raised one and a half million dollars. It would've been just 1 million if it weren't for COVID but then they got delayed and everything, but they're basically heading towards becoming profitable after having raised $1.5 million for a rapid diagnostic. Within two minutes they take a swab of a dog's ear infection and they can tell if there's a bacterium present, is called a Pseudomonas aeruginosa. If it's present, they need antibiotic treatment.
The dog needs antibiotic. You don't want to be giving antibiotic to dogs needlessly. What these guys are doing is they have a path to a profitable business with $1.5 million in funding, academic founder with some helpers who came along the way who have business experience, and so forth. Gil, I'm not the only one working on this. Tony Kulesa at Petri, they're associated with Pillar VC in Boston.
He wrote a post recently and there he is doing a conference online by the time this podcast launches at the conference would be held. It's the idea of the founder CEO, of the founder being the CEO of the company because academic founders, these life science companies, they can't afford to hire a professional CEO, three-quarters of a million dollars. Somebody with experience with those what that person's doing that person is--
Gil Syswerda: Well or a significant equity stake.
“These people [academic founders] have tremendous skills. If we can help them turn those skills in the direction of something useful, I think they can create a lot of value.”
Sal Daher: Significant it can happen, okay? I've seen CEOs being seduced to become founders but it's really hard. What I'm trying to do is I'm trying to help angel investors who are not necessarily disposed to investing in the life sciences to become more conscious of what it is that life science founders need, what these founder/CEOs need, and how to support them in what they're doing. These people have tremendous skills. If we can help them turn those skills in the direction of something useful, I think they can create a lot of value.
There's a lot of support for it. I-Corps and all these things. Then a little bit of advice along the way I saw Ed Golic when he first pitched at Walnut, and then when I interviewed him like 18 months later, he had really sharpened his approach. He was looking for six sensors and doing a lot of stuff, he was getting raising money to pay a full-time CEO.
At that time he had quit his teaching job. He was fully committed and he was developing a product that required just one sensor. They identified instead of urinary tract infections, they were shooting at dog ear infections which are a more achievable target. I think I've got a dozen of these companies in my portfolio now that fit that screen. I think that is, for me, and living in Cambridge as I do, I'm surrounded by companies like that. I also have a connection with Purdue University, which is a sponsor of this podcast, Purdue Entrepreneurship. Purdue has a lot of life science companies as well. I see these companies that are doing stuff that's really close to the ground, a few million dollars will get them really far.
Gil Syswerda: That sounds great. You have a methodology, you have domain knowledge, you know people who you can bring in to assess these things, so that's a pretty powerful combination.
Sal Daher: Oh, speaking of that, speaking of the main knowledge and bringing people to assess, that's one of the big differences from the usual type of angel investing that we see where you see a software company, you can know what business they're in, and you already have a mental model of what that company is doing and what you're really gauging as whether the founders are really committed or not, if they're capable, they know what they're talking about, does the need really exists, and then you take a bet on them. With a life science company, you need to do a lot more work.
Even if you narrow your focus to a certain area, there's a lot of studying that you have to do to understand. You have to bring in people who understand the field to look at it and see if that technology can work, and then you bring in other people, and they're not the same people. They're going to be totally different people who understand what the market is.
These might be founders on adjacent spaces that you're friends with and they'll give you a direct opinion on the stuff. Then you can get a sense. The due diligence for a life science company takes a lot more time. I would say easily 10 times more time and more effort than due diligence or at least that I've seen people do on these software-based companies.
Gil Syswerda: A similar thing in AI-based companies, because AI nowadays is, in a lot of ways, it's gotten a lot easier because deep learning and open-source frameworks and so on. You can put together some demos and tackle some customer problem. It all looks good on the surface, and maybe even have customers who are willing to pay for a prototype and so on.
Then when you find out what the customer's true needs really are, what the marketplace is really expecting and it's outside the bounds of what you can download for free, the models or the frameworks, then suddenly, the AI gets really, really hard. Usually beyond the scope of a typical startup. The devils in the details on these things. It's very hard from the outside to spend enough time sussing out just exactly where those pitfalls are.
VistaPath Bio Using Off-the-Shelf Machine Vision to Create Value in Pathology
Sal Daher: What’s AI and what's just a very mundane tech? I recently invested in a company called VistaPath Bio. What they're doing is very, very mundane stuff in machine vision, but in a very high-value situation. They're basically using machine vision to get the size of pathology samples, like a piece of a prostate that's been taken out by a doctor. It's not rocket science. It's just how big it is, what shape it has.
Gil Syswerda: I looked at that company.
Sal Daher: You did? It's not an AI company. It's a workflow company, but they're doing really well right now. They've gotten a lot of traction because they solve a big problem for pathology labs, reducing labor, and so on. It's a very mundane application, and there are thousands of those. Those are not that difficult to evaluate. Not taxing. This is not an AI problem. A little bit of this stuff can go a long way. It's not a deep technology company.
Gil Syswerda: I'll tell you a little story, a year or so ago, I was asked to look at a startup that uses machine learning. I had a conversation with them and tried to do a deep dive into the machine learning and see what they were really doing. What they were doing is they had access to a set of data, and they'd run a database query across it and pull out some values, add them up, if a value is greater than some number, they did one thing.
Otherwise, they did something else. That was their machine learning, an if then else statement. You could even make an argument, "Okay, I guess maybe it's machine learning, depending on how they got the values or so," but wow. That's a big, big stretch in calling that machine learning and AI.
Sal Daher: There are cases where just a little bit of that stuff can be very useful. It can save errors, it can speed up processing. This is intake of pathology samples, VistaPath, is the intake of pathology samples. They're not doing What PathAI does, which is, the end of the workflow, which is to look at the pathology samples and decide Pacelle is cancerous or not, and so forth. Basically, doing pathological analysis, what they're doing is just like a workflow. I think that is very within reach what they have.
Gil Syswerda: Yes, if it's that understandable, right?
Sal Daher: Yes.
Gil Syswerda: Even if they use "machine vision", all they're doing is checking to see a pick on blob is on a slide, and that provides value to a customer, then that's an easy ask. There's a whole range, there are some really sophisticated things you can do with AI, and of course, you can develop even more sophisticated things.
Sal Daher: PathAI is doing a lot of that. They're hoping to get to the point where they're better than a pathologist at recognizing cancer cells or different types of cells, and they've raised like $160,000,000. They bought Poplar pathology, which is a well-respected lab. They're going to be taking that AI and applying it there. I imagine that could be economically a very powerful thing.
We've talked about advice for founders, talked about advice for angel investors. I have a list of things here, I understand that you did not want to get through this without getting through the HP interview. Would you tell us the story about the HP interview?
[crosstalk]
[laughter]
Gil Syswerda: The story?
Sal Daher: Yes.
Gil Syswerda: Okay, are we on the storytelling time, because--
Sal Daher: Yes, let's do some storytelling.
Gil Syswerda Stories
Gil Syswerda: -you didn't ask me my BBN story; I was waiting for you to ask me.
Sal Daher: I wanted to get to the HP story first, then we get to the BBN story.
The HP Interview or Everything I Needed to Know I Learned in Third Grade
Gil Syswerda: All right. it's a cute story. I was doing campus interviews at University of Michigan, and I got invited out to interview at Hewlett-Packard. At that time, I wanted to work in genetic algorithms AI, but also a lot of other stuff, parallel computing, computer graphics, operating systems, and so on. I was being put through the wringer at Hewlett-Packard from group to group to group, and in the middle of the day, I see a sea a cubicles. I get led to a cubicle, and there's a woman that has a double cubicle with a desk and a table.
She says, "I'm really busy. I'm so sorry, but I'm just having to just give you a programming problem." Puts me at the table with a blank sheet of paper and a problem statement. Says, "Just sketch out on the blank piece of paper what the solution is." All right, it's in three parts, the problem. Part One, is build a certain data structure. Part Two, is build this other thing, it has some methods on it, and then Part Three, combine the two into a something that can compute the following. All right, so I read all three and I go, "Huh, I don't understand it." Feeling like a complete idiot.
[laughter]
I don't know how to do problem three. I'm going to pause right there, I'm going to jump to third grade, Mrs. DeVries where she gave us a pop quiz one day. Oh, my God, pop quiz. Everybody hated pop quizzes, right? Something like 10 questions on this quiz. The first one is, you got to do this little math problem, and next one is, what is the verb of the sentence whatever, but the instructions before the questions started said, "Read all the questions before you start."
[laughter]
Sal Daher: Right.
Gil Syswerda: Yes, I read through them like I'm a good little boy. You get to question seven, or something like that, and it says, "Write the answer to question four on the blackboard."
[laughter]
You got to walk to the front of the room and write down your answer. I think the second to the last question was, stand up and say your name out loudly. The last one is ignore all the questions above, just sit quietly at your seat until we get the quiz is done.
[laughter]
I read through the whole thing, I get to question 10, I go, "Huh." I just sat there, and I watch all the mayhem happen, it was a lot of fun. Now, back to Hewlett-Packard, right?
Sal Daher: Yes.
Gil Syswerda: I read question one and then question two again, and question three, and there's a logical contradiction there. You cannot do three, the way the problem is stated. I get the attention of this woman, I said, "Excuse me, I think you may have maybe a problem in the way this question is formulated." She stands up, walks over to me, pops the quiz out of my fingers, grabs the blank sheet of paper because I hadn't written a thing on it yet. She says, "I'll take you to your next interview."
[laughter]
The right answer was, don't write on a piece of paper.
Sal Daher: It doesn't work.
Gil Syswerda: I'm thinking, "Oh, thank you, Mrs. DeVries” This is otherwise, I would have been sketching out one and two before I even started on three.
Sal Daher: The version of that that I experienced, Gil, probably about the same time, was a sheet of paper with a bunch of questions on it. The first instruction was read everything. Before you do anything else, read to the end." The second instruction was, "Tear the sheet in half."
[laughter]
Sal Daher: Everybody in the class is tearing it in half. The third question was like, "Pat yourself on the stomach," and the other one was, "Rub your head." It kept going like that, until the last one, which was like, "Ignore all instructions except for the first one." Nobody in my class got it.
Gil Syswerda: Is that right?
Sal Daher: Nobody. Nobody, me included. Everybody, you're so impetuous, and you want to get to it, instead of going all the way to the end. That's a great one. Tell me the BBN fish story.
Growing Fish at BBN Labs
Gil Syswerda: All right. I was at BBN labs, this was at 10 Moulton Street, if you're familiar with that.
Sal Daher: BBN is the company that built the internet.
Gil Syswerda: That's right. It was a pretty big company, and they had a labs division, kind of like Xerox PARC used to be. It was a really fun place to work. Anyway, 10 Moulton Street is the ideal building for doing research in. If you have a job where you need to design some sort of complex where are lot of creative work gets done, get the plans to 10 Moulton, duplicate that building a bunch of times.
It was basically offices all on the outside of the building, a square building. Then a hallway, and on the inside are conference rooms and elevators and so on. A lot of window space. I went to BBN because of Dave Davis, because he was doing genetic algorithms. Dave was interested in hydroponics, so he had a window office, and he and I were talking about how we could possibly grow produce in our office.
Sal Daher: [laughs] I like where this is going.
Gil Syswerda: We tried a few plants, right, and nothing really wanted to grow because the windows had some sort of a reflective layer, and it was stripping out certain frequencies-- [crosstalk]
Sal Daher: UV light.
Gil Syswerda: Yes, UV light or whatever. I think I came up with the idea that maybe we should try growing fish. Do a fish farm. So, we researched it, and ordered a bunch of tilapia and a couple of 50-gallon tanks. One for his office, one for mine. We were on opposite sides of the building. The tilapia were these itty bitty little fish, like half-inch long. We had these aquariums with tilapia, and these tilapia started growing like crazy. Tilapia just grows. We had more and more fish mass in our aquariums. We started roping other people in to get aquariums and take some of these fish. Eventually, our aquariums had fish that were five, six inches long, fat. Each aquarium would have 10 pounds worth of fish moving around in it.
Sal Daher: Woah. Just a lot of fish detritus, I imagine, to pick up.
Gil Syswerda: At our peak, in that building, there were 15 people who had aquariums. We had this big distributed fish farm, and then it got too much. They kept growing, some of our fish started having babies. Next thing you know, we have little baby tilapia along with all those giant fish. The intent was always to eat the fish, eventually. One weekend, we went at the end of the parking lot behind Dave Davis' condo, and set up tables, and had a fish-fry. A bunch of people showed up, the fish farmers plus their significant others, and we ate a whole bunch of fish.
There were a whole bunch left, plus they were making more, so then time went on, and then we had another one. We had two big fish-fries out of it. We kind of got tired of the whole thing, and eventually, we just ended the whole thing.
Sal Daher: Threw them in the river or something.
Gil Syswerda: I think a lot of them ended up in Fresh Pond.
[laughter]
Gil Syswerda: Right across the-- [crosstalk]
Sal Daher: [unintelligible 00:43:56] crowded with the--
Gil Syswerda: No, they're tropical fish. They came out of Africa, they're an African pond fish. When winter came along, I'm pretty sure none of them made it through [unintelligible 00:44:07] [crosstalk]
Sal Daher: They're all dead.
Gil Syswerda: Yes, but it was fun.
The Interview at Lawrence Livermore Research Laboratory
Sal Daher: That is a great story. Tell me about the story of your campus interview with Lawrence Livermore Research Laboratory. Are they a weapons laboratory?
Gil Syswerda: Yes. Nuclear weapons. I've read some articles about them that did a lot of other stuff as well, so they're on the list. I filled out the application, right? There's a line on the form that says, "What is it that you want to do?" Because they were Lawrence Livermore Labs, I said, "I want to do socially relevant work in the fields of parallel computing, blah, blah, blah." I figured a socially relevant thing might just strip me out of it, but they called me in for an interview. I said, "Uh-huh."
Next thing you know I'm in this little room with a desk and the guy, he's a grey man in a grey suit. By the way, I now understand his pain because I've done campus interviews that are not anywhere near as fun as you think they'd be.
Sal Daher: Yes, I can imagine extremely repetitive.
Gil Syswerda: It's a grind, you're talking to students who don't really know how to interview well, as one after the other, plus, you're traveling from place to place. He probably flew in the night before, was tired. He clearly hadn't read anything. He has to turn to zero prep, I'm sitting across, he's reading the thing, and I'm watching. I'm going, "I'm just watching him to see when he hits that socially relevant line to answer." [chuckles]
Sure enough, I see him stop there and just suddenly wake up. He looks over at me like he's looking at me for the first time. Looks back down at the sheet of paper, and he reads the thing out, he says, "Socially irrelevant work in the field that"-- He reads the whole sentence. When he looks up, he looks at you right in the eye and he says, "You know, we build nuclear weapons, don't you?"
[laughter]
Sal Daher: Irrelevant.
Gil Syswerda: Anyway, I think the upshot of that is there was a mutual understanding that maybe we shouldn't move forward on this one.
Sal Daher: What do you think? One of Mother Teresa's charities? You build nuclear weapons there.
[laughter]
Gil Syswerda: I thought maybe you did other stuff, too.
Sal Daher: We're building nuclear weapons when we're not raising tilapia and having fish fries. We also do fish fries here, it's not just BBN.
Gil Syswerda: That would've been great if you popped something like that up.
Sal Daher: Oh, we have fish that glow in the dark.
[laughter]
Gil Syswerda: It was a fun encounter with somebody doing campus interviewing. Later, I got my comeuppance because I did the same thing. Grinding it out on-campus interviews.
Gödel, Escher, Bach & All That
Sal Daher: Well, how about the Douglas Hofstadter story?
Gil Syswerda: All right. Douglas Hofstadter, he's a genius that marches to the beat of his own drum. He's always been an audible box other than mainstream thinker. He wrote the book Gödel, Escher, Bach.
Sal Daher: Oh, yes.
Gil Syswerda: Won the Pulitzer Prize for it, he totally deserved. It was just something that hadn't existed on the planet before. It was my reading of that book that actually eventually prompted me to go back to school, that was a big part of it. It just happened to be whilst at University of Michigan that Douglas Hofstadter joined the staff. He only was there for a year or two, he didn't like it and went back to Indiana. I took a class with him. There's your class and some of the stuff, he eventually asked us to write a paper. I'm an engineering-type person, my writing skills back then weren't necessarily up to the snuff of the English department.
[laughter]
I get the paper back, and there is so much red ink on this paper. He's highlighting all the things that he didn't like about my style of writing, the grammar, punctuation, and [crosstalk]. It wasn't just me I've been looking and there's red ink on everybody's paper, for instance. He stands in front of the class, and he proceeds to tell us that in the way he views it, there isn't much difference between the idea in the paper and the way it's presented. There are two sides of the same coin. He says, "If you're a sloppy writer, even if your idea is good, I won't be interested.
Sal Daher: You're not going to get far with that idea. Even if you're not a very stylish writer, if you've really thought something through, you should be able to explain it.
Gil Syswerda: Well, it's clearly important. Having good writing skills is clearly...a great skill to have in life.
Sal Daher: It's tied in the thinking skills!
Gil Syswerda: Well, a little bit. You can think clearly about something and then express it poorly. He made a stronger statement in class. He said, "There's really no difference between an idea and how that idea is presented." He asked us, "Could you all please get the book, The Elements of Style by Strunk and White?
Sal Daher: [laughs] Classic.
Gil Syswerda: "Read it before the next paper." I bought that book and sure enough, it was eye-opening. There's a lot of stuff I didn't know about writing. The next paper, I paid a lot of attention to my writing style. I even had someone in the English Department review my paper to make sure I got it. In the bibliography, I listed all my sources, and since there's no difference between an idea and the expression of that idea, I listed Strunk and White in the bibliography because obviously-- I get the paper back and there's this big red circle and question mark around Strunk and White.
First of all, what professor even reads the bibliography? He did, so I went and talked to him about it afterwards. I said, "I explained my reasoning. You said there's no difference and Strunk and White was a big part of the formulation of this paper. I was using the property." It was a little bit of a joke, right? He did not see the humor in that at all. Not at all. It's just like, that joke went right over his head. He was disappointed actually.
Sal Daher: He's a genius. You expect him to have a sense of humor too?
[laughter]
That book is such a groundbreaking book. He doesn't have to be telling good jokes.
Gil Syswerda: That's a huge book too. A lot of people started that book and didn't finish it. I started doing this thing asking people, "How far did you get in the book?"
Sal Daher: [crosstalk] me too.
Gil Syswerda: Very few people actually finished that book.
Sal Daher: I didn't either. How about the story about snack and the enraged Soviet scientists? What is snack?
The AI Researcher and the Daughter of the Soviet Scientist
Gil Syswerda: Snack. This is another BBN story actually. When I was at BBN, a group of us formed the Seminar on Natural and Artificial Computation. It wasn't just BBN people, it was people thinking machines, people in the Rowland Institute some of the area universities like Harvard and MIT and so on. It was this informal group about 20 people, something like that. We met at the Rowland Institute every Friday. I don't know if you're familiar with the Rowland Institute, I can go into that a little bit. The Rowland Institute is right on the Charles River. It has the nicest conference room I've ever been in.
There is this beautiful oval table, which is four inches thick wall of windows overlooking Charles River in downtown Boston. To get to the conference room, you have to walk through a Japanese garden with the sand and everything else. It was just a complete high-end place where a bunch of scruffy people like us [crosstalk]. Here's the thing, it was the Seminar on Natural and Artificial Computation at the Rowland Institute. We made ourselves up to be pretty prestigious. Cambridge's who's who of people coming through giving lectures universities.
We would invite them to come and present to us. We had some really well-known people come in here and give us a talk, and that was a lot of fun. Somebody invited a visiting scientist from the Soviet Union. This was right about the time the Soviet Union was breaking up, but somebody from Moscow. I didn't know much about him but hey it's just like, we're doing this all for fun anyway. He's giving a talk and it's about measuring the auras of things. The theory was that everything has an aura, like an electrical aura. I know [unintelligible 00:52:49] this is junk science.
He had some impressive equipment that he showed pictures of and also his whole focus was to take pictures, the auras of people and tried to determine if they have diseases, that's the aura which highlight what was wrong with them. Here is one test subject, and it's a strapping young man, he had just a pair of briefs, and the shows us his aura, and so on, he says, "By the way, that's my son." The next test subject is pretty striking young woman also just wearing a pair of briefs. Stewart Wilson, I've mentioned him I think the last time but if I haven't, I need to talk about him a little bit and give him some thanks.
Anyway, Stewart Wilson, he says under his breath but loudly enough that we all heard, "So is that your daughter?" It was funny. I'm sure a couple of us snickered a little bit. The scientist went ballistic. He got so mad. I'm not even sure what it was. Somehow Stewart had just crossed some cultural barrier that should not have been crossed because just an innocent statement like that scientist got so angry, he could barely talk anymore. I thought he was going to storm out the place. That was our encounter with a Soviet scientist.
[laughter]
Sal Daher: You look at stuff they have a very- American society is not very hierarchical. Even when there's hierarchy, people joke across hierarchies, and they make fun of their boss. When people come from societies that are highly hierarchical, where people have these positions, and you don't joke about that. People take these things very seriously.
Gil Syswerda: I know. I agree with that. I don't understand what in that context made him so mad. I'm not even sure we found out whether it was daughter or not. Because the whole thing just kind of blew up at that point. [laughs]
Sal Daher: It must have been a little bit sensitive, this sort of funny type of science.
Gil Syswerda: Stewart is easygoing gentlemanly guy too. I'm sure he had absolutely no intent to piss that scientist off. Right?
Sal Daher: Yes. That's funny.
Gil Syswerda: Kind of just slipped out.
Simulating Professor Paul Scott
Sal Daher: I understand that you simulated a professor at the University of Michigan?
Gil Syswerda: Yes. This was Paul Scott. Paul Scott was actually instrumental in me getting into University of Michigan in a very short order. Wrote a letter of recommendation and spent a lot of time with him. University of Michigan was ahead of its time in computing. It had a system called Confer, which is kind of like Slack, but probably way more advanced than Slack. This is a system where you could have conversations, threaded conversations really easily. The whole AI group at University of Michigan was on this thing and talking to each other all the time. Paul Scott in particular was on it all the time. He just wrote these essay-length things. Talking about this and that everything else.
I got it in my head one day, there's a lot of content. Paul content on the system talking about AI. I had this idea about how to simulate a person based on the writings.
[laughter]
It's a little bit reminiscent of the language models we see today. Some of the big ones that came out in the last 18 months, like GPT-3 and BERT. The idea is the same. You analyze a body of text and then you build up a probability model for how one thing falls after another in whatever language you're analyzing. The systems today, they're hugely complicated. The probability distributions that they're creating are just really sophisticated. They have hundreds of billions of parameters, up to trillions now. Not my system.
My system had a few million parameters. The approach is completely different, but it was still built at a probability model. I could feed it a body of Paul Scott text, and then it would start regurgitating Paul's Scott texts right there. The thing is, it sounded just like Paul Scott. He got a particular style of writing. He kind of had a rambling tone to him anyway, which really helped out simulate him. I went and talked to him. I said, "I've built a simulation of you. Do you mind if I just pretend I'm you and just start feeding output of the system into our discussion forums?" He says, "Ok."
I started doing it. It's like you have this thread of conversation. I cherry-picked a little bit because these things are just spew out reams of Paul Scott. I picked something that sounded relevant to whatever discussion was going on, and I copied and pasted it in, pretending I was Paul Scott. Of course, people were trying to infer what it was that Paul was trying to say because even though the syntax and the sentence structure was pretty good, there was absolutely no semantic content there whatsoever. [laughs] It was just all statistical noise that sounded meaningful, but the meaning was all in the eye of the beholder.
Because guess what? There was no Paul Scott there. This thing went on for about three days. Then finally, somebody started challenging. Says, "I'm not actually sure if this is Paul Scott." [laughs].
Sal Daher: Took him three days, huh?
Gil Syswerda: Yes. Took him three days. I fessed up and told him what I'd done, anyway.
How to Have a Board without a Board
Sal Daher: That is so funny. Hey, Gil, let's do the following. Why don't you tell the Machine Insight a Marty G story? Then let's wrap up our interview. Just giving us an idea of where AI is headed and what it takes to start an AI startup today. Because we touched a little bit on that already. Let's talk about Machine Insight.
Gil Syswerda: All right. We started Machine Insight. This happened after Optimax. If you remember from last time we talked, we had some pretty comprehensive procedures in place for making decisions. When we came to loggerhead and so on. That turned out to be really valuable and [crosstalk].
Sal Daher: That's right. The CEO was very constrained in what decisions he could make. Also, the individual players had a lot of latitude in decision-making.
Gil Syswerda: Truth. 99% at the time. Jeff Herman was a CEO, just when about his business made decisions.
Sal Daher: Right. It's kind of like chairman of the board instead of CEO?
Gil Syswerda: No. Even different than that. The other two founders is true for all of us. Any one of the others could challenge a decision and say, "This one has to be unanimous." We would get together and hash out what's this decision should really be. If we couldn't come to a unanimous decision, we truly couldn't. This never actually happened, then we hand the decision off to somebody else, and at that time it was our board. Nobody wanted to go to our board and say, "We can't make a decision, you guys [crosstalk]. That was going to happen. Sometime later, Jeff and I start Machine Insight. That was designed as IP holding company. We're going to develop IP in there and then license it to other startups. It was set up as an LLC, and the two equal partners because we're both equal partners in crime at that point [crosstalk].
Sal Daher: Jeff is Jeff Herman?
Gil Syswerda: No, this was Jeff Palmucci.
Sal Daher: Jeff Palmucci, okay. The other Co-founder?
Gil Syswerda: The co-founder of Optimax. He's the technical co-founder. Jeff Herman focus on sales and marketing, and Jeff Herman was also really, really good at contract negotiation. He was great to have onboard. Optimax would not have been a success without all three of us doing our things. That worked out in terms of co-founders. Now we have two people, each with a 50% ownership of the company. We wanted to devise something in case we disagreed with one another. We're going to do the same thing.
We need to be unanimous, and we can't make a decision, you're always going to go to somebody else. It's an LLC, we don't have a board, so we needed a virtual board member. We went through Bentley, who we both know that we would trust in making the decision in case we come to loggerheads. We picked this guy, his name was Marty. Then we decided it's actually not really necessary for Marty to know that he has-
[laughter]
Sal Daher: He's a board member.
Gil Syswerda: -Not officially a board member, but acting in the context of making a board-level decision for us. We figured he doesn't really have to know until such time as when-- It was actually in our corporate docs that Marty was the decision-maker. We never bothered to tell him because then he's going to want to be involved with the company, yadi yadi yada. We never had to invoke him. That all went by, and Marty never knew that he was actually part of the Machine Insight.
Sal Daher: [laughter] The threat of bringing Marty in was enough get the two of you to decide.
Gil Syswerda: Yes, that's all you need, because then we'd have to go crawling to him to explain we included it in our corporate docs, could you please make the decision for us? That would've just been too embarrassing for us.
Sal Daher: That is so funny. Great. Gil, let's wrap up, let's get to, where do you think artificial intelligence is headed? Just like a very high level, something that's accessible to the general listener.
Gil Syswerda: The end game for artificial intelligence is artificial general intelligence, which is human-level intelligence and beyond. We talked about this a little bit the last time. Was that afterwards? I can't remember.
Sal Daher: Yes, I think we talked some of it online and some of it offline.
Gil Syswerda’s Perception of Where AI Is Today
Gil Syswerda: There's a feeling amongst people that have a lot of money that the first organization that creates artificial general intelligence is a winner-take-all. That's the end game. There's a lot of money at play here trying to be the first. Deep Mind is a Google subsidiary is doing it, open AI is doing it.
Sal Daher: Yes, we went through the list of the likely candidates. Who would be the likely candidates to get to AGI first.
Gil Syswerda: These organizations are spending a lot of money and they have a lot of researchers, and they're generating tons of AI results by-the-by. Much of what they're doing, they publish and make available to open-source, really powerful models. Having to do with language, computer vision, it's now moving into planning and reasoning and so on and so on. To build these models requires huge teams and literally tens of millions of dollars just to build one of these models, but you can use them.
That's kind of where AI is at today. Deep learning, which came on the scene at the end of 2012 and just exploded onto the scene has pretty much wiped out all other AI. There's very little use of other AI techniques anymore. There's some standard off-the-shelf stuff like random forest. They're in libraries and you can apply them, but all those techniques are being subsumed now into the field of statistics. It's statistics and statistical models and then deep learning, which makes its own models. As a startup, there's actually very little room to do R&D and be innovative in the space anymore. We did it with FeatureX and Machine Alpha, we had some unique techniques that worked on financial data. Financial data, because of the structure that data there's still room to innovate. Things like machine learning language understanding, and some of the big bullet points of an AI, you're much better off pulling stuff off the shelf, and using these big frameworks and precomputing models, and so on.
As I pointed out earlier, that can be a good thing to do. If you can figure out a market need and solve it with some off-the-shelf technology, and it's not necessarily off-the-shelf. You have to do some modification to it, and so on; configure it, and maybe do some training on some specialized data, and so on. If you can get that to work, you can do that with relatively small dollars, and use compute resources on Amazon, AWS, and maybe make a success.
Your barriers to entry, technically, are really quite low, because anybody can download these things and start using them. The techniques are also really quite powerful so you can do things now for businesses that you weren't able to do before. Either, because the expertise required was just too large, or the amount of money you'd have to spend was too large. There are opportunities being opened up. You have to be careful, though, because if you don't fully understand the market needs, the point I made earlier, and these techniques that are out there that you're using, don't actually fulfill that need, come close, but not close enough.
You're probably not going to be able to extend whatever it is that you're using, into a usable state. Things get really hard and really expensive and you need really talented people to start modifying these models and so on. I feel this in a broad nutshell where we are today, and it changes every day. The rate of progress on AI is something the general public doesn't really understand. It is just exponential at this point. A little scary and it's a little hard to see where it's going to go in the near term but there are startup opportunities there.
Sal Daher: Well, I think that is a very useful thing for people who are thinking of starting AI startups. It's also a little bit of a wake-up call to the general audience to say, "Wake-up AI is much further along than you suspect."
Gil Syswerda: Deep Mind has come out with a position paper just in the last few weeks that says, "We think that we now have all the pieces in place to achieve AGI."
Sal Daher: Artificial general intelligence.
Gil Syswerda: Artificial general intelligence, get to the first version. Once you have the first version, you use that to make the second one so and on, and then there's a runaway train. That's a strong statement to make, by a big tech company like Google. You just can't discount it, because the progress they have made in building AI models to do really sophisticated things, is real. If they’re putting their reputations on the line in such a direct way and say, "We think we have the pieces. All we have to do is execute now."
Sal Daher: Excellent.
Gil Syswerda Highlights People Who Were Influential in His Success
Gil Syswerda: Sal, I just want to highlight a couple of people who I've already mentioned quite a few people in talking, and hopefully I did them justice. In terms of John Holland, I just want to say he was a true genius. He invented genetic algorithms, and he won the MacArthur Award for that. Even though I didn't agree with him, in the end on some of his things, that doesn't take away from his genius, and I didn't mean to imply that if I did.
Also, the people who have worked on classifier systems. The reason Jeff [Palmucci] and I succeeded in getting that to work is because we framed the problem differently. Holland initially was really interested in adaptive systems and applying classified systems to adaptation of living organisms. We reframed the problem to just understanding current situations. That's a different kind of a problem. The way for that was shown by Stewart Wilson, a research scientist at the time at the Rowland Institute.
He had produced a system called XCS published a paper on that, and that was a motivation for me to then start looking at classifier systems eventually resulted in the ECS. I’ve got to thank Stewart Wilson for that tag because he has some real insights along the way. I built a computer vision system based on some insights that he and Edwin Land had in terms of how the human vision system work, including the mapping from the retina to the visual cortex. Or Retino-Cortical Mapping, so that ended up being a pretty interesting computer vision system, which I had to give him credit for as well.
I worked closely with a Tudor. I already talked about how great it was to work with that team and Paul Tudor Jones. One person I do want to mention out of that team is Steve Evans. Steve and I worked very closely with one another. I don't think I've ever met a mind brighter than his. He has a mind like a steel trap, nothing gets by him, and he's got a memory to match. He never forgets anything. Steve, it's pretty good working with you, if you happen to hear this.
With Jeff Palmucci, he's my business partner for a long time. I used to think I was kind of a hotshot programmer, and then I met Jeff Palmucci. He's easily two times better than me, just completely put me into my place in terms of developing code. I did eventually figure out why he was so good. He has interesting insight, but it's technical as well. If we have time, we can just dive into that a little bit. A quick little story about Jeff, too, we are raising money. This is, I think, in Percipio. We were a little worried about putting Jeff in front of VCs because Jeff is just-- Now, Percipio was the hedge fund that we did.
Sal Daher: Percipio was your first start. Oh, I'm sorry, hedge fund, okay.
Gil Syswerda: Jeff is just a complete straight shooter. He's just like you, but more brighter.
Sal Daher: It's too dangerous.
Gil Syswerda: We're trying to give him some tips about how to talk to VCs because it feels like-- We would imagine Jeff is going to say, "Well, we're working on this really cool technology, and we're going to become famous for it, or trying to make the world a better place or whatever." All of that's fine, but VCs would actually like to make money. Anyway, then we're talking to the VC, sure enough, we got a roomful of them. Somebody asked a question, like, "Jeff, why are you doing this business? What really interests you?" Jeff says with a complete straight face, right back up at the guy, he says, "I want to become so rich and famous for it."
[laughter]
The whole room burst into laughter. He totally got away with saying that. We've already talked about Douglas Hofstadter along the way, so that's it.
Sal Daher: Well, on that note, I think we can sign off. Well, Gil, it's been a great adventure. You've been a great sport to make time to let us pick your brain on all these topics. A rank amateur such as I am, speaking to somebody who's been doing AI his whole life. I tremendously value all the time that you spent with us.
Gil Syswerda: Well, Sal, it's just really fun talking to you. I hope you have some long-suffering listeners because we've been talking about a lot of things for a long time. Good luck with the editing.
Sal Daher: The reality is that some of my longest podcasts have been some of the most successful. My interview with Ed Roberts, it's hugely popular, and it's 89 minutes long.
Gil Syswerda: I haven't listened to that one yet, so I'm going to--
Sal Daher: You will laugh. Ed is a hilarious guy. He's very generous, a great-souled individual. He's just a great guy. Anyway, Gil, I'm really grateful to you. Thanks for making the time. This has been a very instructive interview for me. I've learned a lot and had a lot of chuckles. Those stories are awesome. I loved the stories.
Gil Syswerda: Thanks.
Sal Daher: Thanks a lot.
Gil Syswerda: I have plenty more where those came from.
[laughter]
Sal Daher: Some of them were actually fish stories.
Gil Syswerda: Yes. Don't get me started on racecar driving stories. Oh my God.
Sal Daher: Oh, yes, we talked about racecar driving last time. Anyway, once again, Gil Syswerda, artificial intelligence genius, and founder. Thanks a lot for being on the podcast.
Gil Syswerda: My pleasure.
Sal Daher: This is Angel Invest, Boston. I'm Sal Daher.
I'm glad you were able to join us. Our engineer is Raul Rosa, our theme was composed by John McKusick, our graphic design is by Katharine Woodman Maynard. Our host is coached by Grace Daher.