Targeting big impact deep tech
Podcast transcription - 23rd January 2019
Alan Cowley: Welcome back to part two of William Tunstall-Pedoe's Invested Investor podcast.
Alan Cowley: In the last episode, we heard how William pivoted multiple times with his company, eventually creating Evi Technologies, which he sold to Amazon and successfully created Alexa.
Alan Cowley: In part two, we hear how William has transitioned into angel investing, as well as some valuable tips for entrepreneurs and angels.
Peter Cowley: Let's move on to angel investing. We met, the first time whilst you were doing that earn out, in this room, a quirky building, which is your home and which I really enjoy being in, in Central Cambridge. You decided to start angel investing. You had already done a little bit before that, but you've obviously done a tremendous amount in the last three or four years. Why did you do that? You'd obviously seen some angels, but why do that?
William Tunstall-Pedoe: Octopus had a group of what they call venture partners, which are high net worth, they invested alongside the venture capital trust funds, partly for historical reasons. They started with a catalyst, as an entrepreneur, they had backed, I was automatically made a member of their venture capital group. I would see the deals as they came through. Obviously, prior to exit, I didn't have very much money to invest, I think I did a couple of tiny investments as an entrepreneur, 5K or less. After the acquisition, I had some real capital, I also had a very substantial amount of income. There was a stock package that was negotiated during the acquisition, Amazon's stock price had tripled during the period of my employment there, to the extent that I had a ridiculous income.
Peter Cowley: Using stock?
William Tunstall-Pedoe: Partly from stock, yes.
Peter Cowley: Okay, yeah.
William Tunstall-Pedoe: It counts as income. The stocks are taxed as income. Also, I had a very large amount of income tax, EIS is a very generous scheme when it comes to income tax.
Peter Cowley: Yes.
William Tunstall-Pedoe: By investing in EIS, I could get involved in start-ups, which obviously I knew quite a bit about, having been a start-up entrepreneur for a very long period, of time. It was very tax efficient, and it was interesting. When I left Amazon, I was given an offer to help with an incubator in Canada, called the Creative Destruction Lab. It seems a bit bizarre that I'm working for an incubator in Toronto when I'm based in the UK, but the simple answer to that question is they asked me first. So, every two months I fly out to Toronto, and I work with 150 or more machine learning start-ups who are going through this incubator.
Peter Cowley: That's huge.
William Tunstall-Pedoe: Yes. It's become very, very successful over the two years I've been there, as well. It was just in Toronto, now they have branches all over the place. So, I have huge amounts of exposure to early stage AI machine learning start-ups there. I invest in those, as well.
Peter Cowley: Right.
William Tunstall-Pedoe: I'm a member of Cambridge Angels, that's another source of deal flow for me, so, between Octopus, Cambridge Angels, and the Creative Destruction Lab, I see a lot of companies, and, as you say, I'm very active.
Peter Cowley: In fact, it's nearly 50 investments you've done over about four years.
William Tunstall-Pedoe: Yes.
Peter Cowley: So, this is beyond the call of duty for most novice angels.
William Tunstall-Pedoe: Yup.
Peter Cowley: Explain that process? Do you fall in love with things too easily, perhaps?
William Tunstall-Pedoe: Maybe, I had a big early success. One of my early investments was in Magic Pony, which was acquired by Twitter within 18 months. There was one follow on round within 18 months, which is exceptional. I think a lot of the more prolific angels have had an early success which encourages them and may be statistically unfortunate, because they may not have that success again for some years.
Peter Cowley: So, you go through this trough of disillusionment, potentially? Which you haven't quite got into yet.
William Tunstall-Pedoe: I haven't gotten into because I haven't had enough failures for that to happen. I have a spreadsheet with a big portfolio. I've had one spectacular exit, I've had one failure, I've got a lot that have been around. I've got them in my spreadsheet at what I invested, they're doing okay, or they're struggling a bit. As you say, it's too early to know.
Peter Cowley: How much interaction do you have with them? Obviously the ones in Canada, you probably see every second month.
William Tunstall-Pedoe: Yes, so interestingly, I spend a lot of time advising start-ups. A lot of the start-ups, I advise, I'm not an investor in, they're going through the Creative Destruction Lab. Part of my job as a fellow there is to help those businesses. I did something like 500 start-up meetings in 2017. A lot of them were from the Creative Destruction Lab, a lot of them were small meetings to help those businesses. I'm less involved than a lot of angels are with the companies I invest in, although I am formally signed up as an advisor for two businesses that I've invested in. One of them is called Fluent AI, which is a Montreal based speech technology business, which I met through the CDL. One of them is called ADA Health, which is a Berlin based AI physician, or symptom checker. You've got absolutely enormous potential, essentially automating a big part of what doctor's do, pointing people at what their next steps are when they have symptoms or a health condition.
Peter Cowley: Does it use a voice interface also?
William Tunstall-Pedoe: It doesn't, but it could do. That's a detail. It's an interactive, conversational interaction with the app. You can download the app. It's out, it has got some very, very deep tech behind it, very exciting. They just raised a huge Series A round, 40 million Euros, I think.
Peter Cowley: From American money?
William Tunstall-Pedoe: No, it's European, they took me on as an advisor as part of that round. I mean, it's more an investment, but advising as well. I'm excited about it because it's got enormous potential, the world has a huge shortage of doctors, particularly in the developing world. Automating a chunk of that with artificial intelligence is very exciting.
Peter Cowley: Are you a board member of any?
William Tunstall-Pedoe: I'm, not, no.
Peter Cowley: Was that on purpose or lack of opportunity?
William Tunstall-Pedoe: Partly on purpose. There are businesses that have asked me, but my time is incredibly short and incredibly valuable, I'm trying to claw it back because I want to be an entrepreneur again.
Peter Cowley: You've been doing it about three of four years. What have you learned during that time?
William Tunstall-Pedoe: I've learned a huge amount from just talking to lots of start-ups. Obviously, I had a very deep, 10 year plus experience as an entrepreneur myself, covering every problem that you've ever had. I've tried to take that experience, to literally, hundreds of other start-ups and help them. In the process, of course, I've learned huge amounts about their problems and what they're facing in a much, more shallow way. In a way that doesn't have the same ownership, I'm always surprising myself at how much I've learned, in the last few years, from the exposure to other businesses. Part of the reason I got into it, was because I thought that I would meet a start-up that I might want to join. That hasn't happened. I also thought I may learn a lot from interacting with start-ups, that has happened. I've learned a huge amount. Because the Creative Destruction Lab has a specific machine learning track, I've been exposed, particularly, to machine learning start-ups. I automatically put them in six or seven buckets. I can see patterns. There is an existing, businesses applying machine learning, there is a diagnosis using the advances in machine vision to automate from scans start-up. There's, basically, all sorts of buckets, all sorts of patterns.
Peter Cowley: Can I ask you which the most exciting bucket is?
William Tunstall-Pedoe: I don't think there's one bucket that's exciting. Obviously, taking advances in AI and applying them is, obviously, a big source of value. There are hundreds of AI businesses that are, essentially, old businesses, applying AI, in order to get a business with a competitive advantage. I'm also very excited by general AI, the more horizontal application, the core tech.
Peter Cowley: Deep minds?
William Tunstall-Pedoe: Yes, deep minds. Exactly. There's not so many of those and they're tougher businesses because they're competing directly against the big tech companies, who are trying very hard.
Peter Cowley: Yes.
William Tunstall-Pedoe: They're also incredibly exciting, I'm very motivated by businesses that, potentially, have a very big impact.
Peter Cowley: Philosophical question, is there a danger in AI? Even the tech giants are getting a little bit concerned about how society's going to cope.
William Tunstall-Pedoe: That's a very big topic of conversation. There's a number, of things people talk about when they talk about dangers of AI. One of which, is mass unemployment, from AI taking jobs, which has some truth to it, but it's exaggerated, I think. I think, a lot of AI is augmentative, rather than completely replacing humans. There are, obviously, clear examples of where that isn't true, like drivers. For example, driverless cars, if they deliver as expected, will make many tens of millions of people unemployed. There's plenty of other examples where they're just helping people, they will change people's jobs, make them more efficient, but they won't replace people's jobs. There's another source of risk that's talked about, which is bias, the black box that declines somebody for a loan, or decides that somebody's going to stay in jail without it being explained. Where there doesn't seem to be a bias, but the data is biased in various ways. That's often cited as a concern. The other one is general AI destroying humanity, taking over the world, “Terminator” type scenario, which is science fiction. It's theoretically possible, but there's a number of big innovations that have to happen between now and then for that to be a source of concern. Obviously, because the consequences are so enormous, people are concerned about it now. It's not something that's going to happen very quickly.
Peter Cowley: Do you mind that people use the term AI rather more often than they should?
William Tunstall-Pedoe: Well, I guess, I don't because I don't have a better term. The cynical interpretation of AI is the things that computers find difficult. Generally, when computers start doing them really well, the AI term disappears. It's no longer used. Historically, it's always been the things that were very difficult to programme computers to do. AI's exciting, now, because computers have made some headway with some things that they had not been able to do for the last few decades, computer vision, is a good example. There are things that you can sit down and programme, and there are things that nobody can programme. Machine learning is starting to solve some of those. I can go to my Google Photos app, type in camel, and get all the photos I've ever taken over the last few years that have a camel in them. There's no way that a software engineer could sit down and write a programme that will identify camels in images. That's just totally impossible. But, with modern machine learning techniques, it's now possible.
Peter Cowley: We're both in the investment called Dogtooth Technology, which, of course, is recognising a strawberry, a ripe strawberry, then working out how to pick it.
William Tunstall-Pedoe: Yes, exactly. The advantages and creative vision have had enabled things to happen that were not possible. Driverless cars is another great example, Dogtooth is a great example, as well. Speech, as well, of course. Speech is completely enabled by more recent learning, but deep learning has accelerated it forward a number, of years from where would have been. It's very exciting, but it's unclear how much further we can go with the same techniques.
Peter Cowley: So, William, a couple of tips for entrepreneurs?
William Tunstall-Pedoe: One of the things that I didn't learn for many years, which I wish I'd learned very early, was the importance of the user experience in product design. Getting a designer involved in your product. The first versions of True Knowledge were really, really ugly to look it. They were very, very good from a technical point of view, but they looked appalling. I didn't regard it as particularly important in those times. I think, one of the most compelling things you can have in a start-up is an amazing demo. One of the most important things for a product, that's facing people, is for it to look good and work well, and be slick. It's not very expensive to get a designer involved, or get a UX person involved, in the early stages of your products, should you need to. I now understand this better than I did 10 years ago. I worked with UX people closely at Amazon, I've seen the importance of this. I think that's what I would've most liked to have known many years ago, as an entrepreneur, is how important that was, and how important a compelling demo is, in persuading people about your business.
Peter Cowley: Thank you. And for angels?
William Tunstall-Pedoe: I think there's two things that I've learned as an angel. One, is that there's many ways that a business can succeed. You can succeed, obviously, just by being successful financially. If you're generating huge amounts of cash, then the business is unambiguously valuable, due to cash flow. You can succeed from acquire, you can be acquired for many different reasons. You can be acquired because you've got an amazing team, you can be acquired because you've got amazing numbers of users, you can be acquired because you've got amazing tech that, in combination with the acquiring business, results in something enormous. That's, essentially, the reason we were acquired. We were never profitable, we had a good team and we had good tech, but it was a combination of Amazon's reach and punch with our tech that made us fantastically valuable to Amazon. I think, the fact that you can win multiple ways is something that is good to know. Another thing that I think about as an investor, is what the upside is. I think a lot about angel investing is the fact that it's asymmetric. The fact that the worst, case scenario is you lose 100% of your money, but, the best, case scenario is you get 20, 30, 40, 50 times your money back. That is what makes angel investing viable.
Peter Cowley: And risky.
William Tunstall-Pedoe: That's right. The point is, it is risky, it's very high risk. It's very inefficient, in all sorts of ways. If you didn't have that asymmetric with those possibilities of big returns, nobody would do it. I think the size of that upside return, I think, it is potentially, underrated by a lot of angels in the criteria. There's a very big difference between an investment that plausibly has a path to a ten, times return and an investment that plausibly has a path to a fifty, times return, in my head. I think you don't need many fifty, times returns in your portfolio to make the whole portfolio successful.
Peter Cowley: Probably one.
William Tunstall-Pedoe: One, exactly.
Peter Cowley: That's the home run, times three.
William Tunstall-Pedoe: That's right. That's my portfolio doing extremely well. The cost of that business may be higher, the price point may be higher than normal, but you need to weigh that against the possibility that it could be a very large return. For example, I have a fast battery charging business in my portfolio that was super expensive.
Peter Cowley: You told me about that. 10 million, or something.
William Tunstall-Pedoe: It was, I'm not sure if the terms are public, but it was an unpriced. It was a convertible, but the cap was very high, it had the potential to transform the electric car industry. It had the potential to transform the mobile phone industry. Obviously, there are competitors, there are problems. It's very, very high risk. There's a very good chance it won't get there. But, if it does, it's potentially worth hundreds of millions, or even 10 figures.
Peter Cowley: Yes.
William Tunstall-Pedoe: I regard that as a very good investment. I think it's very likely that it'll be a zero.
Peter Cowley: Yes.
William Tunstall-Pedoe: I regard that as a very good investment. I wish I had another 10 or 20 of them. Even though it was incredibly expensive, I still think, if you put it through the lens of what the plausible upside return is, you look at these things differently. Not all angel investors think the same way. I have also invested in businesses where, I think they are very unlikely to be worth more than 20, million, but have got a much higher probability of getting there.
Peter Cowley: Yes.
William Tunstall-Pedoe: But they look cheap and you can also do the opposite. You can look at the price. The price is very, very reasonable, they have a lower risk path to a 10, times and that's also a good investment.
Peter Cowley: But if you look at your full wealth asset allocation, probably you've got some cash, you've got some property, maybe. Et cetera.
William Tunstall-Pedoe: Yes, yes. I have a lot of public equities, as well.
Peter Cowley: Yeah, yeah.
William Tunstall-Pedoe: That's right. That's the other thing I like about angel investing, is the diversification. It doesn't necessarily correlate that closely with public equities. That's also part of the reason I do it, as well. I've got a chunk in assets that are very different from the global stock market.
Peter Cowley: And somewhat more boring.
William Tunstall-Pedoe: Or not.
Peter Cowley: Or not.
William Tunstall-Pedoe: Well, it depends on what you mean by boring.
Peter Cowley: Cash is pretty, boring.
William Tunstall-Pedoe: Cash is boring, yes. Yes, indeed.
Peter Cowley: Final question, which I ask everybody. I think you're probably about 15 years younger than me. You've already said you want to be an entrepreneur. What do you think you'll be doing when you get to my age, which is 62?
William Tunstall-Pedoe: Novelists sometimes talk about how they've got two more big novels in them. I think when I left Amazon, I felt that I had two more big businesses in me. So, I'm not sure I still feel like that, but I don't feel like I'm even close to pushing retirement.
William Tunstall-Pedoe: I feel like I've got one or two big further businesses in me.
Peter Cowley: Each one is 10 years.
William Tunstall-Pedoe: If each one is 10 years, then the next one-
Peter Cowley: You're going to be a lot older than me.
William Tunstall-Pedoe: Yes, maybe.
Peter Cowley: So, 75 years old, you're still doing 60 hours a week?
William Tunstall-Pedoe: Well, I'm in my late 40s now. So, two lots of ten years. It's getting old, yes.
Peter Cowley: Yes.
William Tunstall-Pedoe: But, in 20 years, that won't be old.
Peter Cowley: Yes. Provided your health remains.
William Tunstall-Pedoe: Yes.
Peter Cowley: Excellent. William, I really enjoyed that. Thank you very much.
William Tunstall-Pedoe: Thank you.
Peter Cowley: So, I'm with William Tunstall-Pedoe, effectively the inventor of the speech system for the Amazon Alexa, and he's going to give us a demonstration.
William Tunstall-Pedoe: A lot of people are familiar with this now. My experience is about 50% of the people I meet have got devices in their home. I'm happy to give a demo, anyway.
Alexa, what are you?
Alexa: I'm Alexa and I'm designed around your voice. I can provide information, music, news, weather, and more.
William Tunstall-Pedoe: Alexa, will it rain tomorrow?
Alexa: It might rain in Cambridge tomorrow. There's a 59% chance. You can expect about 2 millimetres.
William Tunstall-Pedoe: Alexa, tell me a joke.
Alexa: Why was the cat so good at chess?
Alexa: He always played a purrfect game.
William Tunstall-Pedoe: Yes. So, the jokes aren't necessarily any good. There was a period when one of my many responsibilities for the product included all the jokes.
Peter Cowley: You wrote them yourself, like cracker jokes?
William Tunstall-Pedoe: I didn't necessarily write them myself, but I was responsible for them all. Obviously, the jokes have to be suitable for all audiences and all members of the family.
Peter Cowley: That was supposed to be purrfect, wasn't it? But it didn't say purrfect.
William Tunstall-Pedoe: Yes, I'm a bit disappointed by that joke, actually. The team that followed me have not necessarily maintained standards.
Alexa, what is the time different between Cambridge and Seattle?
Peter Cowley: Cambridge is eight hours ahead of Seattle.
William Tunstall-Pedoe: So, that's just showing that this isn't all canned. Obviously, there's some calculation involved there. It could have been any two places.
Peter Cowley: But it could have been Cambridge, Massachusetts?
William Tunstall-Pedoe: Well, she knows I'm talking about Cambridge, England. There's some ambiguity there, as well.
Alexa, what is the factorial of 20?
Alexa: 20's factorial is two quintillion, 432 quadrillion, 902 trillion, eight billion, 176 million, 640 thousand.
William Tunstall-Pedoe: So, that shows two things. Firstly, she's extremely good at mental arithmetic. Secondly, she knows the names of some very, very big numbers.
Peter Cowley: I've tried that with a hundred factorial once.
William Tunstall-Pedoe: Did you?
Peter Cowley: Yes. It was a rather long time.
William Tunstall-Pedoe: Yes. I've learned 20's fairly good for demo, but you can do it with bigger numbers.
Alexa, set a time for five seconds.
Alexa: Five seconds, starting now.
William Tunstall-Pedoe: Obviously, that could be longer. But, for demo purposes, waiting five minutes for the timer to go off is not going to do it.
William Tunstall-Pedoe: Alexa, thank you.
Alexa: That's what I'm here for.
William Tunstall-Pedoe: Alexa, stop timer.
Alexa, play BBC Radio Four.
Alexa: Here's Radio Four FM.
William Tunstall-Pedoe: So, summoning any radio station instantly is hugely useful. No dials to twiddle.
Alexa, stop.
William Tunstall-Pedoe: There's hundreds of radio stations available through third parties and directly through the BBC, as was the case there.
Peter Cowley: That's built in as standard?
William Tunstall-Pedoe: Absolutely. Instant access to media is super important from an UX point of view. Similarly, music. Millions of previous tracks.
Alexa, play Call Me Maybe.
Alexa: Call Me Maybe by Carly Rae Jepsen.
William Tunstall-Pedoe: That could have been any of a million or more. There's various services.
Alexa, stop.
So, there's various services you can buy. Prime comes with music. There's also a particular music package you can buy from Amazon, that has even more.
Peter Cowley: Do you have to pay for that?
William Tunstall-Pedoe: Yes. Well, there's some music that comes, along with the music in your music library, that you can access. If you're a Prime member, which most people are, there's a big library. There's also Amazon Music, with an even bigger library. There's also third, party services. If you're a member of Spotify.
Alexa, play Yellow Submarine on Spotify.
Alexa: Yellow Submarine remastered by The Beatles from Spotify.
William Tunstall-Pedoe: This is accessing a third party.
Alexa, stop.
William Tunstall-Pedoe: I'm sure I could have got that from Amazon Music, as well, but it's any third, party music providers, as well.
Then there's routine stuff. Alexa, what is 83 kilogrammes in stone and pounds?
Alexa: 83 kilogrammes is 13 stone and zero point nine eight pounds.
William Tunstall-Pedoe: Alexa, what was the score in the Manchester United game?
Alexa: On Monday, Manchester United beat Crystal Palace, three, two, in the Premiere League match. They're playing in the Premiere League tomorrow at 12:30PM, at home, against Liverpool.
William Tunstall-Pedoe: Obviously, access to that live event type stuff. The other thing is there's many thousands of what are called skills, which are, essentially, third party capabilities that people have added. There's thousands of them.
Uber, Deliveroo, the third, party apps that we use a lot are actually linked in. Train Journos. My morning workout is mediated by Alex through another third party skill. There's many, many thousands. One thing that Amazon's way ahead is actually getting that out early and getting huge numbers of developers involved.
Peter Cowley: The platform effect.
William Tunstall-Pedoe: Yes. Alexa, who was President of the US when Donald Trump was a teenager?
Alexa: John F. Kennedy, Dwight D. Eisenhower, and Linden B. Johnson were the US Presidents when Donald Trump was a teenager.
William Tunstall-Pedoe: This is my favourite, show off, question. Obviously, huge amounts of reasoning that goes in there.
Peter Cowley: Yes. Teenager is 13 to 19.
William Tunstall-Pedoe: It's a period, of time that needs to be reasoned, from the date of birth. There's huge amounts of information about which president, which periods of time. Which of those overlap, with this period of time from our calculations from the date of birth. All done-
Peter Cowley: That wasn't hard coded?
William Tunstall-Pedoe: Absolutely, not hard coded. I like demoing that. It shows the reasoning and well of knowledge calculations of the system.
Peter Cowley: Good. Excellent. Thank you very much, William. Alexa, thank you.
Alexa: You bet.
Peter Cowley: Excellent, thank you very much.
Peter Cowley: Thanks for listening to another Invested Investor podcast. You can subscribe to all future podcasts via our website, Invested Investor.com, or via a number of podcast platforms online.
Peter Cowley: Remember, you can order our book online. Be sure to follow us on Twitter, LinkedIn, and Facebook to get most up to date, interesting, and insightful content from the Invested Investor.
