The transcript from this week’s MIB: Benedict Evans of Andreessen Horowitz is below.
You can stream/download the full conversation, including the podcast extras on iTunes, Bloomberg, Overcast, and Soundcloud. Our earlier podcasts can all be found on iTunes, Soundcloud, Overcast and Bloomberg.
ANNOUNCER: This is Masters in Business with Barry Ritholtz on Bloomberg Radio.
BARRY RITHOLTZ, HOST: This week on the podcast. I have an extra special guest. His name is Benedict Evans and he is a partner add the famed technology venture capital firm in Silicon Valley, Andreesen Horowitz, perhaps known to the aficionado as A16Z.
I have actually some swag from a recent — I have a blue A16Z had which matches my car and I just leave that in the car when I’m driving around trying to look cool to a very select group of tech geeks. If you are at all interested in the development of technology of ecosystems of autonomous everything, of smart this, of are really where the puck is going to be, the future of technology, you will find this to be a fascinating conversation, he absolutely has an encyclopedic knowledge of what’s taken place and why and has tremendous insight into the likely direction that the space is going.
So with no further ado, my conversation with Andreesen Horowitz’s Benedict Evans.
I’m Barry Ritholtz, you’re listening to Masters in Business on Bloomberg Radio, my special guest today is Benedict Evans, he is a partner at the legendary venture capital firm Andreessen Horowitz, where he works as an analyst and does a weekly consumer newsletter blog, what have you, covering everything from mobile platforms to AI to autonomous everything. He is also a recommended follow, he began his career as an equity analyst in investment banking moving on to strategy and business development at such August firms as Orange and NBC Universal. Benedict Evans, welcome to Bloomberg.
BENEDICT EVANS, PARTNER, ANDREESEN HOROWITZ : Thank you for having me.
RITHOLTZ: Do you prefer Benedict or Ben?
EVANS: Benedict is better for SO.
RITHOLTZ: Sure. There is an advantage to having an unusual first or last name for SEO purposes, I’m convinced that if my name is Barry Smith, no one would know who I am. But we can talk about SEO a little later. So you graduated Cambridge in 1998 right in the midst of the dot com and technology boom, not too long before the bust. What was it like coming out into the banking world in the midst of that era?
EVANS: I actually joined an equity capital markets team, so we were doing European tech IPOs in 99, and say for about nine months, I thought well, this is going to be a fun career.
RITHOLTZ: Right, this is easy, you just throw it out there and everybody buys it —
EVANS: You achieve some random thing in a sovereign wealth fund, you put in an order for three at a time (ph), and then in an about the early spring of 2000 everything started coming down and we sort of forget now that we talk about dot com bubble but you know, particularly in Europe, there was also a mobile bubble and there was also the fixed line broadband fiber bubble so you kind of have had three bubbles that kind of had happened more or less independently for different reasons and kind of merged and combined into one enormous thing.
RITHOLTZ: I would —
EVANS: And when the market went down for three years.
RITHOLTZ: I would argue you had a semiconductor bubble, you had a software bubble, you got a hardware bubble, you had a storage bubble, it all started a bandwidth bubble, in the in the US, we had fiber-optic and telecom bubble.
EVANS: Well, you had a tech bubble, then you also had a mobile bubble which only for very distant reasons, was happening at the same time but was happening at the same time and then there was a fiber bubble and yes, so they all combined and we got this enormous kind of explosion and an enormous pop and the market went down without stopping more or less for three years.
RITHOLTZ: So were you always the tech guy or was your background more finance or something or none of the above?
EVANS: So I did my degree in history which is really an analytic subject so here is a bunch of information, try and work out something interesting or meaningful or insightful to say about it or something original and interesting to say about it. And so I went from that into sort of fairly conventional vision into investment banking, and sort of quickly decided that the (INAUDIBLE) was kind of project management and whereas research was probably something I was likely to be better at.
So I went into equity research and I did — so I went and did European mobile stocks for a couple of years.
RITHOLTZ: And that was in London or in New York?
EVANS: New York and London.
RITHOLTZ: And did you spend any time in New York or did you go straight London to Silicon Valley?
EVANS: No, so I mean, I came to New York for work, because I worked for NBC Universal, I used to come over here. But no, the first time I actually worked properly in the States was to go to work for A16Z.
RITHOLTZ: Also London with really just a brief sojourn in New York City.
EVANS: Yes, I mean I worked for Orange so I went to Paris and I worked for NBC Universal when I came to New York but my job was — the job was always in London.
RITHOLTZ: So you are covering telecom stocks, you are covering mobile, you’re covering a lot of the technology-related space as an analyst, what was that like in the midst of wild overvaluation and collapsing share price?
EVANS: So it is kind of interesting, I mean there are a couple of different things there. One of them is it interesting in hindsight to look at them in hindsight in that they were the go go (ph) grow exciting dynamic safety (ph) disruptive companies, you know, mobile had gone from nothing in 1990 to oh my god, everyone on Earth is going to have one of these things and they were really kind — they thought of themselves as products and they connected everybody and that was it.
And you know, there is obviously, there is still a lot of stuff going on in emerging markets but like, everyone in Europe got a fine and that was it and then they went from being amazing, they went from being Google to be water (ph) companies in like the two years —
EVANS: But say you look at the — all the stuff we are doing off market, this was all kind of in concept videos and presentations from mobile operators in like 97, 98, 99, you were going to do all of this stuff and they were going to do it all.
And of course it all happened but it wasn’t done by them and so that perspective of how we kind of — I think there is a lot of case of that, for example, if you look at cars now, you have all this futurology of what is going to happen and what cars will be and how everything will change and you have kind of the renderings of the glass cars which is like the renderings of the glass phones in 2000 when there were no phones with color screens, and you think we might get like three quarters of this right. But like the last quarter of it is going to be where all the money is and that is the difference between it being Nokia and Microsoft and it being companies you never heard of like Apple and Samsung and Google, forgotten like Apple, and no one have ever heard of Google.
So it’s interesting to look back and think how do we think about what was going to happen then and how much of it was right and wrong or what would you have had to have said then in order to get it right.
RITHOLTZ: So you’re describing a form of historical futurism because when we’ve what we’ve seen over the years is expectations of the future turn out to be wildly either overoptimistic or pessimistic but rarely right on the nose, is that a fair assessment?
EVANS: I think that’s right. I mean as I said, I think there are sort of several bits — there are several lessons you can learn from like — not as much the bubble per se, but kind of the way people thought about it and so there was this whole idea in the early 90s of the famous thing called the information superhighway, which the whole — gives the name conveys the fact that it was going to be kind of centrally controlled and it would be the Cable Company, a News Corporation, and 20th Century Fox and the New York Times and they would kind of get together every six months and decide what you were going to have.
RITHOLTZ: And how did that work out?
EVANS: And of course, what we had instead was comissionless (ph) innovation in the open Internet and there is a central authority and anyone can do what they wanted to some extent.
And so that is interesting if you compare that now with the way, for example, people talk about cars, all the way to people talk about what is happening to TV now. I think the kind of the — as I just said, the fact that you could have got — you could’ve guessed like 80 or 90 percent of it and still have missed the important part is also fascinating. You could’ve said in 19, in 2000, OK, everyone will have one of the things, everyone will have Internet on their phone, it will be a real operating system and all the feature phone, it will be open Internet, therefore it will not be the carriers.
That would’ve — can describe today’s world perfectly, but you would have still have said that well then, it will be Microsoft and Nokia that do that.
RITHOLTZ: What is amazing is there was this AT&T commercial I very vividly remember, I think Tom Selleck did the voiceover and he talks about what the future will be like, and I think the —
EVANS: — on the beach.
RITHOLTZ: You will.
And it turned out to be dead right except for the fact that AT&T has almost nothing to do with it.
EVANS: Exactly and one of the ways I described this when I talk to telcos now is its though a municipal water company looked to the mineral water business and they said you know, come on, we got brand, we got water, we got trust, you wouldn’t buy water from a company you don’t trust, we should be doing this.
RITHOLTZ: Never heard of.
EVANS: And they hire Mackenzie and Wolf Andersen (ph) and they build the whole thing and like two years later, their first pallet of water rolls into Wal-Mart and it goes into the shelf and they look at it next to 200 brands, and they think, hold on a second, something not quite right.
RITHOLTZ: So how did you go from working in London or Paris as a telecom analyst to I know I’m going to move 6,000 miles away and become an Imagineer —
EVANS: I just asked them for a job. I don’t know why people said it is hard to get a job in (INAUDIBLE), you just ask them for a job, they say yes.
RITHOLTZ: That is it.
RITHOLTZ: It was easy. How did you, well, how did you first hear of Andreesen?
EVANS: Well, so there is a process here, and as I mentioned, I actually did my degree in history which is analysis and so the question was sort of picking up and looking for things where you can apply that to analysis and I left University and became a sell side analyst and as sort of many people will know that for many various reasons, stopped being a fun thing to do or even a thing that anyone could do.
So I think when I was in telecoms, Merrill Lynch Europe had something like 50 European telecoms analyst, and the last time I looked, I think they had less than half a dozen.
So that industry have ceased to be fun and so I left and went and worked in strategy first at Orange and then at NBC Universal and channel 4, again both in London. So from sort of sitting on the outside in, sitting on the inside out trying to work out what we do and how do we understand this.
And it’s kind of interesting because the kind of the questions change and suddenly you go a spreadsheet that lists all the stuff you’ve been trying to work out from the public financials and then you just want to work out different stuff but the process is, this whole process of this thing.
So I did that for a while, I left NBC Universal the week the General Electric share price fell 50 percent which is obviously a direct causal relationship.
RITHOLTZ: Right, right, sometimes you know things happen and you can just draw a straight line from A to B.
RITHOLTZ: I recall seeing the Wall Street Journal piece, Benedict Evans departs NBC Universal, GE stock cut in half.
RITHOLTZ: It went up by the same amount the next week —
RITHOLTZ: That was after the financial crisis.
EVANS: That was in the financial crisis.
RITHOLTZ: It was.
EVANS: An old and boring story of what happened there.
And then I went as — I worked as a consultant in London and so I was doing a lot of the sort of strategy consultancy and was producing research reports for around the European media and telecom space and so I was writing both about kind of what’s happening in fashion magazines and record retail and DVD retail would also, what are Google and Facebook doing, what’s happening with smart phones and so I started a blog and more or less the same time I went on Twitter at that time which was kind of new and so you are easy to get noticed if you are doing stuff.
And I was kind of writing stuff, the way I describe it is basically I would write stuff that the sell side analyst or senior strategy person or someone at Apple could write except that either they couldn’t publicly publish it or they would be writing for a different audience, they would be writing for a, you know, buy sell hold audience and or they wouldn’t have the analytic background to write it.
So a sell side analyst could write this stuff but they’re working for different audiences, a senior person at a company knows all of this stuff but they are not used to writing stuff and they can’t publish it anyway.
Somebody in Mackenzie knows all this stuff, but they are not allowed to say that is not going to work.
So I was kind of an interesting kind of niche in the Venn diagram, kind of a little segment on your Venn diagram as somebody who had an analytic strategy background was used writing about the stuff and explaining it, could say it in public at a time when (INAUDIBLE) people were doing it.
And so for those reasons, I sort of got noticed and you know I spent like two or three years writing about posts and getting like 100 page views a month and then I went through a period where I was getting a couple of thousand page views a day and that sort of happened in the course of 2013.
RITHOLTZ: And then —
EVANS: And I kind of picked on my bag and thought, well what else do I want to do, what do I want to do next? Where can I deploy this and you know you have those conversations that (INAUDIBLE) should go and do this venture capital and where is the place to go and do venture capital and the global customer, San Francisco, and what was the kind of firm where it felt like this kind of innovative approach to explaining things and adding value in public would work and A16Z was sort of at the top of that list.
RITHOLTZ: And you reached out to them as opposed to —
EVANS: Yes, so I got a couple of introductions, I went in and I said, well, this is sort of what I do is have a place for this in the firm, does this fit within the firm, is it useful?
And they kind of said yes.
RITHOLTZ: Just like that.
EVANS: More or less.
RITHOLTZ: So lots of other bloggers should be inundating A16Z with their resumes and saying look, I’ve been doing this also —
EVANS: Well, I don’t think you need a resume, it is kind of like, you know, if you want a sales job, you should be able to get a meeting, if you can’t get the meeting, you shouldn’t have a sales job and it is kind of the same, you know, you prove that you can do the job by doing it.
RITHOLTZ: That makes a lot of sense.
So the focus that you have today is no longer telecom the way it was but certainly the concept of mobility as a source of possible changes in technology is a key factor. What are you focused on these days
EVANS: Well I think talking about mobile now is a bit like talking about PCs 10 years ago.
EVANS: Yes this is the center of everything but it happened, and so we are not arguing about iOS versus Android or whether everyone is going to have one of the things or apps versus the web all of those kind of things, there are no interesting conversations anymore, it’s like arguing,, is everybody going to get onto the web, yes, now what? Next question.
EVANS: So you look, we say that one answer is well, what are the next questions? What are the next kind of megatrends happening? The second is there is a lot of conversation around what happens with that stuff, what happened with the stuff that we already had.
So 10 years ago, what we do with broadband and browsers so let’s talk about the — let’s talk about social, let’s talk about what happens once everyone has a PC, now what happens once everyone has a smartphone and everyone is on Facebook.
So we have those kinds of conversations. The world that we just built, what happens to that, and then we think, well, what are the next things? So machine learning, autonomous cars, mixed reality, cryptocurrency, what the next fundamental trends which at the tech industry the way first to PC and then mobile shaped the tech industry in the last 20 or 30 years.
I think then within that, there is a kind of you know, for what I try and do, there is the question of looking for the arguments or looking for the question so one of the places where that is an on the one hand on the other hand is it going to look like this or is it going to look more like that?
Are they going to be — not — will there be autonomous cars? Yes, not, will it be in 15 or 20 years? I don’t know, it is more like are there going to be winner takes all effects in this, is it going to look like Android or is it going to look like ABS where there’s like a widget that everyone buys and it doesn’t matter.
Is what can happen in mixed reality is it’s going to be something that is going to be every manufacturer, was it going to be (INAUDIBLE) concentrated, how do we think about machine learning, how we try and understand what that might change inside the company?
And so kind of there is like what re the topics and then how do we work out what the kind of the useful questions to talk about might be within that?
RITHOLTZ: So you’re less extrapolating current trends out to some future date and instead thinking about well, here is where we are several steps beyond that, what are going to be the subsequent developments that this might lead to?
You are really several steps ahead.
EVANS: That is right, I mean that is partly a consequence of the state of the industry, say like five years ago, okay what — it was a horse race, who is going to win between Apple and Google, what us going to happen, is Apple going to survive even though Android has got this open story and this open source story and so on.
So is there going to be room for Apple? Is there going to be room for a third entrant? What will happen to Blackberry, and will Blackberry be killing (ph) on a niche, will Windows Phone be able to break in?
So all of that, it was the —
RITHOLTZ: I’m sorry, Windows —
RITHOLTZ: I don’t recall that.
EVANS: Yes, so it’s a painful subject for some people.
And so it was a horse race.
RITHOLTZ: Steve Balmer in particular.
EVANS: Yes, he tweeted (ph). But that was — are your (INAUDIBLE) in place, that is American phrase, you know what was going on. Now you know that we know what happened and we don’t have those kind of day by day tactical questions around autonomous cars because I don’t want autonomous cars, we don’t have those kind of day by day questions around mixed reality because you can’t buy a mixed reality headset yet, (INAUDIBLE) glasses yet.
So the character of the question changes because we are at a different point in the S curve.
Because when the S curve is going near vertical, then you are trying to work out, oh my god, what is going on, is the shit going to blow up, when is it going to start flattening out.
Where we are now is we got the old S curve has flattened out right at the top of the scale and we are talking about we even build on the top of it, and the new S curve is kind of under the radar, if I can mix my metaphors.
One of the ways I described it is kind of a good Manhattan metaphors is like we will pass a construction site every day for six months and there is a bunch of construction workers standing around scratching their back sides and not doing anything, and you think, these guys are lazy, then you walk past on Monday morning and they put 15 stories of steel frame and makes you think, well, they were really busy other weekend.
RITHOLTZ: You are missing the slow practical improvement.
EVANS: Exactly, and then there is a period at the end where they were like, putting the facade on and doing all the fit out and again, now, this is boring.
And so mobile is at the stage where you know, you are putting the facade down and you are doing the fit out and the other stuff are still kind of hole in the ground and you can really see what is going on, and so that means the character of the questions change.
RITHOLTZ: In the early stages of investments, there is no data, there’s no discounted cash flow model, you are really dealing with two founders and a PowerPoint presentation and some numbers which are more or less best guesses.
EVANS: Yeah so the way — it is an interesting shift from being an equity analyst because you’re right at one end of the risk reward profile, you know, if you are kind of buying T-bills as of the other end, you’re at the stage where half of the deals you do will return less in invested capital and that’s the plan.
EVANS: And 5 percent will produce more than a 10 X return and that gets you a kind of a 3X return over 10 years and so everything you do has to be capable of being amazing and if it is capable of going from two people in a PowerPoint to being an amazing thing worth hundreds of millions of billions of dollars, it kind of has to be implausible and crazy and there have to be a bunch of reasons why it might not work.
And therefore what you have to do is to suspend disbelief and not think here are all the reasons why this might not work but what if this did work, what would it be?
And are these the people who can make that happen?
RITHOLTZ: Which is a very, very different approach than thinking about all right, will this new widget or this new management team or whatever sell enough to move the earnings-per-share calculus of this publicly traded company?
I think the kind of the key way to think about Silicon Valley is it’s a machine for running experiments and most of the experiments won’t work and that’s the plan. And yes, you know, you could do some experiments, that is clearly a terrible idea and it was never going to work and you could kind of mess it up and blow up the lab and you know, people will look down on you for that.
But no one will look down on your for the fact that you ran an experiment and then produced a negative result, that was just, OK, well, we tried and did you run the experiment right? That’s a different question. If you ran the experiment, it didn’t work, okay, that’s fine, and the ones that do work justify the whole exercise and pay for the whole exercise and produced mobile phones, so a produced Apple and Google and so on.
RITHOLTZ: So let’s zoom in on that a little bit when you’re at this side of the funnel when you’re looking at these really early stage companies, if half of them are effectively money losers for you for the firm as an investment, how do you conceptualize what you’re looking at? Is it the founders themselves? Is it the idea? Is it something that hey this is just so crazy it might work? What is the thought process like when especially where you said where there’s an endless stream of people coming to Silicon Valley to pitch ideas to VCs.
EVANS: Well, so I think most venture capitalists look at this stuff and they see in the same way that you can argue a little bit about emphasis, but there is a question of what is the market opportunity here and are these people who are going to be able to work out find — discover that market opportunity because inevitably, it is not going to be exactly the original idea or it is going to be something sort of adjacent to that, you will kind of twist around and find it.
RITHOLTZ: Is it —
EVANS: Not so much I won’t iterate, I would say. Pivot is more like a we built our company on that premise and that didn’t work at all, so now let’s try something else. That’s not really the same thing.
It is more kind of we were kind of battling and we sort of moved around we got we made we found it worked — there is this whole sort of concept in Silicon Valley of product market fit.
And so you have sort of a product and you try to work out what will the market be and what will the product be around this space until we can find something that meshes and takes off. And so the earlier you are, the more, in a sense, you’re betting on the ability of the founders to find that, but you are also betting on is this a great market and is there some angle or some way that they are going to find in order to make — in order to turn this into a thing?
RITHOLTZ: So what I’m hearing from you is future growth prospects are 90 percent and valuation is well —
EVANS: So, it depends.
EVANS: But the more progress you make and the more numbers you have, the more that you start looking at metrics and the less that you start thinking about potential sort of the super, super early stage, then the valuations will tend to look the same, as you go further in, then you start getting much more specific about what how well are they doing what do we think this looks like?
Then it gets — sort of gets kind of case-by-case and you know, you get up to a company that has got billions of dollars in revenue and then you are doing DCF and you’re doing multiples like anybody else.
But at the super early stage, in a sense, doing a DCF on two people with a PowerPoint is just an exercise in self-deception.
EVANS: You could do — you could sort of say, well, if they managed to sell those things to 2 billion people, will there be some value there? Trying to do a DCF on that, well, what does that get me? It doesn’t tell me anything.
RITHOLTZ: You are just making up numbers —
EVANS: Yes, and the things that really work in a sense, it doesn’t kind of matter if this is and if you got to take any kind of hypothetical example, we got a $50 million seed fund and you make a significant investment in something that turns out to be worth $50 billion, does it really matter if it is worth $40 billion or $60 billion, the amount —
RITHOLTZ: For your 60 to start —
EVANS: The amount that is returned to your fund is sufficient that those numbers don’t really make any difference.
RITHOLTZ: So that is really fascinating and Mark Andreesen (ph) when we sat down and had a conversation said something similar but looking back at a perspective of 20 years of doing this and if you go back and if you would pay double for everything, it wouldn’t mean any difference whatsoever.
EVANS: Yes, I think there is a famous investor whose name is — I sure think he was somebody in Hollywood, and so a famous producer he said something if I said yes to all the ones I have said no to and no to all the ones I have said yes to, I would have come out in exactly the same place.
So there is always, you know, this is kind of the indexing story, that always — that you can always to proceed arguments and say well then we should just nobody knows anything and there’s a little bit more to it than that.
RITHOLTZ: Let’s talk a little bit about technology, you have a quote I really like, you’ve several quotes I really like and let’s start with the one or two of these and see where they go. All social apps grow until you need a newsfeed or newsfeeds grow so you need an algorithm, all algorithmic feeds grow until you get fed up seeing the wrong stuff and leave for a new app with less information overload.
So is the nature of all tech rinse leather repeat and then something becomes an incumbent, becomes successful, and the new guy is just going to come up and eat their lunch.
EVANS: So I have noted through a blog post in my vein which is called something like technology determinism. And so there will — as it might be half a dozen things which are just steady processes and there are half a dozen things that cycles, so steady process, the most obvious one would be Moore’s Law, there is process of technology moves from research labs to startups to big companies, you can imagine my kind of half a dozen of those then there are cycles and so there are cycles where you go from bundling to unbundling. You go from the client to the server and back again.
You go from —
RITHOLTZ: And the public markets we had conglomerates and then deconglomeriztion and then reconglomerization.
EVANS: Exactly, exactly, my colleague here, Steven Sinofsky, he used to run office, he said all products expand until they can edit photographs.
RITHOLTZ: That’s very funny.
EVANS: Word can edit photographs, Excel can edit photographs, all products expand until they can edit photographs.
RITHOLTZ: That is the endpoint of software.
EVANS: And so there were kind of, you know, inside comment about feature creep or whatever it is you want to say and so there were these kind of inevitable processes or inevitable pieces of logic that kind of flow through.
Now the one that you were — sort of quoted particularly, was an observation about — I actually wrote a blog post about this last week, instead of a combination of Dunbar’s number Zuckerberg’s law, Dunbar’s number is like you know like 150 or 200 people that you would friend them on Facebook at the very least, and you’ve got these social apps which make it easy to post stuff and share stuff and because it’s one to many, you are not emailing it to someone or texting it to someone, you can post quite a lot because you don’t feel like you’re kind of imposing on people to do that.
But you got 200 friends and they post five things a day, okay now you got 1,000 things a day on your newsfeed and you can’t read those, and so this is the logic that gets Facebook to producing what is called an algorithmic feed which is just like engineer speak for let us try and work out which of your friends we care about and maybe we should put those at the top.
And let’s work out that you like these kind of thing and you don’t really like news stories from the Guardian, you prefer to look at pictures of babies, so let us put the babies in front of the news stories from the Guardian.
And you kind of — you come up and you create that and so now you’ve got instead of so to speak, a random sample which is open the app, what have people posted in the last hour because I’m not going to scroll the last hour’s worth of post, so actually it is random the random point back to being what time do you open the app?
So that is the linear feed, the chronological feed is random.
So then you say maybe we should the stuff that is important to the beginning but then you think, OK but it is, now we are arguing well why isn’t the Guardian or the New York Times at the top? It should be up at the top because it’s important, there is a public benefit to that and you got this wrong my friend posted the thing I wanted to see and I didn’t see it and so you get that sense of well maybe this isn’t actually working and you have Russians trying to game it and you have all sorts of problems of trying to make that feed work.
Then you can say well actually what I want to do is if I really care about this stuff and I want people to see it, I will send them a text message, I will do it in WhatsApp, I will do it in Facebook Messenger.
But then I’ve got 15 parallel conversations or 20 parallel conversations with people and then we start creating WhatsApp groups where 20 people from school or can work and all talk to each other and then everyone is like posting normal stuff and you think, I really like to have a screen in this app but just showed me like the important stuff from all of these chats that I would see it, and maybe that should be sorted by which of the ones that I want to see and say like you can create the same process over and over again.
Now, this is an opinion, this may be — entirely be wrong, this may not be how it works, but you can certainly sort of see that problem that if you create tools that let everyone you have ever met share anything that they’ve ever been interested in, then you are not going to be able to read it all.
RITHOLTZ: Well, this leads me to the very related quote, 50 percent of Facebook’s engineering effort goes to stuffing more noise into your newsfeed and the other 50 percent is working on ways to filter it out.
EVANS: Yes, or another expression of the same point that is a classic joke I read in my book by Castiglone (ph) which is somebody’s dug a hole to be able to create the foundation to put up a building, and they asked their friend what shall I do with all this earth that has come out of the hole and their friend says we will just dig another hole and put it in there.
Well, just sort of a (INAUDIBLE) you can’t do it like that. And if you created a system that lets anybody you’ve ever met send you anything that they feel like sending, then you are not going to be able to see it all and there is not like some magic algorithms that is going to make that work, you are only ever going to have a sample.
RITHOLTZ: So I want to talk to you about some of your favorite technologies and future projects. Do we want to discuss it all what’s been going on with Facebook this year and everything from Russian bots to the just changes with Cambridge Analytics and scraping?
EVANS: Well, first of all, I want to say that Oxford Analytics is actually not nearly as good as Cambridge Analytica.
EVANS: So I think there is a bunch of kind of unresolved feeling about Facebook, is it private or public? If I post stuff in the newsfeed, will my friends see it or not, what does that mean? Do I want to share? Do I want to share this stuff or not? And like we sort of understand that if you search on Google for something, it’ll show you, or try to show you what you asked for even if it’s something you shouldn’t have searched for.
We don’t really feel like if I eat my racist uncle posted that story on Facebook, should I see it or not? How do we think about that?
RITHOLTZ: Well if you like it for repost it —
EVANS: But we also sort of saying, well maybe Facebook shouldn’t be showing that at the top of the list of the wall, he is my uncle and he did post it. So we have a bunch of sort of I don’t think we have a clear sentiment about the other extreme, we are confident about that, we’re comfortable with the fact that our banks know how much money we have.
RITHOLTZ: For sure.
EVANS: We’re comfortable that our mobile operators know where our phone is and that there is a kind of a legal apparatus around that —
EVANS: And if the police needs to know, then they can find out but it’s just not available to anybody. I think we sort of have a feeling around Google, mostly. I don’t think we have like a resolved feeling around Facebook. Now this particular story is sort of fascinating because of how many pieces there are to that. So Facebook creates this develop a platform that allows you to install an app that can access your information and also access information about your friends and there is a sort of a bunch of reasons why that would be useful, like I wanted to create in a calendar entry or I want to you know, I wanted to see who else is using this app.
And a large part of like the advocacy around the tech industry would hold on with the bad, Facebook is bad because it is closed, people need to be able to get their information out — you need to be able to — other people need to be able to use this to innovate. So there was this whole suit of ideological argument that Facebook would be evil for not doing this and Facebook needed to do this.
And so you create this platform, it turns out that people are able to exploit that and do stuff with it that was not really anticipated.
RITHOLTZ: Do you think it’s not anticipated because from my perspective, I look at it as by the design that it was supposed to be.
EVANS: OK, so yes and no. So let’s get to your analogy.
Did you remember what macro virus is?
RITHOLTZ: Sure, absolutely.
EVANS: So Office is supposed to be an open development environment, the people suing them for antitrust is suing them for making it closed and hard for third parties to work with.
So they create all these APIs and they create this whole macro language. One of the things on page 15 of the textbook is you can make a macro run when you exit a document, page 48 of the textbook, you can get it to look at your email addresses, page 72 you can get it to send an email. Okay, so I get an email — a Word document, I open it, it emails a copy of itself to everyone in my address book.
OK, that is not what we expected.
RITHOLTZ: But clearly that is not its intended —
EVANS: It’s not its intended purpose, exactly, but all of the individual APIs was there and soy you got this period for Microsoft where they are thinking okay so are we supposed to not have macros, are we supposed to be close now? How do we think about this?
And they had to go through this like 180 degree turn as they went from thinking we should make it as easy as possible for anybody to do this to what would happen if — I’m just going to moderate my language, what would happen if a bad person, what my seven-year-old would call a dingley head, decided to read the textbook, because it’s not like you found bugs in it, they are doing stuff within the textbook that was in the manual but you won’t expect them to use them in those ways.
And I think there is a very strong parallel there with what Cambridge Analytica was doing which was you are able to install an app, the app can get your information, the app can also get your friends information, if you say yes it will get your friends information, yet, but we didn’t expect that people would use that to exfiltrate 80 million people’s profiles, just as we can expect some would make a Word document that can email a copy of itself to a Monday people.
RITHOLTZ: So now, let me — I want to go toe to toe on technology with you with you which is clearly a mistake on my point, but Microsoft notorious for having all these weak security setups and easily exploitable, that is such a foreseeable issue granted there is a little bit hindsight bias —
RITHOLTZ: Right, but you know, Microsoft is home of the exploitable security error, my relationship with Amazon and with Apple is that I pay them for stuff and expect a different level of trust and a different level of — hey, I’m already giving you money, don’t exploit my personal data for other reasons.
EVANS: Well, let me give you hypothetical, you can install an app on your smart phone, it can pop up a box and ask your friends, and there is an awful lot of sensible reasons like you are playing a game, who else, which of your friends are playing the game?
EVANS: Install the Instagram, who are your friends so you can follow them like —
RITHOLTZ: If you want to split a fare with somebody, can we access your contact —
EVANS: Many basic logical reasons why you want to do that, okay, so that app has just downloaded 600 people’s home addresses, is that a breach?
RITHOLTZ: Well, has it download their email address or their home address?
RITHOLTZ: Or have they looked at the home address and then send —
EVANS: So it needs their email address or their phone number to work, that is the — there is no other way to identify. So it’s got 600 people’s email addresses. OK, is that a breach of privacy?
Is that Apple’s fault?
Well, maybe —
RITHOLTZ: A little bit, yes, I would say it is a little bit.
EVANS: But it is not that Apple popped up and said that it was his capability, Apple pops up the same, do you want this app to have your address or not and this sort of a point where it’s not kind of black-and-white, it is not like somebody hacked into Google and gave them your email.
EVANS: And so I think Facebook has been sort of — they have this record of pushing this thing to the kind of the outer envelope for the last 15 years and I think what’s happened here, slightly ironically because they actually closed off all of these APIs like a couple of years ago. And sort of what happened is like this stable door was open, if you were a developer, if you were in tech, if you went to the developer event, they stood up on stage and said hey look, we leave our stable’s doors open so you can all this stuff, isn’t this great ?
And their stable doors were open for like five years and then — well, maybe this isn’t a good idea, we are going to close the door now, and a bunch of you will say, evil Facebook, you shouldn’t close the doors, you should allow freedom of access.
And now here we are in 2018 and people go wow, a load of people went into the stable and stole all the horses, what a shock. And Facebook like — yes, we knew the doors were open.
So a bunch of you can kind of see this as they try and work out how we talk about this, how do we think about this stuff?
RITHOLTZ: We’ve been speaking with Benedict Evans of Andreesen Horowitz, if you enjoyed this conversation, be sure and check out our podcast extras where we keep the tape rolling and continue discussing all things technology. We love your comments, feedback, and suggestions, write to us at MIBPodcast@Bloomberg.net.
Check out my daily column on BloombergView.com, you can follow me on Twitter @Ritholtz. I’m Barry Ritholtz, you are listening to Masters in Business on Bloomberg Radio.
Welcome to the podcast. Benedict, thank you so much for doing this, you’re talking the stuff that I really find endlessly fascinating including the responsibility of Facebook to either have an open platform and what that risk entails or to control their own APIs and control who accessed what from your feed.
But nobody is talking about and by the time this airs,, he will have, Mark Zuckerberg will have already done his congressional testimony but there is still a tremendous amount of responsibility on the individual user who willingly said, hey here’s a ton of private personal information about me, try not to mess it up.
I was always too skeptical, I was always to what he wants his information, it can’t be for a good purpose?
EVANS: So I always kind of looked at it from the other end which is I think people have — well, I think for years, loaded from the other end which is I think people have full I think the years and years and years, people have just sort of presumed that nothing on Facebook was private.
RITHOLTZ: Right, that is correct.
EVANS: And therefore just sort of treated as a public forum.
RITHOLTZ: If you’re smart that’s what you presumed or if you were I should say knowledgeable that should have been —
EVANS: I even think beyond that, so I have kind of had people say things like I don’t use Facebook, I wouldn’t say that on Facebook messenger because I don’t want it to be public.
EVANS: And of course, Facebook Messenger isn’t public, I mean, technically Facebook can —
RITHOLTZ: Well, anybody can screen shot and anybody can see —
EVANS: Well, yes, it’s more like people thought it was public stuff on their public profile, and I went, I joined Facebook and I forget when it was launched in the UK, but everyone’s profile was public, I mean it was public if you were in the London Network.
EVANS: And so anyone could join the London network and so it was public. And then so there was a period of like a year when like, everybody I have been at school at the University, we had a profile on Facebook and it was public, and I could go and look, and then everyone kind of turned the privacy things on.
But I just —
RITHOLTZ: And even then, it’s not full —
EVANS: I don’t feel like you would say stuff on Facebook that you expected to be secret —
EVANS: Is what I’m getting at.
EVANS: And I don’t think many people ever actually thought that. Which comes again to my point of sort of unresolved feelings, you can kind of — you can kind of — you can go to the extreme and say, everything on Facebook should have been completely private and everyone should have understood that and everyone you know —
RITHOLTZ: Well that is not realistic.
EVANS: But I don’t think that is actually very realistic, I think it was much more kind of nuanced and fuzzier than that. And I don’t you could argue everything in a WhatsApp group should be private, but I think that’s a very different kind of form to posting on your news feed.
RITHOLTZ: When I ask people, you know, sometimes you have to come out — the discussion from an oblique angle and I like to use radio as an example. I ask people, “What is radio sell?” Invariably they say, “Advertising.” That’s the wrong answer. The advertisers are the buyers, radio sells an audience and Facebook more or less has the same business model.
EVANS: So, I find this interesting because there’s a common meme even kind of the tech industry where people say, “Facebook sells your information.”
RITHOLTZ: No, they sell you as a part of an audience.
EVANS: Well, it’s interesting because in a literal sense, if you’re an advertiser on Facebook, you don’t get — given a zip file of the profile of everybody who saw that. So, in a literal sense, it’s not just true.
In a kind of a metaphorical sense, where they are sort of — they are selling the fact that they know — like even on kind of a metaphorical sense, that kind of struggles me. I struggle with that as a statement, but I think the fact that people have these conversations reflects again to sort of unresolved feelings about how we should think about this stuff.
It’s the same with this idea that’s going around now, that all Facebook’s problems can be blamed on the ad model.
You say well, if they had a subscription model, would they have developed a platform? Would you know that they’ve been posting your personal stuff to your newsfeed, like why would that have made any difference?
RITHOLTZ: And you would still have the same opportunity for apps to come in and scrape that data before they close the barn door.
EVANS: Yes. I mean I’ve spoken about this with somebody, I think you could say, on Twitter, he’s somebody, a head of a journalism school and they pasted like an explanation of what Cambridge Analytica had been doing. In fact, this was three weeks ago. It was interesting simply that they were three weeks on, we’re still having people still writing explanations of what happened because it’s all sort of fuzzy and they opaque and no one quite understands what it means.
RITHOLTZ: It’s complex and people like simple narratives with definable good guys and bad guys. That makes it easier to do. Nuance sort of gets lost on cable television, to say the least.
All right, let’s talk about some of your favorite technologies. Autonomy — fill the blank, autonomous cars, autonomous whatever — how do you see the development of AI in autonomous everything.
EVANS: Have we got another hour?
RITHOLTZ: I do, I don’t know if —
EVANS: So I think — what can we say about this? So,first of all the reason that we’re talking about autonomous cars now is because of what we call AI, which really means — this new technology called machine learning or technology that just started working called machine learning, which offers the prospect that a bunch of problems around autonomous cars might be solvable in a way that they really didn’t seem to be easily sold before.
So in that sense you could say autonomy is a spinoff of machine learning or the fact that we are interested in autonomy is a spinoff in machine learning, but there’s also some other stuff that machine learning does as well.
As we go to autonomy, the sort of — the way I kind of talk about this is I have a slide with a picture of a horse’s carriage from 1985. But it’s a carriage with no horse, otherwise, nothing’s changed. And I think that is what I hear when people say, “Driverless car,” that you’ve taken the steering wheel out maybe but like, nothing else has changed and it will still drive around like every —
Autonomous cars mean cars that drive like people but we are making mistakes, we’re breaking the speed limit. And that’s just a really short sighted way of thinking about this. There are kind of two kind of building blocks to think about.
The first is when we have a fully autonomous world in decades 20, 30, 40 years depending on what you think these cars will look like, and of course you have periods where some places will be much — go much quicker. You might say Manhattan is autonomous only in a week, in 2030 or something.
But at that point, there are no accidents.
RITHOLTZ: Or certainly much less —
EVANS: No, basically no accidents because all the accidents are caused by human error and you have much less congestion because you don’t have traffic waves, you don’t have accidents which causes sort of congestion or something, you can have vehicles on freeways driving 100 miles an hour two feet apart from each other. So you radically change what congestion looks like, you radically change what the vehicles look like.
So you can have a vehicle that will never — you know, calling it on-demand vehicle in Manhattan. It will not go over 20 miles an hour, it can be a golf cart. If you’re going to go to JFK then you would send a different vehicle.
And so you could radically redesign the vehicle in the same way that when got rid of the whole theoretically redesigned vehicle and you can also radically redesign the city or change your assumptions about the city in the same way that we did when we got rid of the horse.
Except that you are no longer trying to just design a city that will constrain the 19-year-old guy on a souped up Chevy Camaro. You’re designing a city based on — you can create rules, you can tell the cars where to go and what they can do and not do. You can have dynamic real-time road pricing. You can say, “Do you want to pay a lot to get there in 15 minutes or are you willing to pay less to get there in 30 minutes?” You can tell the cars at any given instance which road sides you should you be taking and how fast they should be going.
And so, what you have is like a change in the structure of a city that has more — a lot in common with like the way the city changes as a result of the car. Kind of the example I often get which what’s wealth — since we’re in Manhattan, is imagine if you live in Brooklyn and it’s November and you want to go to that cool new restaurant in Manhattan. How are you going to get there? Well, you could walk 10 minutes in the rain to the subway station, “Nah.” You could get a cab, presuming you can get a cab; it’s going to cost you what — $20.00 or $30.00 each way.
You could drive, but one of you can’t drink on the way back and you’re going to have to park and you will pay for parking and it may take you 20 minutes to find a place to park, then let’s not go out to that place.
Okay now, in case we are a fully autonomous world. You raise your watch. You say, “Hi Alex, I need a car.” And the pod is around the block, stops outside your door within 30 seconds. There is no congestion. The pod just takes you there, it drops you off outside the door. If it’s your pod, then it goes and waits for you somewhere where parking is cheap or it waits for you somewhere else. If it is an on-demand pod, it immediately starts driving other people around, so there’s no parking. And then you can go home and you can drink as much as you like.
I mean, remember all of those photographs of European and East Coast cities from before cars and you think — well the streets look twice as wide, because you don’t have cars parked down both sides of the street, it’s what we could go back to.
And say you have like kind of radical change in how we think about what a city is. Therefore, in — what does a gas station mean, obviously but also what does big box retail mean, where are you willing to shop, and what does your commute look like? Today, so I live in San Francisco, which calls itself a city and I work in Menlo Park which is an office park, 45 minutes drive south, if there’s not much traffic. I should say in hour, I’ll admit to 45 minutes south.
Now, supposing there really is no traffic. Well then, I can get there in half an hour. But supposing, I’m able to read instead of drive I might be able to live further away and spend longer in the car because I don’t need to just sit there staring at the road all day. So, where do you live? Where do you commute? Where does retail go?
All those sorts of changes that happen like — the stuff that happens as a result cars, which is basically what I’m talking about. There is this great saying that it was easy to predict mass ownership of cars than how to predict Walmart? How to predict drive-through ATMs. That kind of stuff, those second and third order consequences will flow out of the stuff.
RITHOLTZ: Let’s talk about smart fill in the blanks, smart home, smart phone, smart cars, smart speakers — what is it about the prefix smart that suddenly everybody wants to move in that direction? Software enabled —
EVANS: Okay, so three or four blocks to talk about here, the first block is smart phone and the supply chain, 1.5 billion smart phones sold last year. All of those chips are available, it’s like a fire hose of stuff to make stuff to make stuff with. So before smart phones, if wanted it competing into to something, you had to use PC components. So, an ATM is a PC, but those are big and they’re heavy and they need power and so on.
Small phone components, it’s much cheaper, much smaller, much lighter. So, suddenly you can make a kinetic door lock and like you can get the parts really easily, plus you could go to the internet plus machine learning, means like a camera can actually be able to like tell us if something is moving or not. So you’ve got all this stuff, suddenly the stuff is not working.
I think the best analogy for this is like our grandparents could’ve told you how many electric motors they had. There was one in the car, they had a vacuum cleaner, there was one in the fridge, they owned maybe five electric motors in total, like there was only one electric motor in the car, that’s not major, that was it.
Today, who has a clue how many electric motors are in your car, there’s like 20. If you tell your grandfather that you can press a button to adjust your window, he’d hit you on the back of the head, but that’s just how it works because that’s just the technology got deployed.
The same thing in your home, there was a period when everything was going to have a DC motor. Now everybody has — you have a microwave, you maybe have a toaster, you may have a kettle, you have a blender. Nobody has all of those things. Everyone in this day and age has a rice cooker, everyone in the UK has a kettle, in America, maybe you don’t have a kettle, but you have a coffee machine, but nobody has an electric carving knife.
And so, what happens is like, there were those underlying components of cheap commodities we are in this period of trying to work out what you should do with them and how they should all get plugged together and should they all be the same system or not, should they be able to talk to Alexa or not, and then you’ve got like the industrial logic.
So like, somebody — the Samsung Board sat down and said everything we sell must have Samsung voice assistant, because then people will be more likely to buy the Samsung fridge that talks or the Samsung dishwasher and it talk or the Samsung this —
RITHOLTZ: How is that working out?
EVANS: Well, it’s great, you see this at CS because you can see like the fridge people thought this is fantastic, I mean metaphorically speaking, the fridge people thought this was a fantastic idea, the dishwasher people are like, damn it. Okay, we’ll put the voice assistant into the dishwasher —
RITHOLTZ: But who needs to speak to the dishwasher.
EVANS: But, A that, and B, they want to sell it to people also had an LG fridge and they sold the fridge because they’ve got the Samsung voice assistant and they’ve got Home Kit and they’ve got Alexa in it because they all need to sell dishwashers, they don’t care about the big strategy.
It’s like when every tiny product had a memory stake and there were some like — the phony group who had said everything is going to have a memory stake, but there are some bits that sort of saw this, they say, this is great and some people were like, “God damn it.” What are we going to do with this?
And as with the electric carving knife versus the blender, like some of these stuff will make sense and some of it won’t. It would be quite nice to be able to say to my oven, “Okay, pre-heat the oven to 350 degrees.
RITHOLTZ: I suppose you have to learn it —
EVANS: You go over and walk over and touch it. I have my apartment. I have an infrared sensor in the bathroom, I walk into the bathroom and the light comes on, it’s fantastic. My parents hate it because they get up at four o’clock in the morning and then the lights will come on.
RITHOLTZ: But then you program that to —
EVANS: No, there’s no — anything, it’s just literally, it is just a WIR (ph) sensor from 40 years ago.
RITHOLTZ: So, here’s where the smart world starts to get more interesting. So I have a long twisty driveway because our house is set off the main road and we have lights along it and the first phase was having a switch would put the lights on and put the lights off and in the next phase was having a timer that has the lights go on at 7pm and off at 11pm, but the problem with that is 7pm, there’s light sometimes —
EVANS: I think I saw this house, I chanced upon it as well.
RITHOLTZ: No, no nothing like that, its’ a contemporary house, but the new switch that is literally going in this weekend is built into the light switches. You can set your latitude and that let’s — whatever you set as your light going on 30 minutes after sunset will change throughout the year regardless and you don’t have to deal with it. So, that’s our sort of smart application software.
EVANS: We don’t have any — your phone changes its clock when the time then changes.
RITHOLTZ: Automatically, right.
EVANS: I think there’s a kind of question of where will the complexity set, and this is like the electric carving knife. There will be stuff that just doesn’t make sense.
RITHOLTZ: Oh, electric carving knives have been around for decades.
EVANS: And they never worked.
RITHOLTZ: Oh really?
EVANS: Well, some people have an electric carving knife.
RITHOLTZ: Yes, I could show you where they took a chunk out of my finger.
EVANS: The other extreme is like — the other way to think about this is, there is this old saying that basically a computer should never ask you a question that they ought to be able to work out —
RITHOLTZ: — on itself or by itself.
EVANS: And that used to mean like, you plugged the printer and the computer should know what the printer is. It shouldn’t ask you what printer it is. Then it means your friend doesn’t ask where you are when you call a car because it’s got GPS, I mean, that’s nice. Then it means like the light should know what the time zone is, so the light should just go on and off.
They showed that the tension point on all of those is, is it actually more hassle to configure the things to do that?
RITHOLTZ: Sometimes it is.
EVANS: Sometimes it is, sometimes it isn’t.
RITHOLTZ: So my beef about Siri, which was the leading voice app to begin with before Alexa began to eat its lunch is asking questions that, it should from the context be able to figure out — this is just ping a database, ping a location, software, ping something. It has access to all that stuff.
EVANS: So this is a different sense of the word smart. So let’s talk about voice for a minute.
EVANS: So, we have this thing called machine — the right way of putting this. So, when you talk to a computer, there are three things that are going on. Step one is, it is taking the audio waveform and turning that into text, so it’s just transcribing it and turning it into words.
RITHOLTZ: Which I think is basic math and —
EVANS: Well no, this is something where it used to work okay like three quarters of the time. You remember using dictation apps like —
RITHOLTZ: — naturally speaking.
EVANS: It sort of work three quarters of the time. And it would get better like half a point every year. The machine learning comes along and it goes from working three quarters at a time to working 95 percent of the time 96 percent, 97 percent, 98 percent of the time. The machine learning gives you this radical change in that.
It also — then there’s a second piece which is you need to just go from the text to a structured query. You need to actually work out what the verb and the noun is and what is it asking, which is a completely different computation for them.
And so, when you talk to a computer, let’s say, two very — almost unrelated things going on. Machine learning also, this is called natural language processing. Machine learning also may now work way, way, way better.
Then you have the third problem, which is, okay, I’ve created this structured query, do I have anything to give it to?
RITHOLTZ: And you should have in the internet —
EVANS: So, this is a problem. Which is what you’ve actually got here is an IVR. You’ve got a voice tray, press “1” for this, and press “2” for that. Machine learning means you can now and we’ve all kind of experienced this thing, you phone the airline and they say, “Tell us what you want to ask about?” And there’s only 15 things you could ask, so they could recognize it, the text, and then they could work out what you’re asking about and they can lead you to the right number inside the company.
RITHOLTZ: They’re pretty awful also.
EVANS: But, what we don’t have — machine learning gives us a ways of automating the transcription and getting the natural language processing to work. But you might ask at any one of three for four million things. And so, effectively what I mean when I say a structured query is, you can fill in a dialogue box by talking to the computer and the computer can work out which dialogue box you are asking for. But somebody has to have made the dialogue box by hand.
RITHOLTZ: If you’re a phone company or if you’re a phone manufacturer and you have a few million or maybe a few billion queries that uses your frame of reference.
EVANS: But, so you can make the top 50 categories.
RITHOLTZ: Or the top thousand categories.
EVANS: Well —
RITHOLTZ: And if you’re going to ask something really —
EVANS: The problem is, the curve gets really, really steep, really quickly, and you don’t know what you can ask. So you can ask, you can get the top five things and you can say you can ask for weather, you can ask for timer, you can ask for unit conversion, you can ask for time in a different city, you can ask me to play music it will get the music right roughly half of the time.
This is you know, you ask it to play an album and it plays you like a weird cover of it, it doesn’t understand.
But then there’s like what are the thousand things you might your EA to do? And how many of get really complex really quickly? Like, can you rebook my meeting this afternoon? Okay, that’s not a simple query. That’s like 500 things that you need to know.
RITHOLTZ: Right, that’s obviously that —
EVANS: And so, the problem is, as I said, what you have with the voice assistant is that you have an IVR that will always understand what you ask. Machine learning means it will always correct and lead you to the right number.
What it doesn’t have is a way to have somebody at that number automated, or automate what that person is going to think, and so you have to create those dialogue boxes one by one. So, I will give you an example. Siri learns how to do cricket. No, I said, no, it didn’t. Somebody at Apple sat down and wrote the cricket module.
Okay, now Siri can do hurling. Okay, well, somebody at Apple had to sit down and write that. And then, Siri can do, “Call me Uber.” Well, someone at Uber had to write that and then it can do, “Tell me whether my flight is delayed.” Someone had to write that and every single one of those, some human being has to sit down and spend an afternoon writing those.
And the problem with this is it doesn’t scale. And in a sense, if you could do that, you would have made HAL 1000 and HAL 1000 is not the aggregate of like somebody sitting and writing those one at a time. That’s the other thing.
RITHOLTZ: That’s machine learning and big data and a bunch of things.
EVANS: And so that’s the gap. You know, we have a way to get the transcription and the NLP to work completely accurately. What we don’t have is a way to get to automatically make a system that could answer any question you could possibly ask.
RITHOLTZ: How far away —
EVANS: If we get human beings to think of what all of those would be. So, that is a description of — so let me answer this question in another way. So, imagine in the 19th Century someone tries to make a mechanical horse, there is no — I mean, you’ve seen the pattern drawn and things, there is no law of physics that says you can’t make a mechanical horse.
It’s just that the degree of complexity required to get that to work was impossible in the 19th Century and arguably, it is impossible, I mean, positive dynamics are now trying to do this.
RITHOLTZ: Pretty close, so that’s like 150 years later.
EVANS: Right. And so, what you can actually do is you can make the bicycle and the steam engine and a bicycle is a lot simpler than a horse and it can’t do everything that a horse can do, but it turned out, that was kind of useless anyway.
The same thing with AI, what people try to do with AI until a couple of years ago was they were trying to write rules, so if you wanted to recognize cat, what you do is you do is detection and you do texture analysis and you try and make something that looks for two eyes and roughly the right place relative to two years, and you try and make something that tries to look for legs.
And hopefully, you try and work out how the hell are we going to tell the difference between a cat and a dog, and it would sort of work like three quarters of the time, it would never really work for the same reason that the mechanical horse would sort of work, but never actually work.
And then what machine learning does is say, “No, no.” You give it a million pictures of labeled cat and a million pictures labeled dog and you let the machine write the rules, that’s what machine learning means.
The machine generates the million statements that will allow it to calculate this difference between a cat picture and a dog picture with 92.7 percent accuracy, and that gives us a breakthrough of a whole class of problem it solves. It needs recognition. It solves speech. It solves language, it solves a whole kind of level of pattern recognition.
What it doesn’t do is give you a techy human 10-year-old. It doesn’t give you general intelligence. Basically what you’ve built is you built an enormous spreadsheet with a million statements meshed together and linked together and you can give this spreadsheet a picture of a cat and a picture and it will tell you whether there’s a cat in it or not and that’s all it will do.
So, another way of thinking about this is like we have this fantasy of domestic robots or look around the home and do the house work, we have domestic robots, it’s called the washing machine. That’s what machine learning gives. You’ve got your washing machine, that gives you a washing machine that can recognize family pictures. It gives you a washing machine that can tell me is there strange behavior happening on this network, it gives you a washing machine that can recognize handwriting. It gives you a washing machine that can do X, but it can only do that one thing. It can’t do anything else, and it can’t train you — it can’t wash your dishes.
And so that’s what machine learning gives. It gives you this step change in capability, so one of the — I am sort of giving you a stack of metaphors here, I think a really good way to think about what machine learning gives us is it is like thinking about relational databases.
So relational databases give you this amazing step change in what you can do with computers. Suddenly, you can say show me X by Y. Bloomberg only- would not exist without this, S&P would not exist without this, just in time supply chain would not exist without this. Totally transformed by wealth.
But everything you use now is a robot, but it’s not AI. Or it is, but it’s like — if it’s AI, then everything is AI. It doesn’t get you HAL 1000.
RITHOLTZ: Let me jump in to some of my favorite questions. Tell us the most important thing that people who work with you don’t know about you?
EVANS: I hate personal questions.
RITHOLTZ: Is that true?
EVANS: I mean, I kind of like you to just e-mail this. I am puzzled by these kinds of questions. I don’t know what is it that people who work with me have no idea. They probably don’t know I have a dog.
RITHOLTZ: There you go. Because sometimes, it’s revealing. Some people have told us deep dark secrets from decades ago and occasionally —
EVANS: There is a joke that the most secure form of encryption is anything you say in the second hour of a podcast.
RITHOLTZ: That’s exactly —
EVANS: Here we are, I can say —
RITHOLTZ: Right, it’s you me and a handful friends that are hearing it, nobody else is going to hear it. Tell us about your early mentors who affected the way you looked at technology and venture investing?
EVANS: I would say, it’s not really about how I look at technology. It’s more sort of mental processes and ways of thinking about things and ways to try and answer questions and so the stuff I did at University, you know, people talking to me about medieval history and it’s not this is not — no, that didn’t happen, this didn’t happen.
It’s more, okay, how do you actually understand what we’re trying to think about here. So, like I remember like my first or second, like you know the first — the way Cambridge history works is you do — you could give an essay type lit reading list and you come back and we do essay, and that’s it. You don’t have to go to lectures.
And so, I think my first essay was the one that of course had one screw up and it was like, well, is the King of England more or less powerful after 1066? And you know, I gave a long answer and my professor said, “Look, you have to think about this. What does it actually mean to be powerful?” And power is the ability to get people to do something they don’t want to do.
And so, you can give all of these other stuff about all of the sophistication of the Anglo-Saxon monarchy and they had there in much more sophisticated coinage and blah-blah-blah.
RITHOLTZ: Men were torn and that was then.
EVANS: That was 200 years later, but like, who is more — he is actually more powerful? Look, William the Conqueror can tell people to do stuff and they have to do it and the King of — the King before the conquest came to England and was like, no one really paid any attention to what he said except for the coin — complicated coins.
So, my point is like it’s not so much that someone told me how this is how to think about technology. It is more how to sort of think systematically and try and breakup problems and try and look for pieces and components and try and kind of look at things in those kinds of ways.
RITHOLTZ: Quite fascinating. Tell us about some of your favorite books, be they fiction, nonfiction, technology or what have you?
EVANS: I don’t know. I am sort of — I am always reading and I am always buying books and those two processes are completely independent. I read almost no business books or technology books. Partly because I feel that all business books are basically a short magazine article padded out to create 50 pages.
RITHOLTZ: There are some authors in particular I can name that that’s their entire over —
EVANS: Well, I think — I feel like, if you have written 15 business books on the same theme, then like you’re either a very bad writer or a very good writer depending on your —
RITHOLTZ: Or a very good marketer.
EVANS: Well, that too, perhaps. I don’t know, I think I read things that intrigue me and interest me. The last thing I read was the “Iliad” which I had never read before.
RITHOLTZ: With a history major, never read the “Iliad” in college?
EVANS: Well, so there’s this great line from Italo Calvino where he says, something like the number of essential works is so great, but nobody, no matter how much they are reading has read more than a fraction.
You know, literature is a sample. You can’t read all of the books. There was a period when you could have seen all the movies, now you can’t see all the movies. You can’t listen to all the music. You just have to take samples and take new things that stimulate you or make you think in interesting ways. Particularly things that aren’t about what you work on.
RITHOLTZ: So give us another example from — that’s newer than 2,000 years old?
EVANS: I don’t know. I am just —
RITHOLTZ: By the way, this is everybody’s favorite question, the one — I get e-mails about this more than anything.
EVANS: I read books, and then I can’t remember what it is, what the last book I read was.
RITHOLTZ: What’s the “Iliad” to you, what has really resonated?
EVANS: The “Iliad” really stood out for me, that was probably a couple of weeks ago. I read this fascinating book about medieval manuscripts called, “Encounters with Fascinating Manuscripts” or something like that, but it’s in a college library in Cambridge, and it’s one of these sort of new kind of genre of books which kind of looked like a big hard-back nonfiction book, but they are all in color or the way through, which has something to do with the development of printing technology, which is — there’s loads of books that have color illustrations all the way through which is our books, which is kind of interesting.
And there’s this system about how books evolved. So, there’s a kind of fascinating observation that books or codices used to be rectangular because you made them out of papyrus, so you would lay your sheets, you know you lay the strips of plant kind of cross ways like plywood and you’d make them square shaped because you had those strips all the same length, so that pretty much is a square shaped of papyrus.
And then you move to what’s it called? Animal skin. What’s the word? I’d forgotten the word. My mind has gone blank now, too. You move to parchment, from animal skins and there’s this great line, animals tend to be oblong. So you get oblong pieces and then you fold them up and I am pretty sure, it’s an oblong book and so, when you go from papyrus to parchment, you go from square books to oblong books.
And here we are 1,500 years later, and the books are still oblong.
RITHOLTZ: And that’s just a residual of the earlier technology of parchment?
EVANS: Yes. Exactly, I think those sorts of — I mean, coming back to you know, why is this interesting? I do think those thoughts of how did stuff happen and why did — what is the thread of causation and how things can be completely random is sort of interesting and how do this stuff change how we think about it?
So, on the flight here, I read a book about kind of lighting — the evolution of lighting technology in the 18th and 19th Century as we go from candles to light to kind of lamps that hold the candle and produce five times more light.
RITHOLTZ: How we got to now or —
EVANS: The gas lights to early electric lights, which use one very bright and then we used tungsten filament and then so the evolution of light. This observation that stuck in my mind and I now need to go and check which is an assertion that Dutch — 17th Century Dutch pictures have windows but no curtains because why would you keep the light out?
RITHOLTZ: Everybody wants the light in.
EVANS: You want the light in. And of course, as we all sort of know, light was really expensive and it used to be a lot cheaper, and now, it is not. But I just thought that was really revealing way of thinking about how your whole sense of the world changes when lighting is cheap. Why would you have curtains?
RITHOLTZ: That’s quite fascinating — the book, “How We Got To Now” has — there are six inventions, one of which is lighting and the process of watching the price plummet over the centuries is really — is really fascinating.
EVANS: The interesting thing that comes out of that is that when light was expensive, the state assemblies would stay up all night. The rich people would get up at lunch time and get to bed at like five in the morning because they could afford to, because they could afford candles.
RITHOLTZ: Every else gets up with the crack of dawn.
RITHOLTZ: What you do for fun? What do you to relax outside of the office?
EVANS: In San Francisco? So, I have a seven-year-old and a dog. So, like my time is not own.
RITHOLTZ: Understood completely. What sort of advice would you give to a millennial or someone who just graduated college that was interested either in a career in technology or venture capital?
EVANS: So, I sort of struggle to give an advice partly probably because my career has been a sort of a series of random lunches and kind of companies going out of business, so I won’t apply for a job at Bloomberg, don’t worry.
I think there’s a — the only thing I sort of observed is there is the kind of things that you are good at doing and the kind of mental processes that you are good at, are you good people or not? Are you good at process and project management and making sure everything gets done right? Is that not what interests you? Are you good at trying to kind of exploit and empathizing with people? Are you good at trying to explain and discuss and argue with people and persuade people?
What are the things that you are good at doing and those are not necessarily what your degree was in and they don’t necessarily specific to jobs, so like, I have contemporary, so you went off and become, in England we call a barrister, which is basically a lawyer who only goes to court. They are lingered lawyers, so listen to this, if they need paperwork, barristers get caught —
RITHOLTZ: A litigator is what we call them.
EVANS: Exactly, but that’s all there. Again, it’s a completely different profession, and I have contemporaries who are barristers and you know, I could have done that I could have gone in and argued for the thing and you know, you kind of you go in and you look at the problem and you pull it apart and you work out which story to tell and how to place that into the law and try and persuade the judge and the jury.
So, in a sense, I have those skill sets, on the other hand, I look at these contemporaries and it says, Charlie is probably the leading up-and-coming advocating complex offshore maritime insurance disputes and I say, well —
RITHOLTZ: That doesn’t sound like fun.
EVANS: I hope he has played a lot of schnapps, but he is interested in the argument and the explaining and the discussion. The fact that it is insurance is kind of not — you handle that really? So, I think you have to just kind of work out or what is it — what are the things that fascinate you and what are the kinds of mental processes that fascinate you.
RITHOLTZ: And our final question, what is it that you know about the world of technology that you wish you knew 20 years ago other than light —
EVANS: — by Apple or something.
RITHOLTZ: No, nothing like that, you know, anyone could look at prices 20 years ago and said “Hey, buy Amazon, buy Apple,” but generally in terms of thinking in terms of broad processes what has helped early in your career had you figured it out sooner?
EVANS: So that actually might be a good way of me framing my technology determinants and piece, because it’s a sort of — I mean, the other thing people talk about a lot in venture capital is pattern recognition that this sort of looks like things that will work, this looks like never work, and why.
And of course, the things that never work eventually, they do work and that’s why there is $100 billion account, so that’s also an interesting kind of conversation, but instead of understanding those patterns of, okay, you have to think, okay, how is this going to fit within this ecosystem. This is — what is the timing for this? It’s those sorts of mechanics of the process that are interesting, I think.
RITHOLTZ: That’s quite fascinating. We have been speaking with Benedict Evans of Andreessen Horowitz. If you enjoyed this conversation be sure and look up an inch or down an inch on Apple iTunes or Sound Cloud, Overcast, Bloomberg.com — wherever your finer podcasts are sold and you could see any of the other 200 such conversations we’ve had over the past four years.
We love your comments, feedback and suggestions. Write to us at MIBpodcast@bloomberg.net. I would be remiss if I did not thank the crack staff that helps us put this podcast together each week, Taylor Riggs is our booker/producer. Medina Parwana is our audio engineer/producer. Mike Batnick is our head of research. I am Barry Ritholtz. You’re listening to “Masters in Business” on Bloomberg Radio.