The following is the text of a talk I gave in San Francisco on December 1st, 2016.
The audience was readers of my newsletter, Exponential View. You can sign up here.
This is a long (7,500 word) transcript of the talk. You can scan it to see the slides and accompanying exhibits if that is easier. Or even read it in more than one sitting….
The text was lightly edited for clarity.
Exponential View has a purpose. In between all the emojis and all the spelling mistakes, this is what it’s about:
- That the rate of change is increasing.
- That the changes that we’re seeing are really inevitable. There’s an inevitability about them and that’s not meant as a techno utopianism, it’s just meant as a kind of pragmatic observation.
- That the relationship between technology, norms, groups, identity, culture, the political economy and business systems is going to change and is going to be strained.
- Given that we know all of that, we have a choice of being spectators or be a participant and I hope that what we do and what we do collectively when we read some of the material covered in Exponential View is to start to become participants. I’m not giving you direction on how to engage, I’m really suggesting that we should be participants and spectators.
This is me on my first day at school in 1976. Back then I lived in Zambia in sub-Saharan Africa. On the right is my friend Rehan, who I reconnected recently through Facebook. He is now known as Dr. Freeze and he does non-invasive body sculpting in Orange County. I can get you a good rate.
But I think it’s important, this starting point is important. We often are inspired from where we come from and what the hell was I doing in Zambia?
My dad was trained as economist and accountant, well he is retired now, but then he was an economist and was down in Zambia building the kind of institutions that we take for granted in countries like the U.S. and the U.K. to make the country function. Zambia had just got independence from the U.K. It needed a deeper civil service, it was having to build its legal system, create its system of distribution and so on. I got an early exposure to the importance of economic institutions for making societies wealthier and making them work.
While I was down in Zambia, which is a land-locked country and doesn’t have great access to the sea and this is the 1970s. We didn’t have a vast range of toys.
One the top right is the Lexidata and it’s a mechanical computer that allowed you to do these general knowledge and IQ tests. The orange packs for kids and the purple ones for teenagers and we had that and I absolutely loved it. It’s a fixed number of questions and even after years of playing it I still wasn’t getting everything right.
But the other thing we had was this Binatone video game console. It had one-bit graphics & plugged into your TV monitor. We had a bunch of games like Pong and a basketball game and of course it’s got a shooting game as well. In fact the only photos I could find of this are low-res because back in the late 70s everything was in low-res.
You can see that I grew up in the context of building institutions in a society that didn’t really have what those that are functioning and early technology. The Lexidata doesn’t even have a chip in it, it is like mechanical computing device. The Binatone does have a chip in it.
These things — institutions and chips-have really influenced me over the last 40 years and six months since that photo was taken.
The age of technology
We live in the age of technology today and nothing tells that story more than this chart. This shows how market capitalization has shifted in just the last 10 years.
Back in 2006 the world’s largest companies were oil companies, conglomerates, electric companies, and finance. Microsoft was in that list. Yet today it’s essentially all technology firms.
That is a phenomenal change in value that has moved from one sector to another. You can calculate it, it’s like two trillion dollars of new value just in 10 years just amongst the top seven firms.
We absolutely live in the age of technology.
It’s important to think about what technology actually does and what technology is and we work in the tech industry but what is it really about?
If you take a look at this chart, looking at 800 years of income in England. You see this curve that goes up and to the right. I know some of you have used this particular chart in your startup business plan but this is real data. This is real data and essentially technology makes us wealthier and only two things really cause problems in that upward trend.
The first is disease. You get the black death, which is a plague that hit Europe, and that makes people poorer. The Spanish flu after World War I does the same thing.
And war. World War I, World War II, the English Civil War, also makes people poorer and what we’re doing through technology is we’re reducing the amount of disease that there is and actually we live in a more peaceful time than we’ve ever really lived as a species. We’re doing a pretty good job certainly with cataclysmic wars and managing them.
And in general it is a positive picture.
We’ve created this global village with free information.
Just think that 10 years ago no human had a smartphone and today 50% of the planet, give or take, has a smartphone. (Note: See a clarification of this.) That is absolutely incredible.
In the next five to 10 years there’ll be a couple of billion more people with a smartphone and it’s worth thinking about how fast that change has emerged.
On the right, you see a screenshot here of RML Farmer. It is a service for Indian farmers, very poor, living out in remote, rural areas and it gives them basically a Bloomberg or Reuters terminal in their desktop, in their hand.
I was lucky enough to incubate this business in 2006 when we delivered it to feature phones, which was a challenge. Ten years on these farmers have access to commodity markets and pricing to make their daily decisions better.
Technology is ushering in a new era of self-expression in all sorts of ways. It’s not just vertical video and Snapchat. Just consider emoji for the sake of argument. When Apple introduced the emoji keyboard back in 2011, you see this incredible rise of the use of emoji within Instagram.
This is Instagram’s data and the decline of the ‘WTFs’ and the ‘LOLs’ because people want to express themselves in new sorts of ways. Just think back to the kind of expression humans had two or 300 years ago and it was much more limited. (Emoji are even making their way into the workplace.)
We’ve gone on to stunning breakthroughs in artificial intelligence as well. Do you remember earlier this year when Google DeepMind’s AlphaGo defeated the world Go champion? I ran around my house yelling, “It’s just like the moon landing and people aren’t saying enough about it.”
I think what’s particularly interesting about this it wasn’t so much that there’s this computational space of 10¹⁷⁰ possible moves, it’s too hard to brute force, and so you have to come up with new approaches.
It was … it was move 37. In one of the games AlphaGo made a move that people didn’t understand. It had played tens of millions of games of Go and it made a move. Lee Sedol seemed taken aback and commentators described it as beautiful.
This is not Deep Blue brute forcing it’s way through chess. This is not a clever algorithm figuring out how to optimize a search space. This was an algorithm figuring out a new way of thinking about something, a creative way of thinking about the problem.
That was the piece that is quite stunning about what happened when AlphaGo played these games and now has kicked off a lot of Go masters trying to understand new lines, new strategies, that they hadn’t thought of in all the studying that we as humans had done of the game.
We’re also seeing these incredible breakthroughs in solar and clean tech and storage. Solar power as you know is declining very, very rapidly. More so than many optimists thought and towards the two cents per kilowatt hour range, which is unsubsizided competitive with anything in the world.
Of course solar needs its buddy, which is storage, and what we’re seeing with storage is also phenomenal. The chart of the right shows energy density in watt hours per litre for electric vehicle batteries. It’s gone up six or seven times in the last seven or eight years year with the resulting decline in the price of batteries. These are incredible, incredible outcomes.
And then there’s breakthroughs in transport. If you think about the primacy of the car, what the car has done to make our lives much richer over the last 100 years, it is mind-bending.
On the other hand, think of what cars now do to our environment, and the city space they take up or the supporting oil infrastructure. The amount of city space the automobile takes from us. We’re the residents after all. And the amount of time we end up spending in the car and taking our attention, the breakthroughs in transport will be paramount.
What we’ve done is we’ve made the rare abundant, even in some quite odd places. Back in the 70s we never had fish in Zambia because it’s a land-locked country. Yet when I came into the U.K. 1980, there wasn’t a ton of fish available.
We figured out some technology around farming fish and there’s that curve that came from a startup business plan again. We’re now producing more fish protein than beef protein. As many of you will know that comes with lots of questions about what actually is the quality of this fish protein for humans. What are the shortcuts or the problems that emerge from intensive fish farming?
It’s fair to say we’ve come a really, really long way. Today we produce 2,700 calories of food for every single person on the planet.
And the average human needs 2,100 calories per day. The average American has 3,800 calories a day.
We have finally got to a point where more humans die from overeating than die from under-eating because there are more deaths from obesity related diseases, like CHD and diabetes, than there are from malnutrition.
In fact if someone does die from hunger, it’s not because of the lack of food, it’s because of a political choice because there is enough food to go around. Ultimately famines are caused by government or pseudo-governments deciding that they need to do this. (Note: this point is made well by Yurval Harari in a lecture I attended in London in 2016.)
There’s this fundamental trend of scarcity to abundance, which sounds really, really great and it’s obviously been fantastic in many ways but it does start to raise questions.
There’s this massive challenge as you know from the newsletter of climate change, which is pretty terrifying in all sorts of ways.
Also these questions around our political consensus and what are we seeing in this sort of funny year, 2016. We lost David Bowie & Leonard Cohen, but gained Brexit and Trump. What is that telling us? And if you’re somebody who observes systems as I do, there’s something interesting that’s happening.
There’s something that bears investigation.
Now I want to use a snapshot of a framework that I use to analyse these issues. I’m just going to touch on these four areas:
- the increasing rate of change
- the incredible growth in efficiency and improvements in efficiency
- the questions around what’s happening with equality and inequality
- the growth in homophily
The rate of change increasing.
We’ve all heard this but here’s actually some data from Horace Dediu from Asymco, which shows going back to 1800 the amount of time it took for certain technology product innovations to penetrate the U.S. The y-axis shows percentage penetration. You can see is not only are the curves getting incredible steep and there are so many more in any given period of time. (I recognise that this particular graph doesn’t show it as clearly as Horace’s data! Here is another view from a different source.)
The kind of product deployment time from an innovation to it being with 20%, 50%, 90% of the population is getting shorter and shorter and shorter. The rate of change is really, really increasing.
The challenge with increasing growth rates is sort of two-sided. On the one hand, the happy side, is that stuff gets better much faster. Difficult diseases get treated. Annoying things that happen in your daily life, like having your food go rotten, get treated because you have a fridge. That’s great but there’s also this issue of discomfort from change.
If you put yourself in the position of somebody who was born in 1920 where the cycle time of products like to the point of which they’re mainstreamed might of been 30 or 40 years. You’re going to live 70 years. You’re going to see two major changes. Radio to TV and then the widespread deployment of telephones. That’s what you’ll see and it’ll take time, a few years, to get into your life and change your life.
Picture the scene for somebody who’s born in 1980 and will live to 120 or 130 and will have an average cycle time of maybe five years but if it’s really getting faster maybe three or two or one year.
The number of significant changes, changes equivalent of going from broadcast TV to internet or fixed-line phone to cell phone, you might see dozens of those in your lifetime. You’ll have to cope with them and understand how to deal with them.
Incredible improvements in efficiency
Since 1980 we’ve reduced the number of people we need to produce a million dollars worth of stuff. It used to be 25 people and it’s now six, so it’s a quarter as many. That’s manufacturing efficiency and that trend is going to continue.
That continues to take a look at the automating of the company. GM back in 1979 was a massive company. Actually might’ve lost money that year but on the top-line revenues it was huge, the fourth biggest firm on the NYSE with a market cap of $15bn and 800,000 employees worldwide. And much bigger by headcount than Microsoft of last year with it’s 100,000 or so. Facebook, which is same market cap as Microsoft, and creating a lot of value has less than 15,000 employees. That trend graph extenuated to the asymptote.
That leads to huge value per employee. So revenue per employee in blue, market capital employee in red. GM doesn’t even register. One of the greatest companies in 1979 probably.
Microsoft, which is an amazing company, compared to Facebook, where revenue per employee in Facebook is more than double that of Microsoft’s and market cap per employee, you know, five, six times more.
But it doesn’t stop there, right? This high efficiency continues.
Google is a phenomenal company as is Facebook. Look at what happens when we look at the number of software engineers it takes to support a million monthly active users.
Google, this amazing company, doesn’t even show on my graph. Facebook barely registers. There’s Instagram around the million level and WhatsApp up at the 12 million level. This is the same curve, the same trend but it’s increasing in leaps and bounds as we have this trend to efficiency driven by the improvements in technology.
So we might even find ourselves getting our first one-man unicorn or even a zero man unicorn, with a DAO and some AI slammed over the top. Something that creates a billion dollars worth of value without any humans involved.
Growth in homophily
That leads us to the third of the trends that I wanted to look at, which is this issue of homophily or what’s known as ‘filter bubbles’. Filter bubbles are not a new thing.
Homophily means birds of a feather flock together and it’s something that we’ve understood for a long time.
Every city has a Chinatown and a place where the Italian community lives and a place where all the doctor’s surgeries are. They don’t mix and mingle at a house by house level in general. If you look at your friends, your friends are quite similar to you.
That’s birds of a feather flocking together. I mean if you hung out with people who are completely different to you with no shared interests, conversations would not be good. Godwin’s Law would dominate.
We’re familiar with that, even in our physical architecture in our physical world and we realize the benefits of it. The Rothschilds in the 18th century realized the benefits of keeping the banking in the family network. The business schools of the 20th century realized importance of the HBS alum network and so on.
But of course in the digital sphere we can accentuate those filter bubbles and this is just a little graph that shows some discussion around the Israel and Palestine kind of ongoing conflict crisis. In the blue is all the people talking about Israeli terms and the green is all the people talking about the Palestinian terms and frankly there’s no bridge apart from the word ‘war’, which is slightly depressing. There’s no bridge and this trend of homophily accentuates.
We think about it just in the basis of Facebook and in Facebook news but it happens in the real world and it happens in our non-Facebook relationships.
Growth in inequality
The next trend just to think about is the growth in inequality and the graph is showing the share of U.S. income of the top 0.1% and what you’ll notice is that the level of inequality, big is worse here, is as high as it was in the 1920s and the 1910s. The steep rise, really since Ronald Reagan frankly, has continued is pretty staunch.
Labor’s share of value, that is what portion of the income of a country goes to the people doing the work rather than the capital that is being employed, is declining significantly and it’s declined by 15% over this long-term study, 60 years down here where as it used to be up.
In other words people are getting paid less of our national income and connect the dots back with the efficiencies that we talked about earlier and capital is getting more and more of the reward.
This is happening in all markets. Look at this longitudinal data for Spain, Italy, and the U.K. and you see these negative numbers. The only country which has seen the return to labor improve recently is Russia. This is not your exemplar.
And the gains are really uneven. I think we’ve talked a lot about the Branko Milanovic elephant chart in Exponential View but what’s happened up here is this is us, up at C doing really, really well over the last 30 years and this is A, this is the middle classes in India and China and they’ve done really well and here is the Western working and middle class, who’s not really seen their incomes improve in 20 years.
First bit of audience participation. I’ll ask some questions just to illustrate this point. How many of us know somebody that earns, as income, $100,000 a year or more? Just put your hands up. Okay.
How many of us know somebody who earns, say, a million dollars a year or more? [All hands go up]. We all know venture capitalists!
Let’s turn to wealth. How many of us know people who’s assets, non-housing assets, are 10 million dollars or more? [All hands go up].
A hundred million or more? [Most hands go up].
A billion or more? [Most hands stay up].
Ten billion or more? [A few hands stay up].
Anyone more than 75 billion? So nobody knows Gates or Carlos Slim.
Okay let’s go the other way around now. How many of us know people working full-time earn $10,000 a year or less? [About half the hands are up]. That’s more than I thought.
How about less than $5,000 a year? [Very few hands stay up.]
This is an interesting point to demonstrate homophily. Let’s look at the global wealth pyramid, this is from Credit Suisse First Boston. The thing is this it the most disingenuous graph and I’ll explain why in a second. The yellow apex is the group of people who have more than a million dollars and it represents not 0.7% of the population. 33 million adults and we all know somebody in that space.
This is base blue layer represents the people that have less than $10,000. It represents 3.6 billion people and probably, including kids, six billion of the planet and basically very few of us know anyone down here.
This CSFB chart is disingenuous because what it should actually show is the size of the wealth held by each of these groups because this tiny top group actually has as much wealth, pretty much give or take, as this massive bottom group.
The challenge with inequality is it really hard to fix. There’s any number of reasons why fixing inequality matters. But I’m going to focus on two non-controversial ones.
The first is that the evidence is that inequality slows down economic growth. You see this negative relationship here, this is the Gini coefficient which is a measure of inequality or equality and this is growth rates. It’s shown that, for a variety of reasons, inequality slows down economic growth. In other words it makes us all poorer.
The other thing is people, when you ask them, see inequality as a major challenge. This is a 2013 survey and even in the U.S. 46% of people thought the gap between the rich and poor was a big problem and in emerging markets it’s really significant as well. In Japan everyone’s had a shitty time for 20 years so not so worried about it. (See also Pickett and Wilkinson on equality.)
Putting it all together
If we just put all of these things together the increasing rate of change, which is that stuff is coming at us much much faster with lots and lots of benefits but everything is there for increasing, including the increase in efficiency that we started to see.
Smaller numbers of people being able to generate larger returns in a traditional capitalist system. With that comes the increasing inequality and a real asymptoting.
This is not a Gaussian distribution where there’s a nice fat middle of people and the number of people who are super wealthy is the same as the number of people who are less wealthy.
This is a power law where the extreme end is getting very, very wealthy and the bit in the middle isn’t.
Add the growth in homophily, which is like-minded people and people who have similar attributes, hanging around together and the architecture of our media is to accentuate that difference.
The issues: externalities, adjustment costs, trolleys
I’ll leave that idea to sit with us for a couple of minutes and just quickly run through three other things that are important.
- One is about externalities
- the second is about terrible clip art and adjustment costs, and … By the way if you type in adjustment in Google Images, it suggests yoga as an additional word.
- The third is about trolley problems.
On to externalities … This is Lois Lane and she’s posted a photo of her and Superman on Facebook. Facebook’s AI system has said, “Do you want to tag Clark Kent?” She’s realized something.
Externalities in sort of layman’s terms are just unintended consequences but in economic terms they are when as a result of a transaction that was about improving social welfare.
I buy a chocolate bar, candy bar, from you which is a great transaction. I get the candy I want, you get the money you want. I then take the wrapper and throw it in the street and it ruins everyone else’s experience.
I’m not being priced for that cost and the worst externality of course that we know of ’cause it may destroy our species is greenhouse gases but there are many other externalities that emerge.
During the global financial crisis the investment banks packaged up all sorts of strange risks in these CDOs and CDO squared and with the fundamental bet that when they polluted the world, the government would bail them out. They socialized the cost and that was an example of an externality.
And externalities often arise out of real unintended consequences.
If you look at sugar. Sugar was something that was really rare for us to have and so it had value as a delicacy. It was hard to cultivate. At some point in the 1800s because of English blockades preventing sugar from the West Indies reaching Europe, we developed the technology to cultivate sugar beet and get sugar from those beets in the ground rather than just from sugarcane.
We massively improved the amount of sugar production. We effectively started to gather infinite sugar which we could plug into people’s systems, overwhelming our body’s homeostatic functions and thus triggering the rise in obesity. (See Ian Leslie’s excellent article on sugar. Interestingly one of the key papers connecting sugar to arteriosclerosis and obesity was published in late-September 1972, the week I was born.)
The people selling the sugar at this point in the 1800s probably didn’t really think about the obesity because they didn’t know about it. This is a great example of an externality that happens just because of some efficiency improvements that happen several decades earlier.
The sugar lobby started to gain power and be wealthy, they started to support research and use that power to pressurize and change the debate. The industry sponsored the first research project into coronary heart disease to play down the signals that sucrose was a risk factor in CHD.
They paid some Harvard scientists 30 pieces of silver or something in order to do that 50 years ago. (It was actually $50,000.)
Little did they know Exxon was doing much worse around climate change around the time. It was a similar pattern and then in the 1970s and beyond you have the U.S. Food Administration food pyramid, which gives the wrong advice to families.
I think it’s important to recognize these things.
Adjustment costs and their context
We really need our world to function in order to function in the world.
What we’ve seen over the last 30 years is that the products that we most need because they create citizens who can live in an healthy and informed way, college tuition fees, childcare, health, are getting super expensive.
The products that we honestly don’t need, TVs, are getting really, really cheap.
One reason is that if you don’t have healthy, happy, educated, informed citizens, you construct friction and friction turns into fighting.
This is a map of 1848 Europe where there was a whole range of revolutions. It was called the Summer of Revolutions or maybe the Whole Year of Revolutions. But essentially from when you look at this over longer historical period, you start to see that when there is a divergence, a misalignment, something needs to give and someone has to give something up to somebody else.
A great example was the end of the 19th century to the early 20th century, labor workers felt that they weren’t getting enough of the pie and they weren’t being treated fairly enough. In some places they changed the political consensus and they got union laws and worker protections or they got the right to vote.
In other places where the system didn’t adjust to them, there were revolutions, right? 1917 in Russia.
It’s really important to think about what adjustment costs are because we tell the story of technology has never cost jobs, net-net.
Technology has always helped create more jobs for forever, or at least the last 140 years. That is true. It’s a great story. And then we talk about “adjustment costs”.
Let’s try to understand what those adjustment costs are.
The U.K. had a very, very thriving coal mining industry. Really as the industrial revolution took off, the James Watt steam engine needed coal and a whole industry borne of innovation, entrepreneurship and resource extraction was born.
At a point at the start of last century more than a million people were involved in the industry. Fast forward to 1980. I arrive from Zambia wearing that jumper.
The coal industry has been in a long decline, mostly because of concerns around air quality in cities. But there are still 250,000 people involved in the U.K. coal industry clustered - homophily, again – in these little communities that were built around the coal industry and that’s all those communities did.
Margaret Thatcher, who was our version of Ronald Reagan, decided that she was going to break the power on the mining unions and shut uneconomic coal mines … I’ve compressed a lot of history there, which she did very, very quickly.
We can now go off and say, “What does an adjustment process look like?”
This is a social-science experiment. So with all the cab drivers and truck drivers going to lose their jobs, it’s just an adjustment process.
We have the same experiment and this is what it looked like. Those are police officers, PCs, police constables, and these are miners. It was a really violent on both sides and a nasty process of adjustment for a year that really scarred communities. Thatcher was coming out of this monetarist, early neo-liberal caucus believing that type of economics. She didn’t believe in welfare nets and social provision for managing the adjustment.
But the impact lasts a really long time. If you look at these communities 30 years later, this is 2013, 30 years later, these are the communities and this is their out-of-work benefit claimant rate and it’s 17, 16%. It’s twice of the wealthier parts of the country. Essentially when we went in and it wasn’t technology, by the way, that did this but it’s just another thing that demonstrates the adjustment issue, and we didn’t put in place the right kind of systems, we created a really difficult setup that has lasted three decades.
Introducing ethical problems of software
We know software’s eating the world. It’s been five years since Marc said this but what does that mean?
It means that software is now the interface by which we access all of our resources, whether it’s as a consumer, as a parent, as a citizen, it’s all done through software.
I want to check my bank account, I go to my phone. I want to buy a ticket, I go to my phone. I want to figure out what’s wrong with me, I go to my health chat bot.
It’s all moderated by software and so in a way product managers are starting to be the people who control access to the resources that we get and those product managers are … Guess what, they’ve got quarterly OKRs.
The question here is, “Is this going to give us indigestion?” Two two little examples of indigestion. The first is fake news which I won’t talk about.
The second is about algorithmic control. In my day job I work at a Norwegian media company and this is Espen Egil Hansen, the editor of one of our oldest papers. (It is about 150 years old.)
He had written a story about war reporting, the importance of war reporting of getting the right kind of discourse around why countries are at war. He published this photo amongst nine others. This is the most famous war photo that our species has. It is of Phan Thi Kim Phuc, a nine-year old Vietnamese girl. She had been hit by American napalm and she’s in pain.
Facebook determined that this was unsuitable content. It was controlling what actually was a fundamental political point, which was war reporting allows us to figure out whether a war is just, whether a war should be continued, whether we should be putting lives at risk. Whether we should be spending money on it. And this is software eating the world, the worst side of it by the way. There are a lot of great sides but this is the bad side of it.
The challenge is that everything starts to look like a trolley problem. The trolley problem is this philosophical game from English philosopher called Phillipa Foot.
The set up is you’re here at a fork in the tracks and this is a trolley coming down. You see that if the trolley runs down it’s normal path it’s going to hit these five people who, for some reason, are lying on the tracks. You can pull the switch and you can send it down this branch and it’ll hit that person over there and it’ll kill one. The test is what do you choose to do?
Onto our second audience participation. In that scenario, I’m not going to give you any more information, how many people pull the switch and kill the innocent person but save the five lives? Just show of hands, how many people do that? [nearly all hands go up.]
Wow you’re a harsh crowd [reflecting on how utilitarian they all are.]
Okay so how many people let the trolley run on its own? [Very few hands go up.] Okay, actually more of you are utilitarians.
You’ll kill the one innocent person rather than the five.
Now let’s think about this scenario.
You’re sitting by the switch and the trolley’s racing down. You’ve read the news and you know that five murderers have escaped from San Quentin. They had been Wall Street hedge funders who had gone crazy, defrauded their investors, and then killed, not their lawyers, but killed some innocent people.
They’re now in prison and they’ve escaped and you can see them on the nuzzle that this has happened. Everyone is sharing on social.
They’re there and you remember seeing this other story in Nuzzle about this orphan girl who had grown up in terrible circumstances and taught herself to play the violin. Everyone thought she was going to be the new Vanessa Mae. She’s crossing the train tracks during her 17 mile walk to her violin teacher. She’s dropped her violin on the tracks. Picture that.
Let’s now ask how many people will pull the lever to save the five and kill the one? [Virtually no hands go up.]
Okay. Our view has really changed hasn’t it?
We’ve made a moral judgement. We’ve really made a moral and an ethical judgement. A hard judgement. Today at some point these decisions are going to be made autonomously in situations a bit like this but in reality they’re being made every single day. For a lot of our discourse, the person at that switch right now is Mark Zuckerberg.
The thing is Trolley problems are really hard to solve. They’re just not that easy. We don’t even have agreement in the room apart from saving the violinist — and even then several people thought we shouldn’t-but we don’t have agreement in the room.
If you look at what happens with these two scenarios. One is a Trolley problem involving death and the other is just monetary loss of the same value and you go to the U.K. or other WEIRD nations. Western educated, industrialized, rich, democratic, they’re pretty utilitarian, right? Save the five, kill the one. If you go to China they’re much less so and the theory is it’s kind of Confucian “let fate run its course”. (See Cultural differences in responses to real-life and hypothetical trolley problems by Natalie Gold, et al.)
We can’t agree in a group of people who are pretty much birds of a feather. Every single one of us knew someone worth more than 100 million dollars and earning more than a million dollars a year. We all work in the same industry and we can’t agree.
And then we have this issue about what do we do with other societies? These are quite hard problems and they’re not just gonna happen with self-driving cars, they’re going happen because, especially with implementing AI systems, they’re will be commonplace.
Why machine learning brings ethical problems to the fore
The traditional programming metaphor (slide credit: Brocade) is that you take some program, you throw it some data, and out you throw some output. You know what you want out. The machine learning approach is that you take some data, you know what kind of output you want, the training’s set, and you generate a program and it executes really mercilessly.
The issue is that the data that you put in determines what comes out here. And so this new engineering approach of machine learning may struggle. I’m sure many of you are familiar with Word2vec, which is a way of basically understanding high order representations of word meanings.
Word2vec can do things like say well, “If it’s man to king then it’s woman to queen. If it’s Paris to France, it’s Tokyo to Japan. If it’s he to doctor, it’s she to nurse. If it’s he to realist, it’s she to feminist.” That’s learned from the corpus that we feed the machine.
It’s not-quite a Trolley problem, it’s a philosophical problem, it’s an ethical problem. From a separate academic paper in our results European-American names are more likely than African-American names to be closer to pleasant than to unpleasant because the corpora of text. Corpora of text reflects a society at the time the corpus was gathered.
It’s 61 years to the day actually since Rosa Parks didn’t give up her seat on that bus and if you think about training a system on Breitbart, you‘ll get a very different set of outcomes from that system.
And progress is slightly Janus-faced because it can be fantastically amazing and difficult at the same time.
Just look at cities. Cities are these hubs of innovation and education where people meeting together, innovate, mix and mingle. Wonderful things, cities have been great for humanity to have these great, environmental footprints in general. But they’re also the origin of populous surges everywhere. Take a look, this is data from July this year.
It was written just after Brexit by Phillip Auerswald. The Turkish vote pattern for the populous is all in the blue area, which is not where the cities are and he was back then showing where the support was likely to come for the current election. What Phillip says is:
Cities are humanity’s greatest invention. They’re platforms in which we share, create, and exchange. They benefit from density but it causes equality and invites a backlash. Achieving exclusive prosperity requires working with, not against, these fundamentals.
One of the reasons why we have this conversation here is that entrepreneurs are really good at getting things done. I’ve made everybody a little bit depressed for the last 20 minutes and now let’s cycle back.
The entrepreneur’s role
We’re fixers as entrepreneurs. We fix things. We look at hard problems, things that can’t be done and aren’t believable when we say them and we try to create something new to address that need. Who else is going to organize all the world’s information?
I really want to address what that fixing might be.
One important thing is about on the one hand fixing broken local communities, which have been really hollowed out partly as a result of business model change. Local newspapers have disappeared, communities have been hollowed out. The source of community information is now talk radio and we’ve read all of the rest of the stuff that’s happened in the last six weeks [since the US election.] [Note: this is obviously a very long term trend. See Robert Putnam’s Bowling Alone, published in 2000.]
On the other hand cities are becoming incredibly important and it’s sort of inevitable that people will flock to cities. They are a cradle of collaboration and creativity that need to be protected in some sort of way.
The other problem is the platform problem. The platform problem is essentially this. If you think back to the share of labor to the share of capital graph, it’s sort of 60% labor now, 40% capital.
Just thinking about Uber, do you think in all the value that’s been created by Uber the share of value that’s going to the capital, the shareholders relative to the labor, the drivers, is 40% capital, 60% drivers? It’s more like 99:1, right?
And that’s a problem for some of the other transit we’ve talked about because we’re accentuating it. By the way it’s doing nasty things to us . My Uber doesn’t come in a minute, cancel, try again. Cancel, try again.
Personally, I’m really interested in things that make marketplace platforms avoid those dynamics. And that brings me to Juno out of New York, a new ride-sharing platform.
Uber’s vig is 25% on a journey. Juno takes a 10% cut but more interestingly 50% of the share capital of Juno is in a pool to be allocated to the drivers, which they will earn over time as they do more rides. So what’s really fascinating is Juno can still construct an amazing business there but it’s a community-owned business in some sense.
It’s also very hard for a business like Uber to compete against. What do you do? Do you say to your shareholders, “Well we’re worth 70 billion but in order to compete with Juno we’re going to give 36 billion to drivers.” It’s not going to happen. I think if you’re making marketplace platforms this is something to think about and think about it in context of new localism.
We have to figure out that consensus between labor and capital. Capital is taking the lion’s share and work isn’t paying and the gains from automation are things that we need but they have other costs. A lot of people are talking about universal basic income, which I think is interesting but in the context of the power law distribution, it doesn’t solve the inequality because people are going further and further away. You can improve people at the bottom here but they still don’t have the level of equality and self-esteem that all of us need for other things to work.
We also need to fix the architecture of social discourse, so we don’t really have a common or shared language because we’ve used technology to accentuate filter bubbles and accentuate our sameness. We saw in the graph above on the Israeli-Palestinian conflict, opposing sides often don’t have a common language.
In general technology has shaped us for the better.
The chart shows cranial capacity of humans and our lovely ancestors and what you see for a seven million year period is a linear improvement… Nothing much is happening for years. Then there is this uptick and, wow, we’re up and to the right.
The possible cause that was our command of fire. Fire is a technology, which allowed us to improve our caloric intake. We didn’t have to spend as much time eating and we had more time to then socialize, plot, plan, build culture, and become the most successful species in the planet.
But I think we’re also starting to ask questions about technology. There seems to be a sense of unease. We seem to be asking more troubling questions about the purpose and impacts of technology. Whether we are in the industry or outside, there is this sense of unease.
This chart the growth of mindfulness research. It is also an exponential curve. People are researching mindfulness and trying to understand how you achieve it.
This question seems to correlate with a bunch of other things that we have covered tonight… because we are most likely asking those questions and searching for those answers.