Warning! Red Alert! At the end of each chapter this book has summaries of the main points and questions, and in particular questions about how the content of the chapter may relate to the goings on at “your employer.” It is clearly meant to be packaged into “continued education” courses offered by the authors’ employer (the Sloane School of Business at MIT) to middle aged (hence the enormous print) middle managers on a two-week jolly / off-site / executive MBA. This, sadly, I discovered only after the book had come through the mail.
I decided to read it anyway, because the authors of “Machine Platform Crowd” are who I thank for waking me up to the fact that we are living through a proper, bona-fide depression. Just like the depression that the discovery of the tractor and phosphates caused in the 1930’s, by dint of putting out of work, permanently, the 25% of the US workforce who used to work in agriculture, McAfee and Brynjolfsson argued six year ago in their 76 page monograph “Race Against the Machine” that we are living a new depression brought about by a “Second Machine Age.” (which is the title of their second, less sombre, book, incidentally) The victims this time round are educated people who do repetitive work that once upon a time (but not any more!) could only be performed by an educated human. Oh, and we’re only getting started.
Sure, blame the deficits, blame the banks, blame the ninja loans, blame the Fed, blame who you like, but only for trying to solve this intractable problem in silly ways that won’t work and thereby for making things even worse. The deeper cause of the depression, however, is that the educated class has educated itself to do repetitive tasks that are increasingly done by machines. Oh, and the 1% of guys who can “run with the machines,” yeah, they are going to be the bad-guy 1 percenters. Ooops.
This was a good way for the authors to become famous (and earn my respect forever) but not a terribly good way to get rich themselves! “Dismal” does not sell. Ask that Malthus fellow.
Boy, have they changed tack since 2011.
“Machine Platform Crowd” is to the machine age what Candide and Ingenu were to the Enlightenment, but with no trace of irony, sarcasm or, indeed, doubt about what promises to be the best of possible worlds.
Let’s start with us humans. We suck at judgement, it turns out. Our Kahneman “System 1” is trash and an algorithm will always beat it. But wait. Not all is lost! That info is solid gold for those of us who are smart enough to understand that we have been doing things backwards. All we need to do is let our “System 1” free to come up with whatever garbage a billion of years of evolution have taught it to do, and then feed it all as input to the algorithm. Job done. Conversely, don’t dare take the results of the algorithm and judge them with your experience. You’ll mess it all up.
Don’t meet the client. Trust that FICO score. Call up the guys at FICO and ask politely if their computer will talk to your loan officer. It may well be able to process his otherwise useless experience into an even better FICO score. (You’ve guessed right, it does not say this in the book, that’s me extrapolating, sorry)
Next let’s look at the machines themselves. In the words of ZZ Top, they come in two classes: First, the kind that we program with some sensible ideas we have (example “adjectives in English absolutely have to be in this order: opinion-size-age-shape-color-origin-material-purpose-Noun” p.70) and then have the machine implement much faster than we ever could. Second, the kind that don’t need us at all and look at reams of data and come back with some answer and we just tell them how well they guessed. (That’s called “supervised learning” for those in the back who are not paying attention)
The first kind has been beating idiots like me at chess since I can remember. The second kind (steered, but not quite, by a fellow Greek, let’s hear it for Demis!) did something much more awesome the other day, it beat the world champion in Go, which is apparently much harder because there are fewer atoms in the known universe (and any undiscovered parallel universes) than there are ways to lay your pieces on a Go board. Maybe. Approximately. Where’s Val Kilmer when I need him?
The second kind of machine does not know what it knows or how it learned it, it embodies the Polanyi paradox. What’s new and amazing, basically, is a machine that, like me, could not tell you why it played the move, but very much unlike me, plays the correct move! All from having played many many many more times than I ever have, mostly with itself. (no, honestly, that’s what it does) And it’s finally possible today thanks to (i) Moore’s Law (ii) Amazon Web Services and (iii) the availability of “big data” to train itself on.
The future is in machines that don’t know what they know, just like us, basically. Instead of asking for the “Highest Paid Person’s Opinion,” also known as the HiPPO, ask this machine. It may not know what a liver transplant is, but if it thinks you need one, stop reading and call Alvin Roth now. (Kidney, liver, it’s all the same, don’t heckle! Instead, pray that it’s as well trained in recommending rare medical procedures as Demis’ Deep Mind is at playing Go with itself. I’m sure it will be fine. Better than the highest paid doctor’s opinion, you can be sure of that.)
The final frontier is that machines are starting to do creative work. Exactly how they play Go without knowing why they played the move, merely that it will probably work, they can design a radiator that looks like no radiator you’ve seen before, rig up a very effective chassis for your racing car, or compose a piece of music that you are likely to enjoy. Amazingly, after all the mumbo jumbo of the first 100 pages in the book, I actually found this (fifth!) chapter to be totally fascinating and convincing.
From “Machine” the book moves on to “Platform,” which actually is a tremendous section. (There are two authors at work here, basically, and this is the one I prefer)
The authors start by explaining that Apple makes 110% of all money made in phones today because 1. it controls the “platform” for all the apps and 2. nobody is counting the money Google is making off of Android. Jokes aside, they make the very strong point that making devices comes down to competing on costs, whereas establishing a platform allows you to reap the positive externalities generated by every single new participant in your platform. The virtuous circle works like this:
• every app that runs on an iPhone creates demand for the iPhone itself
• if many people own an iPhone, this creates a bigger supply of good apps
So 700 dollars is a lot of money for a telephone, but not if it can run 700 apps you’d pay a dollar each for, and things work out such that it will. And Apple no longer has to find its competitive edge in squeezing down the costs of making its beautiful devices.
Next come up the “two-sided” platforms, which are all the craze. I do not remember my days of starting book2eat.com that fondly (a good 40 VC’s were aghast at the fact that I was trying to sign up both restaurants and punters at the same time, and another 200 never answered my calls or emails) but the point is that if you manage to fill more seats the word gets out and you sign up more restaurants and if you’ve signed up more restaurants punters are likelier to make a booking on your platform and this creates a dynamic that becomes irresistible. So much so, that for example a company called ClassPass (that filled exercise studios with punters who were happy to be told at the last minute where to take their yoga mat to) had to revise its pricing policy (p. 178) due to the unmitigated success of the business model in filling exercise classes!
Next up is an analysis of Uber. Here I finally understood something that had always been a mystery to me: Uber’s obsession with low price is so huge because that’s where the demand curve is most elastic: All the way to the right, the demand curve is almost horizontal. People switch over from the train to Uber at some point, basically, and then all your cabs are full all of the time. The chart is on p. 213 and I totally buy it. Congestion in that city would be awful, but that’s down to the choice of the taxi example, of course. The point about why double-sided platforms love a low price is very well made.
The way ratings have solved the asymmetric information problem is discussed here too, with the inevitable name-dropping of the authors’ colleague down the hall who got the Nobel Prize for first posing it. Oh, and don’t worry, Airbnb will not kill the hotels industry, because hotels cater to business travelers. Don’t forget, it’s all good! The fate of the cabbies and the longshoremen is left to the reader to ponder, but they can all go work for Travis, no?
The third, final, section of the book is about “the crowd”
It starts by explaining that crowdsourced software rocks, basically, and that the ingredients you need are:
1. Openness: who knows what motivates them, but people do contribute
2. Noncredentialism: good software can be written by absolutely anyone
3. Verifiable and reversible contributions: unlike a novel, you can test software quickly and undo any poor contributions
4. Clear outcomes: via a good license you can assure coders that their work will remain public
5. Self-organization: people do best the work that they WANT to do and if that leads to a fork in the code, so be it
6. Geeky leadership: it’s the kind coders will write code for
Hayek and Polanyi make another guest appeareance in the chapter, but it works fine without them.
A further important point about the crowd is that very often the contribution to a project that makes the big difference does not come from an expert in the field. Indeed, the knowledge that will cut the Gordian knot may come from somebody with knowledge that is “far away” from the subject, i.e. somebody who will be surplus to requirements 99% of the time and thus a team of experts could never afford to have on board: “The expert you know is not the expert you need,” basically. He’s somewhere in the crowd and that’s where you must reach for him.
From there we jump to a painfully banal chapter on Bitcoin that addresses the points about cryptocurrency I could not care less about (like the mechanics of how it works, described just inaccurately enough for you not to be able to actually work it out, or the effete generalization that blockchain may be more important than Bitcoin,) barely gets into the aspects that matter (is it money? How does it provide store of value if there can be a million different otherwise scarce and deflationary varieties of Bitcoin? How will it compete if nobody will accept to be paid tax in it?) and complains that the Chinese are now dominating it.
By that point, we find ourselves in a final chapter that’s nothing to do with Machine, Platform or Crowd, but is there to make the customers feel good who are attending the seminar and assures them that companies will still carry on being necessary in the future, because they are who possesses “the residual rights to control: the right to make decisions not specified in contracts.”
Given that, presumably for the benefit of the same attendees, there is a need to sketch (very well, I must say, see pp 154-156) how the demand curve and the supply curve work and what happens when they intersect, one can safely say that this short paean to the firm is guaranteed to go over their head. But it must make for a warm conclusion. And now, back to work!