Yuval Harari:

"21 Lessons for the 21st Century" (2018)

(Copyright © 2019 Piero Scaruffi | Terms of use )
This is the worst of Harari's books. It mostly contains trivial stereotypes that have become very popular thanks to (really bad) books on A.I. and (really bad) Hollywood movies. I was pleased to read the chapter on "fake news", which is very similar to what i wrote when the debate on "fake news" started in 2016. See Fake News is the Hallmark of Human Civilization . But i rolled my eyes when i read his opinions on the "useless class" that, according to him (an to other popular books) will be created by A.I. If A.I. is such a bad idea, why not ban it altogether? Because it is not what these popular books think it is.

Harari believes that mass industrialization created the working class and the A.I. revolution will create a new "useless" class. This kind of writers, conveniently, never define what they mean by A.I.: if my $5 watch (that can do something that no human can do) a form of A.I.? If my dishwasher a form of A.I.? What exactly is this A.I. revolution? Just that a computer beat a world champion at the game of go? And when is this catastrophic unemployment going to happen? It's been predicted for so long. By now all truck drivers should be unemployed due to the widespread use of self-driving vehicles: have you seen any self-driving vehicle in your neighborhood this morning? Just one?

Every generation since the invention of the computer (and perhaps even before the computer) has been terrified by the prospect that these "intelligent" machines would make humans obsolete. Instead each step in automation has created better jobs. Technology has taken us from the caves of the stone age to today's comfortable air-conditioned homes and high-rise office buildings, with splendid plumbing. Mass unemployment has never happened. Unemployment in fact is a chronic problem in countries where automation has lagged behind. Greece has the highest unemployment in Europe, not Germany. Japan and the USA have very low unemployment, and high average salaries. Africa and Latin America have high unemployment and low salaries. That is not to say that automation doesn't cause social problems, especially to the hyper-specialists of the past (hopefully being replaced by younger generations that are much better at reinventing themselves as technology changes). Automation creates new jobs, and often jobs that nobody predicted. When the computer was invented who predicted that it would create a whole new industry of smartphones, ecommerce, shared rides and so on? Of course millions of traffic cops lost their jobs when traffic lights were deployed at every intersection, and typists lost their job when wordprocessors replaced the typewriter, and film-developing shops disappeared when the digital camera was invented, and booksellers lost their business when websites started selling books. Some of them certainly had trouble finding new jobs. But their children now probably have better jobs, jobs that didn't exist back then. Humans simply take the new technologies and invent new jobs that use these new technologies. Then humans automate these new jobs with new technologies and the cycle continues. Algorithms can probably do everything that we can do today, but not what we will do tomorrow because what we will do tomorrow is precisely whatever it is that algorithms don't do today.

Actually, algorithms can't do most of the things that Harari thinks they can already do. Maybe some day they will become more intelligent, but right now Harari wildly overestimates their intelligence and wildly underestimates their stupidity. There is a reason if you haven't seen any self-driving car today during your commute.

He also takes for granted wildly exaggerated claims that have been published in popular magazines, claims that no serious A.I. researcher would make. Machines that can easily and accurately recognize your emotion from a video of you? They are as real as flying elephants. The day there is a flying elephant, we can discuss the socioeconomic effect of flying elephants. Until that day, we are just speculating on something that may never happen. We are taking science fiction and discussing it as if it were science.

I have written about this paranoia in my book "Intelligence is not Artificial". My personal prediction was (in the middle of the Great Recession, while unemployment was skyrocketing), and still is, that all this software (call it A.I. or simply call it "software") will enable just the opposite: hyper-employment, i.e. negative unemployment. See the chapter Jobs in the Age of the Robot - Part 6: Hyper-employment of my book (in the 2019 edition). The most important question in 21st-century economics may well be: What should we do with all the superfluous people, once we have highly intelligent non-conscious algorithms that can do almost everything better than humans?

Unfortunately, books like this one sell a lot more copies than the books that advise people (especially young people) on what to study in order to take advantage of the new automation. People who read books like this one will probably never develop the skills to survive and prosper in the world of tomorrow. Hence this kind of books are a sort of self-fulfilling prophecy: this book's readers are indeed likely to be the "useless class" of the future.

This book reveals Harari's staggering ignorance of economics, technology and even art (his chapter on machine art is funny).

Read his previous books and skip this one.

In April 2019 Harari was invited to speak at Stanford and emphasized two points that were already mentioned in this book. The first one is that: algorithms know us better than we know ourselves. Harari correctly points out that it doesn't take much because we are so clueless about who we really are that even a bad algorithm can still do better than we do at analyzing ourselves. But in my opinion what algorithms do is simply to make us buy things we don't need, something that radio and TV commercials, and ads in newspapers and magazines, have always done. The science is not all that different, if only Harari studied it. In fact, the staggering difference is how stupid the algorithms can be compared with a traditional ad in a traditional magazine: make the mistake of browsing "flights to New Zealand" and for days and weeks you will see ads about flying to New Zealand. Is this the algorithm that is supposed to know me better than i myself do?

The second point that Harari emphasized is that A.I. could be the technology of world domination in the 21st century, the modern equivalent of nuclear weapons. This too seems far-fetched. First of all, the difference is that nuclear weapons really exist and have been demonstrated, whereas A.I. is far from being "intelligent". Second, Harari is always and only talking about one specific kind of A.I., deep learning, the kind that is based on the combination of big data and powerful computers. That indeed is difficult for most countries to achieve (because they have neither the data nor the supercomputers), but i find it difficult to believe that this glorified version of statistics will grant the USA and China the kind of superpower status that came with thousands of nuclear weapons.

I still think that most of what Harari and others discuss has little to do with the tool (A.I. and automation in general) and a lot to do with the "vast algorithmic bureaucracies" that we have created all around us. See