• Home
  • Artificial intelligence, automation and labor

Artificial intelligence, automation and labor

This post is also available in: Italiano (Italian)

The New York Times of February 26, 1928, published on eight columns, at p.129, the famous “March of the Machine Makes Idle Hands” article, with the subtitle “Prevalence of Unemployment with Greatly Increased Industrial Output Points to the Influence of Labor-Saving Device as an Underlying Cause.”
More recently, the Economist (2016) has resumed that title by opening a series of eight articles: the title of the first article, “March of the Machines,” is preceded by the specification “Artificial Intelligence.” Two words that well clarify the difference between the situation in 1928 and in our days.
The second article in the series, entitled “The return of the machinery question,” echoes the title of Chapter 31 of “On Machinery” by Ricardo (1821) and also that of Marx’s “The Fragment on Machines”.

Machines, labor, and artificial intelligence

Quoting Ricardo (1821), we can easily underline the key point:

(rif.31.25) The statements which I have made will not, I hope, lead to the inference that machinery should not be encouraged. To elucidate the principle, I have been supposing, that improved machinery is suddenly discovered, and extensively used; but the truth is, that these discoveries are gradual, and rather operate in determining the employment of the capital which is saved and accumulated, than in diverting capital from its actual employment.

In the perspective of labor, the problem lies in (i) the extreme acceleration of the revolution in production and (ii) in the quality of change, now with the intelligence of the machines.

Quoting again the Economist series, “After many false starts, artificial intelligence has taken off.”

We were close to this revolution, but we were waiting for more powerful and less expensive computers to have the possibility of a paradigm shift: switching from teaching to the machine what to do, to teach it how to learn! In technical terms, the transition to machine learning and especially to that part of automatic learning based on artificial neural networks, with the name of deep learning.

For the specialist in the field, this is not true artificial intelligence; maybe they are right, but the question goes beyond the subject of this note. New learning machines can replace workers, especially because they can deal with very complex problems, anyway showing intelligence.
To get a measure, though very rough, of the phenomenon, see Figures 1 and 2. In the first, we find the installations of industrial robots in the three-year period 2013-2015. The car is overcoming the crisis by installing over 250,000 robots, a huge figure.

Figure 2 gives us a projection of the phenomenon until 2019. Considering the three areas, in the final year, about 400 thousand industrial robots will be installed, mainly in Asia.

Observing the pace of the change

The development of production worldwide has considerably limited the perception of the acceleration of the ongoing change. This consideration has been true mainly until the arrival of the crisis, in 2008.

In Brynjolfsson and McAfee (2014), respectively director and co-director of the MIT Initiative on Digital Economy (http://ide.mit.edu), we find material of great interest on the ongoing change. The problem of concentration of wealth in a few hands is not new: it is now at a growing stage, but it is not negative in itself until even the less well-off people perceive the improvement of their condition. Instead, when people who want full-time jobs find it only part-time or do not find it at all, it is confirmed that while the benefits of new technologies are real, they are not enough to offset the growing gap between personal situations. A trend that is only partially due to the recession and that, above all, seems destined to be a not transient phenomenon.

The accelerated process can so produce unexpected effects.

In a changing society

So, the machines in place of the workers, in new and surprising fields. And the job?

Do we need to ask: are we inevitably condemned to work?

Everything is going to change, but the phases where change accelerates are the hardest for people. If the machines replace an ungrateful and tedious job, it is certainly a good thing. The replaced person, however, loses the job; he/she can find a better one. But if at the same time there are many persons who lose their jobs and very few who find a new one, they need reliable forms of social protection. If the framework is that there will always be less work, we need a complete rethinking of the organization of the society, making the transition periods less traumatic.

What to do?

Taxing robots, as proposed by Microsoft founder Bill Gates (various news reports at the beginning of 2017 attribute this proposal to him) is a diminutive way to deal with the problem, perhaps a temporary solution, certainly not a structural choice.
For an economist, taxing robots is equivalent to taxing capital, which is entirely legitimate, but we are now in a perspective of a true Copernican revolution. We need to imagine something of completely different.

Psychologists are horrified when an economist states that the work will be for a few and at a very different pace from the current ones. Work as a source of social relationships and personal satisfaction is deeply interlaced to the design of life that almost everyone considers positive and natural. I do no doubt the need for social relations, but do we need to work for that?

An extreme perspective (but, maybe, not so extreme)

How will I have an income if I do not work, but will I still have to buy the necessary goods? At the center of the answer, we have another question: who will produce the necessary goods? If the robots will produce nearly anything and robots will also produce new robots, who will be their owner and so the owner of the resulting goods? It is now challenging to imagine this transformation, and we see in perspective an infinite sequence of obstacles generated by the various transition phases.

Science must be aware that it will not be possible to evade that problem, trying to reconcile the tensions with remedies inspired by welfare, acting on the income of the citizenship. It is necessary to create new foundations to regulate the participation in collective life, given that most of the work will be done by machines and computers. And it will not be easy to decide who will have to give them the orders.

The central question is exactly the last one: to understand who will give the orders the robots (and who will be their owner, which is a non-irrelevant corollary). If you deal with this point, everything else becomes secondary. If robots produce robots and they are collective property, the goods and the services produced in that way will be extraordinarily abounding. The prices will tend to disappear, the money will no longer be necessary. Eliminating the money and the accounts, many other jobs, possibly survived to robots, will no longer have reason to exist.

To imagine the world without money can seem close to extravagance or madness: it is instead the design of a new society that has overcome both the scarcity and the related conflicts, and it is more protective and respectful of people.

Objection: without prices and without the profits of the distribution and production network, how to determine what to produce and for whom? The enormous difficulty of economic planning has led to the collapse of the Soviet Union. The computing systems and the data available were inadequate and paradoxically it was easier to plan Sputnik’s orbit than to calculate how many socks to produce for every area in an immense country. Now, with super-sophisticated computing facilities and with big-data, Amazon and its (few) competitors know how to continuously restock the decentralized warehouses, minimizing stocks, but assuring deliveries mostly in twenty-four hours.

The full change can require twenty or fifty years or so; the apparently negative effect of the intelligent machines on society, with over-productions and missing work places, is manifesting its effects now, in the early part of the 21st century. Temporary solutions are related to income taxation formulas for people under a certain level of income, but we cannot handle this transition if we do not have clear in mind the long-term consequences pointed out here.


Brynjolfsson, E. and McAfee, A. (2014). The second machine age: Work, progress, and prosperity in a time of brilliant technologies. New York: W.W.Norton & Co.

Ricardo, D. (1821). On the Principles of Political Economy and Taxation. London: John Murray, third edition (first edition: 1817). Online at http://www.econlib.org/library/Ricardo/ricP.html

Krugman, P. (2017). Maid In America. Online at https://krugman.blogs.nytimes.com/2017/02/24/maid-in-america/