The artificial intelligence revolution: from machines to thoughts.
Published on:Systems&Enterprise, October/November/December issue, theme: from the industrial revolution to the cognitive revolution
The revolution of Artificial Intelligence (AI), in particular the generative one, can be framed as a "fifth industrial revolution” or a “digital cognitive revolution”, and its effects are following some patterns similar to previous industrial revolutions, but with new dynamics. While the first revolution (from around 1750 to 1850) marked the transition from agriculture to industry (with the invention of the steam engine and mechanical weaving), putting the income of landowners and artisans in crisis, the second (from around 1870 to 1914) saw the advent of electricity, steel, mass production, Taylorism and Fordism, transforming manufacturing and transport, and the work of workers and engineers. The third revolution, that of information technology, automation and the internet, which occurred from 1970 to around 2000, radically changed the service, office and finance sectors, creating shocks in the world of office workers, technicians and bankers. The fourth, which started in 2000 and is still ongoing, has seen further transformations in digital, in the cloud, in big data, and has involved all sectors. But it is the fifth revolution that interests us now, that of AI. For the first time, it is having a massive impact on cognitive, creative and decision-making sectors, and on activities such as those of doctors, lawyers, advertisers, managers, highly cognitive and well-paid professions. Naturally, the transition between successive industrial revolutions was, and still is, gradual, with large areas of overlap: for many years, for specialized professional skills such as, for example, structural calculation in engineering or diagnostic imaging in radiology, there has been software capable of simulating the skills of excellent professionals, whose work is increasingly limited to the introduction of inputs and the validation of outputs.
Increased productivityat lower cost
Always, however, the effect of industrial revolutions has been to enormously increase labor productivity, initially especially in the primary and secondary sectors. This initially created major employment problems. As is known, the growing replacement of workers with machines led, at the beginning of the 19th century, to the violent Luddite revolts, so called for the first legendary (in a literal sense: there are no certain data on his existence) leader Ned Ludd, a worker who started the revolt as early as 1768, destroying, in a fit of rage, a mechanical loom that had taken away his job. Most economic historians agree that the protest, which was understandable (although difficult to justify due to its violence) in the short term, was not so in the medium and long term. If it is true, in fact, that the increase in individual productivity (quantity of product produced in an hour by a single worker) meant that the same quantity of production could be achieved by a smaller number of workers, it is also true that - precisely thanks to this increase in productivity - production costs were reduced enormously; consequently the demand for the finished product increased even more (the United Kingdom became the world leader in the production of fabrics), and to meet the growing demand the companies had to (re)hire many workers. This has happened for all productive sectors: in agriculture, for example, at the beginning of the 19th century the majority of the workforce was employed in producing just enough food for a world population of around one billion, while today a modest (in the most industrialized countries, very modest) percentage of workers produces food in abundance (even if, unfortunately, poorly distributed) for over eight billion people, while the majority of workers have moved to the secondary and, above all, tertiary sectors.
More generally, globally, per capita income is now (at constant values) approximately 10 times that of 1850.The average Italian is 15 times richer than he was in 1880. IWorld GDP is 250 times that before the first industrial revolution.
This increase in productivity, however, did not follow the same dynamics for all the industrial revolutions: in fact, if for the first three the impact was relatively rapid, but with (very approximately) linear growth, for the fourth and fifth, thanks to Moore's law, the growth (with exponential law: the computing power doubles every 18-24 months), in the first years, starting from very low values, was rapid, but not 'explosive' as it became in the period next (let us remember the well-known fairy tale about the inventor of chess). Even in 1987, Robert Solow (awarded the Nobel Prize for Economics in the same year) could quip: "We see the computer age everywhere, except in productivity data." However, the following decades changed everything: while the first three industrial revolutions were characterized by macro-discontinuitiesqualitative, the passage from the third to the fourth, and – even more – from this to the fifth,isin fact characterized by a macroscopic leap in scalequantitativein storage and calculation capabilities. To get an idea of how large this leap is, consider that, while a common snail has around 10 thousand neurons, a human being has around 100 billion: we could deduce that we are 'only' 10 million times more 'intelligent' than a snail, while from the first microprocessors to the current ones the multiplication factor is well over 1,000 million (and the growth does not seem to be stopping at the moment).
The importance of soft skills
Under this pressure we can believe that many of the intellectual activities that up until now have been almost exclusively 'human' will soon be automated, such as case law research, the automatic compilation and analysis of tax returns, advanced automatic translations; even in the world of writers/directors/advertisers we see more and more use of AI software. We talk about "human-machine hybridization”, in the sense that managers and analysts will have decision support from AI, software engineers will only check and assemble automatically written codes, and creatives will be able to limit themselves to providing crude (but as ingenious as possible) inputs to AI software.
Where previous revolutions automated physical force or calculation, now part of thinking is automated, and even specialized and well-paid intellectual work is potentially replaceable. Who has (had?) protection forstatus(qualifications, master's degrees, experience) can be supported or surpassed by AI systems. The value will then move toability to use AI(prompt engineering, algorithmic supervision), whilehuman soft skills will remain important: empathy, relationship, authentic creativity. Already today there is less and less demand for designers in engineering companies, despite the constant, or even increasing, use of project managers.
In a world driven by artificial intelligence, it is not enough to know how to use the tools:you need to understand how to interactwith them and, above all,what can (still) only be achieved by a human being. Lesoft skillswill be increasingly important to interact with people and machines, to verify AI output, to avoidbias, make decisions. Communication capacity will be more complex, as it will be necessary to interact together with human and AI teams.
Of course, you won't all need to be AI developers. Rather, it will be necessary to become professionals who know how to use it, evaluate it, guide it. The most resilient and useful profile in the next 20 years will likely be that of a “hybrid expert”: someone who combines deep knowledge of a domain (e.g. law, healthcare, communications) with a real ability to use and supervise AI.
What will be the employment impact of this (for now) latest industrial revolution? Until today, as seen, the increase inproductivityacted as an almost punctual counterpoint (at least on a global level) to the growth ofproduction, but can we believe (or, even more, hope) that this growth will continue indefinitely? Fortunately, the "immaterial" production sectors, at least in industrialized countries, seem to be growing more than the material ones, whose environmental effects (primarily climatic ones) are increasingly less bearable, but in any case - despite the (evidently insufficient) efforts made so far - these continue to increase. As noted by David Keeling, not only is the concentration of CO2 in the atmosphere increasing, but it continues to do so at increasing rates: if in the 1960s the annual increase was 0.7 ppm, currently the growth is over 2 ppm per year. And then, how "immaterial" are the production of "virtual" services or financial instruments actually, just to give some examples? Tourism is not only made up of hotel hospitality, but also air travel, and the Bitcoin network consumes between 90 and 164 billion kWh per year, equal to approximately0.5%of global electricity consumption.
If the increase inproductionswill no longer be able to fully compensate for increases in productivity if we have to move towards productive realitiesquantitativelystationary, even ifqualitativelyalways evolving, there is only one way to avoid this leading to dramatic increases in unemployment: reducing the time dedicated to work. A century ago Keynes, in his speech on "Economic possibilities for our grandchildren", had predicted, for our era, a substantial reduction in working hours (up to just 15 hours per week), thanks to increases in productivity, with the same production volumes. It was precisely the enormous growth in GDP per capita that thwarted this prediction, but not completely. In fact, it must be remembered that, over the last 150 years, in industrialized countries, and therefore also in Italy, there have actually been progressive decreases in working time, which was "unlimited at the beginning of the Industrial Revolution, saw the first limits to twelve and then ten hours a day by the end of the nineteenth century, the conquest of eight hours a day after 1917 (the year of the Russian Revolution), finally the introduction of the Saturday public holiday and the increase in paid holidays over the years From that moment on, despite the enormous increases in productivity generated by technological and organizational innovation, the trend of working hours also stops and even reverses throughout the West: since the 1980s, contractual hours have remained stable while those have actually increased, for example with the use of overtime; in recent years, contractual hours have also tended to increase considerably, due to the progressive postponement of pension requirements" (Craviolatti, 2014).
The centrality of educationin the development of critical thinking
Currently, in fact, unemployment levels are stable, or even decreasing, even if new employment appears increasingly concentrated in sectors with high precariousness and low profitability, not in danger of replacement (only at the moment, until - for example - drones replace riders, and robots carers). But even in the boom in investments in AI there are signs of a decline in employment: according to Indeed, in the US the demand for software developers more than doubled between 2020 and 2022, and then progressively reduced by around75%between 2022 and 2025. It seems that, progressively, exclusively 'human' professional spaces are reduced to relatively few positions of 'brilliant' leadership of research and technological development, increasingly closer (industrial research finds that approximately85%of the AI technological solutions created become industrial products within two years), while the mass of 'routine' developers is reduced precisely thanks to their replacement with AI applications.
The scale of investments in AI talent is certainly impressive, which has made some very bright young computer scientists into billionaires, distancing them enormously (from a remuneration point of view) from 'normal' researchers. Zuckemberg's decision to found the Meta Superintelligence Labs division, led by Alexandr Wang (a very young ex-CEO of Scale AI, which he created before even graduating) and Nat Friedman, to aim for the development of a new superintelligence, was recently evaluated positively on the stock market. The initial investment is in the order of tens of billions of dollars, but the number of new employees seems very small, at least in top positions: the initial team has around 50 elite researchers specialized in multimodal (text, voice, image, video).Meta has recruited at least 11 top researchers from OpenAI, DeepMind and Anthropic, with offers ofsigning bonusunprecedented up to $100 million to top talent.
All this places the institutions (governmentsfirst of all, but also schools and universities) in the face of enormous problems, and above all macroscopic risks such as that of a further increase in the already enormous social and economic gaps, of employment upheavals, of excessive delegation to AI (cognitive atrophy), of digital exclusion (those without access or skills will be cut off), and of ethical and legal ambiguity (who is responsible for the operation of a self-driving car, or - in a more or less dystopian, but possible, future of a 'robocop' on patrol in our cities?). To prepare for this future we increasingly need an education and research system capable of educating critical, creative and adaptive people, keeping pace with the transformations imposed byUseeChina, world leaders in these transformations: this is the enormous challenge to be faced immediately.