Will the AI revolution benefit everyone or will it massively destroy jobs? The question remains open according to Frédéric Cherbonnier, for whom artificial intelligence could well revolutionize our societies and help them in their quest for innovation.
Will our societies always be capable of innovating and benefiting from technological revolutions? This question has been pitting techno-pessimists against techno-optimists for years. The advent of artificial intelligence (AI) seems to be changing the terms of the debate and proving the latter right.
The main source of pessimism is the paradox posed by the development of information technology. A phrase attributed to Nobel Prize winner in economics Robert Solow has become famous: “You see computers everywhere but in the productivity statistics”. Specialists in economic growth such as R.-J. Gordon believed that we would never again experience technological revolutions on the scale of those of previous centuries.
In fact, it took almost thirty years for the American economy to reap significant productivity gains thanks to computers. As economist Erik Brynjolfsson and his colleagues have shown, one explanation for this delay is that companies could not really take advantage of this new technology without reorganizing themselves1.
It will also take more than thirty years to observe the impact of the electric battery developed at the end of the 19th century on economic growth2, and much longer to see the effects of the invention of the steam engine3 at the end of the 18th century. Considerable time may be needed to make technologies exploitable or to adapt to them.
On the other hand, it took less than five years for the development of large language models (LLMs) to lead to concrete applications (particularly with chatbots such as ChatGPT). Consulting firms such as Goldman Sachs are already predicting considerable productivity gains in the next ten years, while the IMF announced last month that this technology could lift Latin America out of economic stagnation.
The second source of pessimism is that it is becoming increasingly difficult to innovate. The productivity of researchers is said to be falling exponentially, by 50% every thirteen years! This is supported by an analysis of the productivity gains generated by American companies, and by specific case studies in the fields of electronics, agriculture and medicine4. For example, the number of clinical studies needed to achieve a certain degree of reduction in breast cancer mortality would have increased sixteen-fold in less than thirty years!
Here again, AI could be a game changer, as a recent experiment5 carried out in a materials science laboratory tends to show. An AI is trained on the structure and characteristics of existing materials. This tool is then entrusted to a group of researchers, who then see their ability to discover new structures and file patents increase on average by almost 40%! But the impact is very uneven, with the most experienced researchers seeing their productivity almost double while the others struggle to derive any significant benefit.
There is therefore reason to be optimistic: artificial intelligence could well revolutionize our societies and help them in their quest for innovation. But will this be to the benefit of all or, on the contrary, will it massively destroy jobs and only benefit a fraction of the highly qualified population? No one can say today which way the balance will tip, as the recent Nobel Prize winner in economics Daron Acemoglu points out. The question is open and the answer will depend heavily on the framework put in place by the public authorities: labor regulations, taxation, education system, vocational training, etc.
November 20, 2024 – published in Les Echos
Illustration: Photo Minh Pham, Unsplash
1Bresnahan, T. F., Brynjolfsson, E., & Hitt, L. M. (2002). Information technology, workplace organization, and the demand for skilled labor: Firm-level evidence. The quarterly journal of economics, 117(1), 339-376.
2Syverson, Chad. "Will history repeat itself? comments on" Is the information technology revolution over?"." International Productivity Monitor 25 (2013): 37.
3Crafts, Nicholas. "Steam as a general purpose technology: a growth accounting perspective." The Economic Journal 114.495 (2004): 338-351.
4Bloom, N., Jones, C. I., Van Reenen, J., & Webb, M. (2020). « Are ideas getting harder to find? ». American Economic Review, 110(4), 1104-1144.
5Toner-Rodgers, A. (2024). Artificial Intelligence, Scientific Discovery, and Product Innovation (in process of publication).
