The great potential of generative artificial intelligence (AI) as a transformative of productive processes and innovation has been highlighted for some time, but the enormous success of ChatGPT is accelerating this vision. In recent months, anything that smacks of artificial intelligence has attracted the attention of the media and, above all, of investors.
S-index
One reason that could support this view is the scant effect that all these expectations are having on productivity. Solow, a Nobel laureate in economics, said in the 1980s that computers were seen everywhere except in productivity statistics. Paraphrasing, we could now say that you see AI algorithms everywhere except in productivity statistics. Since 2005, the productivity of most countries has stagnated or even fallen, as in the case of Spain, Italy or Portugal.
Some authors consider that there are reasons to mistrust the potential of these new technologies on productivity. In the first place, Robert Gordon, one of the most influential economists in the United States on the subject of productivity, considers that the current revolution cannot be compared in influence with the previous industrial revolution, where the generalization of electricity, the combustion engine internal, chemicals, running water, antibiotics, etc. It was a radical transformation.
Secondly, various investigations show a very significant drop in research productivity, one of the main sources of growth for this indicator. For example, the number of researchers required to double the density of transistors on a chip every two years (Moore’s Law) is more than 18 times greater today than it was in 1970. The productivity of research in improving the results of agricultural production has been falling by 5% per year for years. And research to reduce mortality from cancer or heart attacks is also much less productive than the great advances that have occurred in the past.
Third, new patents and scientific articles have less disruptive and more incremental ideas. This happens in all sciences. One reason for this less disruption may be the aging of the population, as there are studies that indicate that the most disruptive research is produced by young people, who have a more “fluid” and less “crystallized” intelligence than the elderly.
However, there are reasons for hope. The first is the need for a certain amount of time for innovations to spread in the economy. It is true that productivity growth in the United States during the period 1920-1973, approximately 2%, fell spectacularly in the period 1974-1995, down to 0.5%. This would seem to prove the modern technoskeptics right. But between 1995 and 2005 there was a substantial recovery reaching almost 1.5%. One explanation could be that innovations take some time to have a significant impact on productivity. Those computers that were everywhere in the 1980s didn’t really change production processes substantially until the mid-1990s. This same delay could occur in the case of artificial intelligence. To realize productivity gains, AI has to spread through the economy, and that takes time. These advances require organizational change and human resources with adequate skills to take advantage of them, especially in the case of small and medium-sized companies.
However, in the case of AI, this process could be faster than in the transformations of the past. First of all, because it is about software. It is not necessary to acquire, transport, install, physical equipment as in the revolution of computers and telecommunications. Let’s remember that ChatGPT had 100 million users in just 2 months. Second, companies like Microsoft or Google are implementing generative AI solutions into their search engines and programs, rapidly expanding the number of users. Finally, the use of generative AI does not require being a specialist or programmer, as it works with natural language understandable by any human. A second reason for hope is the complexity of measuring the productivity generated by these activities. It could happen that the effects of generative AI are not directly seen in official GDP or productivity statistics due to problems in measuring inputs and outputs. This measurement difficulty is common in the valuation of intangibles.
Even admitting that the impact of generative AI on productivity may take some time to materialize and be reflected in the statistics, perhaps excessive expectations are being produced about its potential to transform economic activity. Time will tell.
