All of this was analyzed in a GSR study by the Global Economy team led by Joseph Briggs who explains that so far there has not been a great impact on productivity growth, but the reason is that although there is still great potential for AI to automate many of the daily tasks of workers, thus saving a lot of time and generating large productivity gains, adoption rates are quite limited at the moment. “The key step, of course, in task automation, is that people have to start using it,” explains the GS economist with a past at the Fed.
According to Briggs, some of the academic literature and economic studies that have analyzed the increase in productivity after the adoption of AI support our view that large productivity gains are possiblesince the average increase in productivity is around 25%.
“Case studies of companies that have adopted AI involve equally large efficiency gains. So there are many reasons to be optimistic. “It’s just going to take a little longer to see these productivity gains come to fruition,” he maintains.
That’s why Briggs reaffirms its forecasts since they do not imply any boost in AI before 2027 and the changes observed in the last year are consistent with the moderate growth scenario in the first three years of its adoption and implementation.
Now, why the slow adoption despite the heavy investment in this technology? For AI to be widely implemented, explains Briggs, many things have to happen, which, for the moment, they are not happening.
First of all, he points out, we must have models that are powerful and trained enough so that they can be useful in daily work. In addition, you have to have the ability to facilitate and answer all the questions that people are going to ask AI models, when they use them daily several times a day in their normal work.
“Both of those things require a large increase in investment in semiconductors, which in turn requires a large increase in investment in network capacity. And ultimately, that is going to require an increase in investment in electricity and energy to support the increase in demand that facilitating consultations will require,” explains Briggs.
In any case, the expert highlights that clear signs are being detected that investment on this front is increasing: the income of semiconductor manufacturers has increased by around 50% since the beginning of 2023 and the revisions to supplier forecasts of AI hardware implies an increase of about $250 billion from a year ago. “So there are many signs that the investment that lays the foundation for the future use of AI is occurring,” he notes.
“Adoption and usage will happen when these pieces are in place and companies start to actually use AI on a daily basis. For the most part, that hasn’t happened yet. We see about 5% of companies saying they use “Generative AI in its usual production, but it is a fairly small proportion relative to the total number of companies that we believe will end up benefiting,” adds Briggs.
This percentage of companies includes companies from the information, finance or insurance sectors, as well as the audiovisual industry. Among the functions for which generative AI is used, Briggs points out that marketing, chatbots or data analysis. “It’s kind of the low-hanging fruit where AI is most applicable, at least in its current form. Ultimately, we think generative AI will automate a broader set of tasks. But that will likely require creating of an application layer to support the broader automation that we see possible,” he explains.
As for the remaining 95%, GSR notes that many executives see the potential of this technology, but also important barriers such as lack of knowledge, concerns about privacy and security or fear of investing too much at too early a stage in its development. This appears to reflect that companies want to ensure they get generative AI right and are therefore taking a deliberate approach to its adoption, according to the GS expert.
“These views are broadly consistent with what we’ve seen in some of the business surveys, where CEOs are asked about their intention to use generative AI. Very few say they expect it to have a significant impact on their business in the next one to three years.. “Most say they expect to see a significant impact over a three- to 10-year horizon,” Briggs said.
It should be noted that since adoption has been very low, it is not surprising that no major repercussions have been seen in the labor market. In fact, the unemployment rate between occupations that are highly exposed to AI automation and those that are less exposed has not changed much.
“It is very possible, and even probable, that the net impact on the labor market has been positive so far. This is in line with our long-term expectations, where we hope that generative AI will not lead to large job losses. Overall, we believe it will create opportunities in sectors or occupations adjacent to AI or in sectors where the workforce has a comparative advantage,” he concludes.
