Only one out of four problems solved with AI will be economically justified in 10 years

The boom of artificial intelligence began, by the standards of human history, quite recently, and many experts already doubt how useful the corresponding technologies will be in their current form. One American scientist claims that over the next ten years, investments in various artificial intelligence technologies will pay off in only one out of four cases.

Image source: pikisuperstar / freepik.com

At the very least, as MIT professor Daron Acemoglu put it on a Goldman Sachs podcast, in economic terms, no more than a quarter of all applied tasks automated by AI will justify the investment in them over the next ten years. For the foreseeable future, “there will be very serious constraints on where the current large language model (LLM) architecture will take us.”

At the same time, according to the professor, the speed of improvement of artificial intelligence systems will not be determined solely by the pace of increasing computing power through the installation of more accelerators in data centers. Requirements for data quality will increase, and it is still difficult to understand how this quality will be improved.

Goldman Sachs experts, based on these arguments, make a forecast according to which AI will affect only 5% of all economic activity in the United States in the next ten years, labor productivity will ultimately increase by only 0.5%, and GDP growth due to this factor will not exceed 0.9 % based on the results of the next ten years. Such an economic effect can hardly be called convincing, and such a forecast will certainly make investors think about more rational spending of funds on the development of artificial intelligence technologies.