Unlock the Editor’s Digest for free
Roula Khalaf, editor of the FT, selects her favorite stories in this weekly newsletter.
The National Bureau of Economic Research has published a new paper from MIT’s superstar economist Daron Acemoglu, dismissing AI dreams as a productivity renaissance, accelerated growth and reduced inequality.
At this point it almost feels like heresy to say that AI won’t revolutionize everything. A year ago, economists at Goldman Sachs estimated that AI would increase annual global GDP by 7 percent in ten years — or nearly $7 trillion in dollar terms.
Since then, Goldman’s predictions have become almost sober, with even the IMF predicting that AI “has the potential to reshape the global economy.” FTAV’s personal favorite is ARK’s prediction that AI will help accelerate global GDP growth to 7 percent per year. 🕺
Professor Acemoglu – a likely future Nobel Prize winner – takes the other side. Alphaville’s highlights below:
I estimate that [total factor productivity] The effects of AI advances over the next decade will be modest – an upper bound that does not take into account the distinction between difficult and easy tasks would be an overall increase of about 0.66% within ten years, or about an increase of approximately 0.064% year-on-year. TFP growth. When the presence of demanding tasks among those who will be exposed to AI is recognized, this upper limit drops to approximately 0.53%. The GDP effects will be slightly larger because automation and task complementarity will also lead to larger investments. But my calculations suggest that the GDP boost over the next ten years should also be modest, on the order of 0.93% − 1.16% over a total ten-year periodprovided that the investment increase due to AI is modest, totaling between 1.4% and 1.56%, if there is a major investment boom.
As Acemoglu says, that’s “modest but still far from trivial.” But as he notes, we also need to take into account the fact that some of the most common AI use cases are bad – i.e. deepfakes etc.
Combating that can stimulate growth in the same way that rebuilding a hurricane-ravaged city stimulates growth, but it still detracts from overall prosperity. Below is the emphasis of Alphaville.
. . . When we take into account the possibility that new tasks generated by AI could be manipulative, the impact on welfare could be even smaller. Based on figures from Bursztyn et al. (2023), which relate to the negative effects of AI-powered social media, I provide an illustrative calculation for social media, digital advertising and IT defense attack spending. These could increase GDP by as much as 2%, but if we follow the figures from Bursztyn et al. (2023), their impact on prosperity could be −0.72%. This discussion suggests that it is important to consider the potential negative welfare implications of AI-generated new tasks and products.
Acemoglu is also skeptical that AI will have a major effect on inequality – and not significantly worsen or improve it. But overall, his work suggests that “women with low education may experience small declines in wages, that overall inequality between groups may increase slightly, and that the gap between capital and labor income is likely to widen further.”
The skepticism is interesting, since Acemoglu is one-third of an influential trio of MIT economists who lead the university’s heavily named Shaping The Future Of Work initiative.
The professor emphasizes that the potential of generative AI is great, but only if it is mainly used to give people better, more reliable information, instead of hallucination-sensitive chatbots and mechanically reconstructed images.
My assessment is that there are indeed much greater gains to be made with generative AI, which is a promising technology, but these gains will remain elusive unless there is a fundamental reorientation of the industry, including perhaps a major change in the architecture of the most advanced technologies. common generative AI models, such as the LLMs, to focus on reliable information that can increase the marginal productivity of different types of workers, rather than prioritizing the development of general human-like conversation tools. The generic nature of the current approach to generative AI may not be suitable for providing such reliable information.
Simply put, it remains an open question whether we need basic models (or today’s breed of LLMs) that can have human-like conversations and write Shakespearean sonnets if we want reliable information useful to educators, healthcare professionals, and electricians. , plumbers and other craftsmen.
Read further:
— The manicure economy (FTAV)
— Note on the investment prospects for the coming year or ChatGPT? Take the quiz (FTAV)
— Generative AI will be great for generative AI consultants (FTAV)