How to get big numbers when predicting the impact of AI on development.

WhatsApp Group Join Now
Telegram Group Join Now
Instagram Group Join Now

Stay informed with free updates

I will admit to ending the debate about whether artificial intelligence will destroy us all. If our digital overlords eventually turned me into a paper clip, at least I would enjoy my last precious moments as a human being. I've spent them considering a different part of the debate, that of how much AI will affect growth. The stakes are a little lower, but there is just as much disagreement. Why?

The main debate is about the scope, scale and speed of AI. Will AI be a force that accelerates automation, or will it also accelerate innovation? And will the effects be food prep's avocado slicer or the microwave? And then there's the danger that while technologists may like to move fast and break things, corporate executives prefer a more sedentary lifestyle.

Several attempts have been made to estimate the impact of generative AI on annual productivity growth, with widely varying results. Last year, Goldman Sachs estimated that in rich countries it could contribute about 1.5 percentage points over a decade.

Shortly thereafter, McKinsey predicted that it could deliver between 0.1 and 0.6 percentage points between 2023 and 2040. And recently MIT's Daron Acemoglu calculated an increase of at most 0.2 percentage points over the next decade.

The difference between these figures is mostly related to differences in speed and scale. Each tries to estimate how much existing work will be affected by generative AI, as well as potential cost savings.

For example, Acemoglu suggests that about 5 percent of jobs will be profitably replaced or augmented by AI in the next decade. (I would argue that my editors have to stick with me or the columns can become very funny.) Even so, the average cost savings on these jobs can only be around 15 percent — or less. If AI struggles to adapt to difficulty where decisions require a lot of context or lack objective measures of success. (I've heard that column writing is very difficult.)

McKinsey says it is clear about the speed of diffusion, based on historical evidence that it takes 27 years for technologies to reach a plateau in adoption after they become commercially available. But Acemoglu seems to be more bullish about the potential for tasks to be automated. In a separate report, McKinsey estimates that 8 percent of work hours in the U.S. could be automated by generative AI by 2030.

Goldman Sachs analysts also believe that a significant portion of work will be affected by AI. But the big difference is over time. They cite the electric motor and personal computing as breakthroughs that led to an increase in American labor productivity of about 1.5 percentage points per year over a decade. Oddly enough, it took them 20 years to get started. In other words, the boom they predict is over “a decade,” not starting now.

In an updated note, Goldman Sachs analysts cited the survey as saying that fewer than one in 20 companies report “using generative AI in regular production.” And they confirm that most of the growth in global GDP will come after 2030.

Questions of speed and scale are important. But perhaps the bigger question is more than the scope of AI. Tyler Cowen of George Mason University recently criticized Acemoglu's paper for dismissing the possibility that AI will perform new tasks or create new things — just chatbots impersonating Shakespeare or Elon Musk. See Acemoglu argues that the industry's focus is elsewhere, for example on digital advertising.

The store can have great benefits. Over the decades the world has invested an increasing share of resources in innovation, with diminishing returns. A study published in 2020 found that research productivity for the US economy has fallen by a factor of 41 since the 1930s.

Optimists suggest that AI can increase those returns and accelerate the rate at which we discover new ideas. Just this week, Google DeepMind unveiled an AI model that could help researchers find new drugs. Ben Jones of Northwestern University points out that the impact on productivity may be even greater than the most optimistic of these earlier automation-based estimates.

“Some uncertainty is certainly healthy,” Acemoglu says of the change brought about by AI, “because we're just at the beginning of it.” Which means there are many other important questions to consider. are, including how the losses of any progress are shared. Along with these, perhaps I'll allow myself to wonder if an AI will one day be powerful enough to turn paperclips into humans again. give

Follow up with Somaya Kanes. myFT And on x

WhatsApp Group Join Now
Telegram Group Join Now
Instagram Group Join Now

Leave a Comment