Summarize this content to 2000 words in 6 paragraphs in Arabic Stay informed with free updatesSimply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.I’ll admit to tuning out much of the debate over whether artificial intelligence is going to destroy us all. If our digital overlords do eventually turn me into a paper clip, then at least I will have enjoyed my final precious moments as a human. I’ll have spent them considering a different part of the debate, over how much AI will affect growth. The stakes are slightly lower, but there is just as much disagreement. Why?The core discussion concerns AI’s scope, scale and speed. Will AI be a force that accelerates automation, or will it also speed up innovation? And will its effects be the avocado slicer of food prep, or the microwave? And then there is the risk that while technologists may like to move fast and break things, corporate executives prefer a more sedentary lifestyle.There have been several attempts to estimate the effects of generative AI on annual productivity growth, with pretty varied results. Last year, Goldman Sachs estimated that in rich countries it could contribute around 1.5 percentage points over a decade. Soon after that, McKinsey predicted that it could deliver between 0.1 and 0.6 percentage points between 2023 and 2040. And most recently Daron Acemoglu of MIT calculated a boost over the next decade of at most 0.2 percentage points.The gaps between these figures mostly relate to differences over speed and scale. Each tries to estimate how much existing work will be affected by generative AI, as well as the potential cost savings. Acemoglu, for example, suggests that over the next decade around 5 per cent of tasks will be profitably replaced or augmented by AI. (I’ll argue that my editors should hang on to me otherwise the columns might become too funny.) Even then, the average cost savings across these tasks might only be around 15 per cent — or lower if AI struggles to replace harder ones where decisions require lots of context or lack objective measures of success. (I hear column writing is very hard.)McKinsey says it is clear-eyed about the pace of diffusion, drawing on historical evidence that technologies take up to 27 years to reach a plateau in adoption after becoming commercially available. But it seems to be more bullish than Acemoglu about the potential for tasks to be automated. In a separate report McKinsey estimates that in the US, generative AI could account for 8 per cent of work hours being automated by 2030.The analysts at Goldman Sachs also reckon that quite a large share of work will be affected by AI. But the bigger difference is over timing. They cite the electric motor and personal computing as breakthroughs leading to US labour productivity booms of around 1.5 percentage points per year over a decade. Awkwardly, these took 20 years to start. In other words, the boom they predict is over “a decade”, not the one starting now.In a more recent note the Goldman Sachs analysts cite surveys suggesting that fewer than one in 20 companies report the “use of generative AI in regular production”. And they confirm that most of the boost to global GDP will come after 2030.Questions over speed and scale are important. But perhaps the bigger question is over AI’s scope. Tyler Cowen of George Mason University recently criticised Acemoglu’s paper for assuming away the possibility that AI would do new tasks or produce new things — just look at the chatbots impersonating Shakespeare or Elon Musk. Acemoglu’s argument is that industry’s focus is elsewhere, for example on digital advertisements.There could be bigger benefits in store. Over decades the world has ploughed an increasing share of resources into innovation, with diminishing returns. A study published in 2020 found that research productivity for the US economy had fallen by a factor of 41 since the 1930s. Optimists suggest that AI could increase those returns and speed up 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 suggests that the effects on productivity could be even greater than the most optimistic of those earlier automation-based estimates.“Some uncertainty is of course healthy,” says Acemoglu of the change brought on by AI, since “we are at the very very beginning of it”. Which means plenty of other important questions to ponder, including how the spoils of any growth are shared. As well as those, perhaps I’ll allow myself to wonder whether one day there will be an AI so powerful that it can turn paper clips back into [email protected] Soumaya Keynes with myFT and on X
