Economist and Institute Professor at Massachusetts Institute of Technology (MIT), winner of the 2024 Nobel Prize in Economic Sciences
According to some studies, although the impact of generative ai is uncertain and conditional on technological advances, this innovation alone could increase us labour productivity by almost 1.5 percentage points per year over a 10-year period after widespread adoption. what do you think of those estimates?
The paper I published in 2024, “The simple macroeconomics of AI”, suggests that total factor productivity, i.e. the sort of efficiency- and technology-related component of economic growth, should increase by about 0.5% to 0.6% over 10 years, so about 0.05% growth every year additionally thanks to AI.
This is much smaller than some of the existing estimates.
There is a very risk that the sector will be dominated by a limited number of big tech companies.
I believe the current direction of AI does not just miss out on great productivity gains, but it also increases the dominance of large corporations, multiplies inequality and creates a variety of social ills.
There is a lot of potential for AI to improve human productivity, but this would require a new architecture of AI based on domain expertise and reliability, which seems to be missing from current AI chatbots and large language models.
As a result, more beneficial outcomes may require new institutions, policies and regulations.
How do you see the role of competition law with respect to ai?
There is a very real risk that the sector will be dominated by a limited number of big tech companies. Current market conditions make it easy for incumbents to dominate because they have all the cash (and can acquire or bury competitors), all the data and huge existing networks of customers.
I believe antitrust has a role to play to allow alternative firms to embrace a vision of “pro-human AI”, i.e. opening up space for entry by new business models that will have a more beneficial impact in the labour market.
What is going to be the future of generative ai?
I think data is going to become an even more impo!tant factor of production than land. It is imperative that data creators should be properly compensated, which would in turn lead to a more equitable distribution of gains and also to higher quality data. This means there is a need to protect both the privacy and the data rights of data creators.
We need to find clever market solutions so that, once payments have been made, the data can be shared with smaller players as well as big players. It would be the industry’s own cost if high quality data cannot be produced.