🐙 AI‘s table stakes: how much is all in?
If LLMs are GDP boosters, how much should governments invest to keep their economies competitive
I’ve been thinking about the increasing evidence that generative AI, specifically large language models (LLMs), could significantly boost labor productivity and drive up GDP.
Many countries have been in a productivity funk for years. Even emerging economies have experienced a slide in productivity growth since the Global Financial Crisis.
This development is likely to motivate governments to act. For governments, if AI (or any technology or policy) offers a way out of the slump, they will want to take it.1
They can’t look a gift horse in the mouth.
When I spoke with economists, Eric Brynjolfsson and Lindsey Raymond, a few days ago, we discussed their new paper demonstrating that labor productivity improved by about 14% in certain jobs.
A Goldman Sachs paper suggested that generative AI could improve productivity in the US by 1.5 to 2.9% over a 10-year period. Globally, this could translate to a 7% growth in GDP. In the context of the UK and US economies, this means an additional $210 billion and $1.6 trillion, respectively.
These staggering numbers should prompt governments to ensure their economies are well-positioned to capitalize on these benefits. Public-private partnerships will likely play a crucial role in this process, with the market driving investment and governments setting frameworks to further optimise the results.
The UK government, for instance, has allocated £900 million to build a supercomputer and another £100 million to train a foundation model. If we scale these numbers by GDP for the US, that would amount to about $9.9 billion. However, the US allocation for a national AI research task force supercomputer is only $2.6 billion over five years, indicating that the UK is making a more substantial commitment on a per GDP basis.
But is this enough, and is it the right approach?