Is GPT-3 the AI tool that launches a productivity revolution?

GPT-3 is an AI tool — machine learning, specifically — trained on an extremely large data set that can generate some pretty amazing text when prompted, everything from creative fiction to computer code. It’s also pretty good at answering questions. The tutorial example given on the OpenAI launch page shows a user looking up “bread” on Wikipedia and asking “Why is bread so fluffy?” As the demo user explains, the tool is able to point her directly to the paragraphs based on the context of the text. That, even though nowhere in the article does it use the word “fluffy.”

Anyway, lots of people are pretty excited by the demonstrated capabilities, with some even wondering if they’re seeing evidence of sentience or “general intelligence.” One AI researcher even wonders if GPT-3 is pretending to be dumber than what it actually is.

Yet even if GPT-3 doesn’t move us way down the path toward the technological Singularity, it could still be a pretty important advance for the economy. Tech analyst Eli Dourado tweets an interesting question: “If an AI like GPT-3 were on the cusp of revolutionizing productivity, what signs would you look for in financial markets?”

Good question. One obvious approach is to see what happened during the late 1990s productivity boom. And what happened was that stock price-earnings ratios and technology stock prices went through the roof. It was also a time when some techies were wondering if the age of supersmart machines was almost here. It’s tough to disentangle the data during a pandemic and depression recovery, but you probably have noticed that tech stocks are on an unbelievable tear. My CNBC colleague Jim Cramer said Monday that the recent upward moves of Amazon, Microsoft, and Tesla were “truly insane and unlike any i have ever seen in my life.” I bet a lot of analysts would talk like that during an AI-driven productivity revolution.

Another way is to dig through the economic data. In the 2017 paper “Are We Approaching an Economic Singularity? Information Technology and the Future of Economic Growth,” Nobel laureate William Nordhaus presents “several tests of whether we are rapidly approaching Singularity.”


There are six tests on the supply side. The conclusions from the empirical tests proposed here is that the substitution tests fail or are ambiguous for four of six tests and succeed barely for two of the six tests. However, the growth trajectories of the variables which pass the test (the share of capital in total income and the share of informational capital in total capital) are extremely slow. Projecting the trends of the last decade or more, it would be in the order of a century before these variables would reach the level associated with the growth Singularity. The conclusion is therefore that the economic Singularity is not near.

So the Singularity was not near three years ago, nor it seems a productivity revolution. But we’ll keep an eye on the numbers — and GPT-3.