This is an excerpt from The Breakdown newspaper. To read the full article, register.
“Design is not everything, but in the long run it is almost everything.”
—Paul Krugman
“Total factor productivity” (TFP) is how economists measure the contribution of technology to economic growth – the sustainable ability of an economy to produce more output with the same quantity.
As such, it is the most important measure for economists, because the way to continue producing more with less is how life is better.
“A country’s ability to raise living standards over time depends almost entirely on its ability to raise its output per worker,” explains Paul Krugman.
Technology is what makes this possible and TFP is how it is measured.
To learn more about the importance of creativity, consider the following: A recent paper from the National Bureau of Economic Research (NBER) estimates that an additional 0.5% of annual TFP growth will stabilize US government spending at the current level of debt-to-GDP.
0.5% on!
That doesn’t sound like a lot, but if it continues over the next 10 years, the NBER estimates it could reduce the projected US government debt by $2 trillion.
More than 30 years, a stable 0.5% boost to TFP would make the US government’s debt-to-GDP ratio 42 percentage points lower than the NBER’s baseline (and 80 percentage points lower than its non-believer).
Seeing how the public finances look hopeless, staying in debt today is a dream that seems too good to come true.
But Anthropic researchers think we can do much better.
Anthropic conducted a survey of 100,000 conversations about Claude.ai to “estimate how long the tasks in question would take without the help of AI, and to learn how it affects the overall economy.”
The ending? LLMs can raise overall productivity by 1.1 percent.
1.1% on!
If 0.5% can stabilize the US government’s economy for decades, what can 1.1% do? It can fix almost anything.
There are reasons to be skeptical of this optimistic prediction, of course.
The study found, for example, that Claude saves teachers four hours of work by creating lessons in just 11 minutes. But estimating how much such time savings can lead to substantial wealth requires a type of financial system that is riddled with assumptions and false precision.
So, while Anthropic is right about saving time, it could be wrong in design: It could be that we are always using AI to save us from doing something useless, like watching more TikTok videos or reading more articles.
In that case, AI may improve our well-being (more free time) but not our wealth (economic surpluses) – great news for society, but no help for governments hoping for a solution to the debt crisis.
On the contrary, there are reasons to think that the Anthropic model also exists hopeless: “We don’t think about the number of adoptions” it says, “or the big results that can come from AI systems that have a lot of potential.”
In other words, his research assumes that we continue to use AI as we do now and that we are still using modern, unchanging, alternative languages. 10 years.
Linguistics get better every few months and we’re only just starting to learn how to use them – so Anthropic is right to say that its estimate may represent “nearly a drop in AI results.”
If so – if 1.1% then at the bottom in the development of AI – we can pay off the government debt and spend more time on TikTok.
And that’s just thinking about AI’s impact on non-physical jobs – just wait until we get robots!
To end such optimism is to assume that the billions of dollars that organizations plan to spend on AI capex and R&D will all be wasted.
Be that as it may – technological change does not always arrive.
But the biggest reason for optimism is that Anthropic’s 1.1% estimate is based only on AI “making existing tasks faster to complete” – his model does not take into account AI’s ability to completely change the way we complete those tasks.
“Historically,” anthropic says, “changes in productivity – from electricity, computers, or the Internet – have not come from speeding up old jobs, but from reinventing production.”
There is no way to prove these new ways of doing things, but it seems that the results will be bigger than what Anthropic tried to test.
Anthropic researchers are careful to disguise their promising findings by explaining the limitations of their methods and documenting the many hypotheses they are developing.
And even if all these ideas turn out well and the invention of AI solves the US government’s debt crisis, policymakers may come back for it.
But because of the 100% probability that everyone seems to be facing an impending financial crisis, even a small chance that Anthropic’s estimates are correct is a reason to change our expectations: The US government budget is not as bad as we think, and the US dollar is not as weak as we think.
In the long run, productivity will be almost everything – and AI may be on the verge of making us more productive.
Get news in your inbox. Check out the Blockworks blog post:
