The world has been reminded of Jevon’s paradox over the past year or so since the onset of AI as an overarching theme in our lives. It has been referred to in the context of the potential job destruction that AI could bring upon us, mostly by the proponents of AI saying that as workers become efficient by using AI, demand for workers will actually go up and not down. In general, it is about the paradox that as a resource becomes more efficient, demand for that resource doesn’t decline but grows as use cases for the resource opens at new productivity levels.
“The original formulation of “Jevons paradox”, by William Stanley Jevons in 1865, was about coal production. Jevons observed that, the cheaper and faster we got at producing coal, the more coal we ended up using – demand more than eclipsed the cost savings, and the coal market grew rapidly as it fed the Second Industrial Revolution in England and abroad.”
It has been demonstrably seen elsewhere too, most notably in the world of semiconductors: “ Today we all know Moore’s Law, the best contemporary example of Jevons paradox. In 1965, a transistor cost roughly $1. Today it costs a fraction of a millionth of a cent. This extraordinary collapse in computing costs – a billionfold improvement – did not lead to modest, proportional increases in computer use. It triggered an explosion of applications that would have been unthinkable at earlier price points. At $1 per transistor, computers made sense for military calculations and corporate payroll. At a thousandth of a cent, they made sense for word processing and databases. At a millionth of a cent, they made sense in thermostats and greeting cards. At a billionth of a cent, we embed them in disposable shipping tags that transmit their location once and are thrown away. The efficiency gains haven’t reduced our total computing consumption: they’ve made computing so cheap that we now use trillions times more of it.”
The author reckons the same is likely to happen with AI, not just in the chips and the compute infrastructure but as new use cases create opportunities for work that currently don’t exist.
“We’re all betting that the same will happen with the cost of tokens, just like it happened to the cost of computing, which in turn unlocks more demand than can possibly be taken up by the existing investment. The other week, we heard from Amin Vahdat, GP and GM of AI and Infrastructure at Google Cloud, share an astonishing observation with us: that 7 year old TPUs were still seeing 100% utilization inside Google. That is one of the things you see with Jevons Paradox: the opportunity to do productive work explodes in possibility. We are at the point in the technology curve with AI where every day someone figures out something new to do with them, meaning users will take any chip they can get, and use it productively.
Jevons Paradox (which isn’t really a paradox at all; it’s just economics) is where demand creation comes from, and where new kinds of attractive jobs come from. And that huge new supply of viable, productive opportunity is our starting point to understand the other half of our economic puzzle: what happens everywhere else.”
To explain what happens everywhere else, i.e, in areas which are not directly affected by AI, he introduces ‘Baumol’s cost disease’, to say that there will not only be increased demand for AI enabled workers, but even workers in areas totally detached from AI could benefit, based on the work of William Baumol who published his work a century after Jevon.
“…the basic argument is, over the long run all jobs and wage scales compete in the labor market with every other job and wage scale. If one sector becomes hugely productive, and creates tons of well-paying jobs, then every other sector’s wages eventually have to rise, in order for their jobs to remain attractive for anyone.
But the odd thing about Baumol’s is how rarely it is juxtaposed with the actual driving cause of those productivity distortions, which is the massive increase in productivity, in overall wealth, and in overall consumption, that’s required for Baumol’s to kick in. In a weird way, Jevons is necessary for Baumol’s to happen.
For some reason, we rarely see those two phenomena juxtaposed against each other, but they’re related. For the Baumol Effect to take place as classically presented, there must be a total increase in productive output and opportunity; not just a relative increase in productivity, from the booming industry and the new jobs that it creates. But when that does happen, and we see a lot of consumption, job opportunities, and prosperity get created by the boom, you can safely bet that Baumol’s will manifest itself in faraway corners of the economy. This isn’t all bad; it’s how wealth gets spread around and how a rising tide lifts many boats.”
The argument fits in conveniently with those who believe new technology is always beneficial to all of humanity. Whilst we came out of the industrial revolution figuring out something else to occupy us with and in all likelihood, we shall this time as well, the challenge is that Baumol’s argument holds only over the long run and hence the lived experience through the transition period even if it is as short as a decade will likely be painful for some. Hence, as someone put it, humans will not lose their jobs to AI but to other humans who learn to use AI well.
If you want to read our other published material, please visit https://marcellus.in/blog/
Note: The above material is neither investment research, nor financial advice. Marcellus does not seek payment for or business from this publication in any shape or form. The information provided is intended for educational purposes only. Marcellus Investment Managers is regulated by the Securities and Exchange Board of India (SEBI) and is also an FME (Non-Retail) with the International Financial Services Centres Authority (IFSCA) as a provider of Portfolio Management Services. Additionally, Marcellus is also registered with US Securities and Exchange Commission (“US SEC”) as an Investment Advisor.
