Don’t be a plumber

Let me pause from considering the symptoms of a possible AI-induced bubble, and instead consider possible AI-induced mass unemployment. Geoffrey Hinton, pioneering AI researcher, gave some career advice last year. He said you should be a plumber:

The jobs that are going to survive AI for a long time are jobs where you have to be very adaptable and physically skilled, and plumbing’s that kind of job.¹

Is this good career advice? No. Here’s an example of good career advice, which is from me to Hinton.

Keep your day job. Stick to computer science and leave the economics to the economists.

Now, while I am an economist, I am not a labor economist nor an expert in the economics of technology and growth. So in this post, I will share wisdom from bona fide experts in these fields. If you want to learn more, I particularly recommend that you read Autor (2024).

Let me make two points. First, incorrectly predicting unemployment due to technology improvements is a time-honored mistake, so Hinton is in good company here. Second, while plumbers are good, and it is certainly fine with me if you want to be a plumber, I do not recommend that you become a plumber due to concerns about AI.

The boy who cried wolf

This is not the first time Hinton has given terrible advice. Consider this recommendation from 2016:

I think that if you work as a radiologist you are like Wile E. Coyote in the cartoon. You’re already over the edge of the cliff, but you haven’t yet looked down. There’s no ground underneath. It’s just completely obvious that in five years deep learning is going to do better than radiologists….They should stop training radiologists now.”²

Hinton was wrong. Here we are eight years later, and while AI has indeed impacted the field of radiology, we currently have a shortage of radiologists in the U.S. and other countries. Thankfully, no one heeded Hinton’s advice in 2016.

Here’s Agrawal, Gans, and Goldfarb (2019) commenting on the situation:

However, the effect of artificial intelligence on the number of workers in radiology turns out, on closer examination, to be ambiguous and nuanced.

… It is ultimately not obvious even whether the number of radiologists will rise or fall, since that will depend on whether radiologists perform the nonprediction tasks and whether overall demand for radiology services rises as radiology becomes more efficient.

Now, predictions of unemployment due to technology are extremely common over history, including the great economist David Ricardo in 1821 as quoted in Hollander (2019):

the substitution of machinery for human labour, is often very injurious to the interests of the class of labourers

It is just not true that automating certain tasks formerly done by humans produces widespread unemployment, at least historically. (It typically hurts a subset of workers but helps workers as a whole.) Instead, technology has made most workers better off by making them more productive. Here’s Autor (2024) discussing predictions of AI-driven disaster:

The most charitable thing I can say about these ominous statements is that they are likely wrong — a flattening of the complexity of innovation into a single dimension of automation. Do these technology visionaries believe that Black & Decker tools make contractors’ skills less valuable and that airplanes outperform their passengers? The latter question is of course nonsensical. Airplanes are not our competitors; we simply couldn’t fly without them.

Now, in the story of the boy who cried wolf, a wolf does eventually appear. Could AI be a case where the wolf will actually appear, and we actually will see mass unemployment in every sector except plumbing and similar fields?

I think there are two reasons that a disastrous downside scenario could be plausible, although as I’ll explain in the next section, in neither case is a career in plumbing the right response.

First, one could imagine that the wolf of technological unemployment will happen due to the speed of AI development. Here’s John Stuart Mill as quoted in Hollander (2019):

I do not believe that as things are actually transacted, improvements in production are often, if ever injurious, even temporarily, to the labouring classes in the aggregate. They would be so if they took place suddenly to a great amount, because much of the capital sunk must necessarily in that case be provided from funds already employed as circulating capital. But improvements are always introduced very gradually, and are seldom or never made by withdrawing circulating capital from actual production, but are made by the employment of the annual increase.

And Keynes (1930):

We are being afflicted with a new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come – namely, technological unemployment. This means unemployment due to our discovery of means of economising the use of labour outrunning the pace at which we can find new uses for labour.

So, under the “outrunning the pace” story of Keynes and Mill, it would be fine if AI were introduced over the course of decades but not fine if introduced over two or three years.

Second, a different manifestation of the wolf of unemployment would be artificial general intelligence or AGI, defined (somewhat ominously by OpenAI) as “highly autonomous systems that outperform humans at most economically valuable work.”

How likely is it that we invent AGI? Well, here is another quote from Hinton from May 2023, and this time he might be right:

I have suddenly switched my views on whether these things are going to be more intelligent than us. I think they’re very close to it now and they will be much more intelligent than us in the future.³

Past technological changes sometimes augmented human abilities (Excel spreadsheets), sometimes automated tasks entirely (the automatic elevator eliminated the job of elevator operator), and sometimes created entirely new jobs (airplanes created many new jobs, such as air traffic controller). The fear is that AGI will automate 100% of all jobs.

What would happen under AGI? Here is Korinek (2023) describing two competing views:

If AI systems can perform all cognitive work and are cheaper, market wages of cognitive workers would tend to fall to the level of AI systems’ operating cost. And if AI systems in combination with advanced robots can perform virtually all human jobs in the future, wages for all human labor would tend to decline significantly (Korinek and Juelfs, 2022). A certain level of demand for human jobs would likely remain from two sources: because consumers may have an intrinsic preference for humans to perform certain services (e.g., law-making) and because our society may choose to retain certain jobs with humans for ethical reasons or for reasons of better controlling AI systems. Yet it is questionable whether these two categories of jobs would generate sufficient demand to maintain the current level of wages.

Other experts - including some economists who have studied earlier rounds of innovations - remain deeply skeptical based on a different perspective of how powerful AI might become in comparison to human brains. They point out that such a development would contradict our experience of the past two hundred years. They observe that in the past, the economy has always created new jobs when old jobs were automated. We have indeed regularly invented new jobs in the past, leaving automated tasks to machines and leveraging our unique human abilities in areas that were hitherto unautomated. However, at a fundamental level, all new jobs consist of recombinations of existing human abilities. Once machines can acquire all those abilities, they would be as good as humans at any new jobs too.

Plumbers are not the wave of the future

Now, if you are considering a career in plumbing, let me say I have nothing against plumbers. While some blame plumbers for the fall of the Roman Empire,⁴ I say plumbers are a force for good. Without plumbing, we would be plagued by filth and disease, and civilization as it currently exists would be impossible. So if you want to be a plumber, go for it.

But let me explain why the advent of AI does not imply that you should become a plumber. Let me sketch out two scenarios over the next 20 years. First, the AGI scenario where AI can do 100% of all human jobs. Second, the non-AGI scenario where AI is impactful but not science-fictional.

First, under the AGI scenario, being a plumber will probably not help. Korinek (2023) says “AI systems are also making robots more capable, which may soon lead to further automation of blue-collar jobs.” So instead of being an unemployed lawyer, you will be an unemployed plumber.

Second, under the non-AGI scenario, it is true that jobs like software engineer, lawyer, or financial analyst will be greatly impacted and perhaps plumbers would not. But that is not a good reason to be a plumber. Let’s do a thought experiment. Let’s say you had a time-machine and could go back to 1900, and talk to the greatest plumber of them all, Thomas Crapper (1836-1910).⁵ Should you tell Crapper that “in 2024, we no longer have the jobs of blacksmith, lamplighter, and scullery maid, but plumbing is just the same, and therefore plumbers are the wealthiest, happiest, and most powerful group in our society.  Make sure your descendants are also plumbers.” No! You should tell Crapper the following: “There will be many new jobs that you cannot understand, such as cybersecurity specialist and reality TV contestant. Make sure your descendants acquire good work habits, a passion for lifelong learning, and a basic knowledge of math and science, or else they will end up as reality TV contestants.”

References