Skip to content

Shocking Revelation: David Autor Warns Reprogramming our Future with AI! The Untold Design Dilemma

“In a series called ‘Economists Exchange,’ David Autor, a professor of economics at MIT, provides an optimistic view on how generative AI can help address the hollowing out of America’s middle class. He believes that AI has the potential to give people without a college education the tools they need to do more expert work, earn higher wages, and close the income gap with top earners. This stands in contrast to the warnings issued by other economists who fear that AI will exacerbate inequality and leave many workers behind.

Autor explains that AI is different from previous technological advances because it can handle problems that require judgment, inference, and flexibility. It learns from examples and statistical reasoning, allowing it to automate tasks that were previously not subject to automation. However, he emphasizes that AI should be used to make people’s skills more valuable rather than replace them.

Autor acknowledges that AI may change the availability of certain types of expertise and reduce the scarcity of professional skills. But he sees this as an opportunity to complement foundational expertise or judgment, enabling people with slightly less elite skills to perform more expert work. He gives the example of nurse practitioners performing tasks traditionally done by doctors, supported by AI tools.

In terms of job postings, Autor’s study suggests that firms adopting AI are hiring fewer workers for tasks that AI can handle, such as information processing and decision-making. However, he believes that this is just part of the job design challenge that comes with new technology. Like the clerical occupation, which has evolved with computers, AI will present opportunities to reallocate and redesign work. Autor hopes that AI can reinstate the value of skills held by people with less formal education.

Ultimately, Autor’s view is that AI can be a complement to human skills rather than a substitute, as long as it enhances expertise and helps individuals perform more advanced tasks. However, he warns that if AI simply provides what was previously scarce for free, it could undermine the value of certain professions or skills.

In summary, Autor believes that AI has the potential to benefit workers, particularly those without a college education, by making their skills more valuable and enabling them to perform higher-level work. He sees AI as a tool for complementing human expertise rather than replacing it, and emphasizes the importance of using AI in a way that promotes inclusivity and closes the income gap.”

—————————————————-

Article Link
UK Artful Impressions Premiere Etsy Store
Sponsored Content View
90’s Rock Band Review View
Ted Lasso’s MacBook Guide View
Nature’s Secret to More Energy View
Ancient Recipe for Weight Loss View
MacBook Air i3 vs i5 View
You Need a VPN in 2023 – Liberty Shield View

This is part of a series, ‘Economists Exchange’, featuring conversations between top FT commentators and leading economists

Globalisation and the advent of office computing hollowed out America’s middle class — devastating communities built on manufacturing and devaluing administrative skills, to the benefit of a small elite.

Generative AI could help redress the balance, giving people without a college education the tools they need to do more expert work, win higher wages and close the gap with top earners. This is the optimistic message of David Autor, a professor of economics at the Massachusetts Institute of Technology, whose previous work on the so-called China shock underpinned a rethinking of US trade policy.

His upbeat view contrasts with the warnings issued by other economists who believe AI could leave many workers behind and entrench existing inequality, long before it pays off in higher productivity.

Autor, who has spent his career exploring how technological change affects jobs, wages and inequality, argues that the latest advances in AI come at a time when workers are in short supply. With the right design, AI can be used to make people’s skills more valuable, rather than to replace them.

In this interview, he talks about the risks AI poses to emerging markets in particular, but also the potential to rebuild middle-class jobs against the backdrop of a tight post-pandemic labour market. 

Delphine Strauss: David, you’ve spent many years exploring how technological change affects workers. What do you make of the wildly varying claims being made on the transformative impact of AI? What have we seen historically and why might it be different this time?

David Autor: I do think AI is different. What makes it different fundamentally is the type of problems it can be applied to. Classical computing is great for following rules and operating tools and carrying out well-understood procedures that we could fully formalise and codify and have the machine execute without any judgment or problem solving, or common sense.

That need to codify everything has been a barrier to many of the things we could automate because many of the things that we do regularly, we don’t know how we do them. We don’t know how we come up with a convincing argument or develop a hypothesis or make a creative drawing. Nor do we even know how we figure out what dishes to pick up in a sink to wash and set a table and cook. We have all these capabilities that we understand tacitly but not explicitly.

The need for explicit understanding was the key barrier to computerisation. AI overcomes that, because it learns tacitly as well. It learns from examples and statistical reasoning or statistical inference and finding associations between things, and it’s able to connect the dots . . . It’s not following a set of pre-specified rules. That allows it to do all kinds of tasks that require judgment and inference and flexibility that were previously not subject to automation.

That can be writing. It can also be many research tasks. It could be decision-making. Many decision-making tasks require discretion — in management, in piloting a plane. So I do think it’s going to change, very substantially, the set of things in which we can use computers effectively.

It’s critical to bear in mind that not everything that we use technology for is automated. Technology often provides us with tools. Our work would be unthinkable without our personal computers. You would not want to be a roofer if you didn’t have a pneumatic nail gun to put on shingles. You wouldn’t want to be an air traffic controller if you didn’t have a GPS, a two-way radio and a radar system.

Many, many technologies provide us with tools, but those tools have different places where they substitute or complement us. Computers have substituted us in carrying out rule-based procedures and complemented us when we need that information to make better decisions.

I think AI is going to change the availability of certain types of expertise. In some sense the last four decades have made professional expertise increasingly scarce, relative to demand, because it cleared out all the bottlenecks of all the other calculations, and information retrieval and sorting and filing. It left us in the decision-making mode. How do we architect; how do we care for a patient? How do we design a piece of software? How do we read a legal brief? And in some sense, all the difficult discretionary tasks were left for people because all the rule-based tasks were executed instantly and costlessly.

DS: Does that mean AI is where automation starts to hit top earners?

DA: It’s possible to say that but that’s not my focus — it’s hard to lament both the rise of inequality and the fall of inequalities simultaneously. Computerisation has definitely helped white collar professionals, the MDs and PhDs and JDs and MBAs and so on. Computerisation has helped make them extremely scarce. We can see that not only in rising wages, but the fact that they work many, many more hours than they used to. I think AI is going to reduce the bottleneck of expertise in some areas, but that can complement others.

There are many paths where you have foundational judgment, acquired through experience or training, bounded by some upper bound of technical or specific knowledge. For example, if you’re a nurse practitioner, you can do many care tasks but not all of them — but in the US, there’s been a big innovation. Twenty years ago, there was no one other than a doctor who could prescribe medicines and diagnose and treat. Now a nurse practitioner can do that job, supplemented by lots of technology that gives access to electronic medical records, to diagnostic information, to supervisory programmes that say, don’t put these two drugs together.

The good case for AI is where it enables people with foundational expertise or judgment to do more expert work with less expertise. The nurse practitioner example is a focal one. We’re taking the most elite tasks and allowing someone with somewhat less elite skills to perform them. Nurse practitioners are highly skilled professionals but they have five fewer years of education than medical doctors.

I think this is something we could use AI for in many other settings. To put it in simple economic terms, the question to ask is for whom is AI a substitute and for whom is it a complement?

If it’s purely a substitute, if it takes some form of expertise you had and provides it for free, that’s not good news. If you’re a London taxi driver and you spent years memorising all the streets of London and now, everyone else can do exactly what you could do, suddenly that expertise is no longer scarce, right?

DS: What did your recent study of job postings tell you about the skills that are becoming more or less valued by employers at the frontier of AI?

DA: We do see reduced hiring at firms adopting AI in some of the tasks that AI is good for — information processing, some software coding, decision-making tasks. But I don’t think that is in any sense a full description of what’s going to occur.

It’s actually a challenge of job design to figure out how we reallocate and redesign work, given the tools we now have available. This often takes a long time to figure out.

For example, the clerical occupation has been dramatically changed by computers. There are fewer clerical workers but those who remain do a much harder job, because they don’t do the typing and the filing and the spell checking. They do the proofreading, the organising of travel and events, the handling of expenses.

I think AI is going to present us with a similar set of opportunities and we have to use it well. My hope is that we can use AI to reinstate the value of skills held by people without as high a degree of formal education.

Let me just try to synthesise that point about complements or substitutes. Imagine, to give you a very concrete example, I want to rewire my house. I have some time this Sunday and I’m going to tear out the breaker box and replace it with a higher amperage breaker box. I could go to YouTube and see videos on how to do that. Now, let’s say I didn’t know anything about electrical work. I’m surely going to electrocute myself or set my house on fire, because I don’t know how to handle high-voltage wiring. I don’t know how to strip a wire. I don’t know how to put circuits together. I don’t know how to do testing. It’d be terrible.

However, let’s say I have some skills in electrical work, but I’ve never replaced a breaker box. Having access to YouTube would be really useful to me. It’s not a substitute for expertise, it’s a complement. Up to a certain point, it allows me to take that same foundational expertise and apply it to a broader set of tasks. I think there are many, many jobs that are going to look like this.

DS: Surely there are occasions where AI boosts one group and displaces another? Google Translate allows me to read a press release in, say, Hebrew, but puts translators out of work…

DA: Absolutely. I did not mean to suggest that there’s a win-win or that everybody’s better off. It is definitely not the case and it never has been. There are almost no technologies I can think of that are important that haven’t displaced one set of people even while benefiting another. That would be true for mass production and its displacement of artisans. It would be true for agricultural technology that caused a lot of people to leave farms behind. It’d be true for automation and manufacturing, especially the computer revolution, and it’s happened in a lot of white-collar work as well.

I definitely think there are occupations, some of them valuable, that will be directly displaced. Translators, some of those will survive, but they will be very, very elite. A lot of people who do service bureau work in the developing world, who translate code from Cobol to Python — that work can be done so much more efficiently now. Or people who take your slides and apply the McKinsey template — that’s a really good task for AI.

Or you could say, I’m going to London this weekend — find me flights and a hotel and a place to eat. You could have a machine do that. I do absolutely think there’s work that will be displaced.

DS: Is it potentially going to have a bigger displacement effect in developing countries?

DA: I’m very concerned about this. There are a lot of professional support tasks that we outsource to the Philippines, that we outsource to Mumbai and so on. I don’t think it’s all going to go right away and some of it will get better. You could say, it’ll just be so cheap and so good that everybody will want to do it all the time. That could happen, like with Uber. Uber made transportation more efficient but it caused more people to work in transportation services because the demand response was so strong.

But I’m not sure — there’s only so many nice PowerPoint presentations the world can withstand. So I am very worried about that. And it’s not exclusively in the developing world. MIT undergraduates have always thought, well, I can always fall back on coding. Now they’re thinking they can’t. It’s going to reach much more into the white-collar skilled ranks than it used to. Educated workers are in general better prepared to adjust to these things but nevertheless it will be challenging.

DS: When one thinks about groups of losers, do you think we’re looking at something similar to the China shock in terms of speed and concentration?

DA: No, I don’t. One thing that made the China shock so shocking was that it was very geographically concentrated — the textile workers are in one place, the commodity furniture workers are in one place and the whole town is built around that industry. There was never a clerical worker shock in the US and there was never a clerical capital of the world. It’s a much more distributed set of occupations and tasks. For that reason, it will not be the kind of concentrated economic blast or bomb, it won’t leave the same type of crater. But it will move fast, because AI will deploy fast.

But the question we should be concerned about is not the number of jobs. We have a labour shortage throughout the industrialised world. I am concerned about the number of jobs in Mumbai, but in the UK, in the US and northern, western Europe, we are running out of workers.

The concern we should have is about expertise. If people are doing expert work that pays well and now they have to do generic work that pays poorly, that’s a concern. It’s the quality of jobs, not the quantity. The problem is technology can make some expertise much, much more valuable, but in other cases, it directly replaces expertise we already have.

Now, we tend to create new demands for expertise. As we moved out of agriculture, work has gotten broader, not narrower. We have hundreds of medical specialties. But there is always a race between technology automating away something that was formerly expert and new things that we’re creating. I do think the set of things that will be automated away by AI are different from anything in the past.

DS: Moving on from AI to the work you’ve done on the post-pandemic labour market — you’ve shown that in the US, the only group of workers that haven’t seen their wage gains wiped out by inflation are the lowest paid. Is that a scarcity effect or are we now valuing what they do more?

DA: It’s hard to disentangle, because so many things happened at once. I think they are moving into better jobs and they’re being better treated in the jobs that they had. A substantial part of this is driven by scarcity. In the US, the unemployment rate was at a historic low, even in 2020. Inequality had been falling since at least 2015. The pandemic accelerated this. It created a big demand surge that created additional labour scarcity. It also disrupted people’s expectations. If anyone in the US had ever felt loyal to their employer before, they certainly don’t now because so many of them were laid off. And of course it also gave people money in the bank, to search for better jobs.

So I do think it changed workers’ expectations of what they could get out of employers. It means that the less productive firms that pay lower wages are less able to hold on to labour. If you think the labour market is operating more competitively, employers can’t underpay workers and expect to hang on to them, that means greater efficiency. Employers won’t like it, but it means workers will be working at better firms and the less productive firms will be more likely to go under. That’s good. That’s the icy fingers of the invisible hand, working their magic.

DS: Do you think this will continue? Or could it end as interest rates go up?

DA: It’s already decelerating. It’s not at the same fever pitch that it was six months ago. But the labour scarcity is going to be with us for a while because of demographics and low fertility and in the US heavily restricted immigration — which is very harmful, both for the US, and for immigrants.

Of course, the central bank can always put the kibosh on it. If we announce we’re having a big recession tomorrow, then labour markets will be slack at least for a while, but they’ll be likely to come back again. Just like we came back from the pandemic, just ferociously.

Because of large populations of workers entering retirement or in retirement, small young cohorts, there’s a tendency towards labour scarcity. I don’t think automation is proceeding in a way that’s going to radically reduce that — it hasn’t. Over the last 40 years we’ve generally had pretty low unemployment rates and low interest rates. We haven’t had an employment problem.

We’ve had a job quality problem, as many people have been displaced from skilled manufacturing and skilled office work by computerisation, and then by trade, and moved into generic non-expert work — food service, cleaning, security. To be clear, it’s not a value judgment. That work is important. Being a daycare teacher or a crossing guard, these are life and death matters. Nevertheless they’re poorly remunerated in every country, the reason being they don’t require great expertise.

This is where technology can be both very helpful or very harmful. It’s helpful to the degree it complements expertise and makes people’s skill set more valuable by allowing them to do more with it. It’s harmful to the degree that it takes skills that we’ve invested in that are the basis of our livelihood and makes them so abundant that they aren’t worth anything any more.

I think we have a real design choice about how we deploy AI. It is so flexible, broadly applicable and malleable that we can do lots of stuff with it, some quite good, some quite bad, and depending on the mental model we have in mind of what we’re trying to do, we will accomplish different things.

DS: Do you think there’s any benevolent authority able to direct it?

DA: It has to be leadership in the private sector. It has to be leadership in universities, even government leadership to demonstrate what’s feasible. Medicine would be the best and obvious place to start. There’s unlimited demand for health care. We could never have enough workers in that sector and if we got more efficient at it, we’d just do more of it. So the question is, could we reorganise it to make it more accessible and less expensive and actually less error-prone as well?

If companies just set the goal to figure out how to use as few workers as possible, they’ll accomplish some of that too. China uses AI for many things, but one thing they’re incredibly good at is surveillance and real-time content filtering or censorship. It’s not because that’s what AI is for. It’s because that’s where they spent their AI money. But we could spend it differently. It really matters what we set out to do.

DS: The optimistic view is that we end up with a situation where low-wage workers are more scarce, more highly valued and have more bargaining power; the middle bump up their level of expertise by means of technology to make their skills more valuable; and people at the top may suffer?

DA: Exactly. If you think about it, computerisation had a big role of basically de-contenting much of the middle skilled work that people did, in offices, in factories. If we could reinstate the value of mass expertise by enabling people to do more in the trades, in healthcare, in contracting construction, in even some of the writing tasks that we do, that would be spectacular. If it comes at the expense of making some elite expertise less scarce I think that’s ok. I don’t think people with PhDs and MDs and JDs are going to be wiped out. They may just not see the same year over year wage growth that they’ve seen over the last several decades. That’s ok. They’ve had a good run and they’ll be fine.

DS: What else could governments or regulators do to give low-wage workers more lasting bargaining power?

DA: I think in the US we have really insufficient labour standards that affect the quality of work. Do you have predictable hours, do you have enough hours? We don’t have mandatory paid leave, we don’t have mandatory family leave. There are many things that could be done. I think they should be done, they hold up the bottom, they give workers more bargaining power. I don’t think they would be nearly as transformational as changing the type of work people do.

I think that we should be highly invested in — not simply labour standards, not simply tax and transfer — but also what kind of work we can create that enables people to use skills.

The above transcript has been edited for brevity and clarity 

—————————————————-