Interview: Aaron Benanav

Common Wealth’s Amelia Horgan spoke to sociologist and economic historian Aaron Benanav about ChatGPT, post-scarcity, and automation.

[.cdw-name]Amelia Horgan[.cdw-name]

What is automation theory as you describe it in Automation and the Future of Work (Verso, 2020)?

[.cdw-name]Aaron Benanav[.cdw-name]

In the 2010s, there was a lot of excitement about the idea that new technologies — digitalisation, internet connectivity, robots, automation and artificial intelligence — were going to end work as we know it. Silicon Valley was supposed to launch us into a wholly new kind of human existence — a post-scarcity world. By the end of the 2010s, and certainly by 2018, you saw the start of a big backlash against Silicon Valley, based on a growing recognition that, in reality, jobs weren't disappearing due to automation. In fact, tech companies were inventing technologies that depended on people still working and which often incorporated new ways to surveil and manage workers and manipulate consumers. That moment also opened up a political question about what these new technologies can do and what they'll be used to do if companies are allowed to exploit those technologies without restrictions. Are we heading for a world in which technology connects people in ways that make it easier for them to work together in a democratic way, or will technologies be used in ways that separate and atomise workers and make it much more difficult for them to organise and build solidarity?

By the end of the 2010s, and certainly by 2018, you saw the start of a big backlash against Silicon Valley, based on a growing recognition that, in reality, jobs weren't disappearing due to automation. In fact, tech companies were inventing technologies that depended on people still working and which often incorporated new ways to surveil and manage workers and manipulate consumers. That moment also opened up a political question about what these new technologies can do and what they'll be used to do if companies are allowed to exploit those technologies without restrictions. Are we heading for a world in which technology connects people in ways that make it easier for them to work together in a democratic way, or will technologies be used in ways that separate and atomise workers and make it much more difficult for them to organise and build solidarity?

[.cdw-name]Amelia Horgan[.cdw-name]

Why does the end-of-work-automation thesis, despite the possibility that it's wrong, have such an appeal?

[.cdw-name]Aaron Benanav[.cdw-name]

Firstly, there’s a Popular Mechanics mindset that a lot of people still have that what carries our society forward and changes it is technology. When we are feeling pessimistic about the chance for social change, we look to technology to bring us to a happier place, especially when we distrust power, politics and the capacity of other people to bring us there. I think that the excitement about technology is related — though I’ve never exactly figured out how to describe this in precise detail — to a belief in technocracy, to the idea that a small group of smart people are coming up with solutions that are going to fix everything. The puncturing of those beliefs is really damaging for a lot of people. I think about this a lot in relation to conspiracy theories — about Covid-19, and even 9/11 in an earlier era. What happens when people who used to believe that technologies and technocracy were going to save them lose faith in those entities and start to believe that they might be doing more harm than good?


[.cdw-name]Amelia Horgan[.cdw-name]

Why, despite all its apparent wonder — the wonderful and brilliant technologies that we have, that we can have a conversation over Zoom, that we all have iPhones and can use ChatGPT — is the story that automation is going to end work wrong?

[.cdw-name]Aaron Benanav[.cdw-name]

There are two ways to approach that question. In my book, I approached it through an economic lens. Economic statistics help us get a better sense of how quickly new technologies are adopted throughout the economy. What those statistics show, without question, is that the adoption of these technologies is taking place very slowly and gradually and that we don't really live in an age of particularly rapid economic transformation. We live in an age of near economic stagnation, defined by a startling absence of broadly based and rapid technological change — at least from an economic perspective. That's a problem that people have had a lot of trouble understanding. In my view, it ultimately has to do with the fact that a lot of the brilliant technologies of the past were applied to transform the industrial sector; the industrial sector is designed and primed for this transformation because it is an entirely man-made environment where people do very repetitive tasks. As such it's really primed for rapid productivity growth, scale economies, all those kinds of things. But since the 1970s, the industrial economy has been shrinking in terms of employment, and we’ve seen the growth of a massive service sector, which is defined by very low rates of productivity growth — it’s very hard to use technology to make those activities more efficient. As the economy has shifted over into services, all the brilliant technologies we're developing aren't diffusing through the economy in the way that people expect. They're not even diffusing through manufacturing in the way people expect, because manufacturing has suffered from the fact that more and more firms and more and more countries are struggling to produce the same sorts of things.

In the 2010s, when Silicon Valley was talking about this new age of robots, and how they're going to transform everything, the manufacturing sector in the US had zero productivity growth; it was the worst decade for productivity growth since they started counting it. That’s one way in which you could say the automation story isn’t working out — we know it’s not really working out because we can’t see it working in the economic statistics.

But a second approach, one which has been most productively followed by a group at MIT, the Work of the Future group, is to go and see how these technologies are being implemented in production and to look at their limits — not just from an economic perspective, but from a technology perspective. Despite the incredible growth in the use of robots, most robots are used to do a very small number of tasks in just a few industries; most of the robots are used to move heavy things from one place to another. The Kiva robot in Amazon warehouses moves very heavy things from one place to another. Robots are also used to do things like welding together pieces of metal, painting, and cutting.

In general, though, the kinds of tasks that robots are being used to do today are the same things that they were used to do before. A lot of the new technologies that people are excited about — the Internet of Things, collaborative robots, all that kind of stuff — have not been adopted for technical reasons. And, in a lot of small and medium-sized enterprises, which account for a very substantial part of manufacturing, there are basically no robots at all, because, even as their prices come down, it’s still very expensive to programme robots. If you work in an industry that requires a lot of customisations, it's very expensive to reprogramme these robots again and again.

AI technologies also suffer from technical problems in their use. A lot of the big breakthroughs have been in natural language processes — this is what ChatGTP is good at — but those advances have happened because AI researchers abandoned the effort to get computers to understand logic and to think in a reasoned way. The new technologies are getting very good at fluency, but to some extent, they’re fluent in bullshit. When you ask them questions that require symbolic reasoning, like maths questions, or logic questions, or even ask them to generate scholarship and quotations, they don't understand what that means. They don't understand the difference between what's real and not real, so they tend to generate a lot of stuff that just isn't true. There’s no way with the current technologies to solve those problems. There are many fields in life where being fluent in bullshit is a great skill, and having computers that fluent in bullshit will improve productivity in some limited sectors of the economy but sometimes what’s needed is actual reasoning — traditional software programming, for example, will still likely be the major way that we programme systems.

 

[.cdw-name]Amelia Horgan[.cdw-name]

ChatGPT is fun to play around with. It’s very good at knowing which word should come next, like knowing the mechanics of a pop song.

[.cdw-name]Aaron Benanav[.cdw-name]
That’s what it does — it’s programmed to predict the next word or a missing word or a sequence of words, and then they give it a lot of the internet and a complex algorithm for predicting the next word. But that’s all it does. It doesn’t have logic.

 

[.cdw-name]Amelia Horgan[.cdw-name]
What do you make of the widespread panic among university and presumably school teaching staff about the use of Chat GPT by students? 

[.cdw-name]Aaron Benanav[.cdw-name]

This is related to questions about our education system in general, I think. Our education system is so outdated — it’s a system designed for the most part to convince peasant children to become factory workers, and it’s not very good at motivating people. A lot of students learn from the education system that they won’t be rewarded for developing their own internal motivations to study and learn and explore, but that they will be rewarded for producing whatever teachers tell them to produce. Technologies like ChatGPT could lead to a productive crisis of the education system and one that I would hope would lead people who control educational institutions to listen to people who study education and consider what better methods are available. Nobody thinks the existing grading system is good, for example. It encourages the kind of mechanical, unmotivated responses that ChatGPT is good at producing.

This is related to questions about our education system in general, I think. Our education system is so outdated — it’s a system designed for the most part to convince peasant children to become factory workers, and it’s not very good at motivating people. A lot of students learn from the education system that they won’t be rewarded for developing their own internal motivations to study and learn and explore, but that they will be rewarded for producing whatever teachers tell them to produce. Technologies like ChatGPT could lead to a productive crisis of the education system and one that I would hope would lead people who control educational institutions to listen to people who study education and consider what better methods are available. Nobody thinks the existing grading system is good, for example. It encourages the kind of mechanical, unmotivated responses that ChatGPT is good at producing.

[.cdw-name]Amelia Horgan[.cdw-name]
The panic seems to speak to an actual set of crises, then. One is about a lack of trust in students on the part of teachers and vice versa, but then about the structural paradox you’re describing too.

[.cdw-name]Aaron Benanav[.cdw-name]

In a system that was good at motivating people to want to learn, ChatGPT and similar things would be an incredible tool. But in a system in which people are asked to pretend to be interested, then it’s dangerous. I hope it produces positive change; I think it’s an interesting tool and I don’t think that the fact that it’s just making stuff up a lot of the time means that it won’t be useful to people in all kinds of ways.

 

[.cdw-name]Amelia Horgan[.cdw-name]

In the debates on work, the questions of quality and quantity of work are often separated, despite their interwovenness both conceptually and historically. Through what kind of practical or theoretical frameworks could they be combined, if they should be, that is? 

[.cdw-name]Aaron Benanav[.cdw-name]

Around the world, there has been a very long-term problem of there not being enough work. At the same time, there has been a political push to reform unemployment benefits and reduce access to welfare and generally create an environment where people need to work to survive. That frantic search for work and job insecurity then encourages job quality to decline. Except when it comes to highly skilled workers, like computer programmers, business owners do not need to pay attention to workers’ needs. Part of the issue here is that it's hard to know how to measure how many jobs there are precisely because the economy is suffused with low-quality jobs that people move in and out of. In the US we have very low unemployment rates. But at the same time, the number of people in work keeps growing because people pass directly from not being in the labour force at all to being employed, without passing through a stage where they are counted as unemployed. 

One important thing to look at is the labour share of income, which functions as an overall measure of the balance of forces between capital and labour. That’s been trending downward in a lot of countries for a long time. When we look at statistics like that, we should think about them as giving us qualitative information. Because when workers are in a poor bargaining position, in terms of wages, they also give up a lot in terms of the quality of work: break times and safety, and they lose the ability to shape how new technologies are implemented in their workplace (in the past few months, evidence has emerged that the lowest wage workers in urban areas have seen their bargaining power improve in the aftermath of the Covid crisis, although this situation is likely to be temporary).

One important thing to look at is the labour share of income, which functions as an overall measure of the balance of forces between capital and labour. That’s been trending downward in a lot of countries for a long time.

When we look at statistics like that, we should think about them as giving us qualitative information. Because when workers are in a poor bargaining position, they give up a lot in terms of the quality of work: break times and safety, and they lose the ability to shape how new technologies are implemented in their workplace.

Under these conditions, workers have had a much harder time fighting to improve the quality of work. As for how to bring struggles over the quantity of work and the quality of work back together, I think that what everyone knows is that a world where workers have more exit power, a world where there was either basic income or much stronger unemployment benefits as well as stronger collective bargaining rights and unionisation — a situation with enhanced exit and voice — would be one in which workers would have a lot more power to shape the quality of work.

There are researchers, mostly in the UK, who have done a very good job thinking about what the qualities are of work that people actually want. We need a positive focus on what — if we had more power, either through economic regeneration or through political institutions and empowered workers — kinds of qualities of work that people would demand.

Part of the reason why that’s so hard is because demands like wage increases are things on which everyone can agree, so unions have historically emphasised wage demands to the detriment of other demands, especially for what used to be called “shopfloor control” (defeats of major union struggles in the early postwar period were also essential here). Money seems to cover lots of the bases of the problems people face — whether that be inadequate maternity leave or childcare or high rent — but there is a complexity of people’s working and non-working lives, there are so many different situations that people are in that point to the wider range of qualitative changes in work that they might desire, and our existing workers’ organisations aren’t really designed to handle that complexity and differentiation.

[.cdw-name]Amelia Horgan[.cdw-name]
How could a democratising work agenda help enhance exit and voice rights?

[.cdw-name]Aaron Benanav[.cdw-name]
Autonomy at work, by which I mean, being in control of your work, is one of the qualities of work that makes people like their work better. Democratising work can come to seem like a big, collective project that is ultimately about collective bargaining at the industry level but we should also be interested in questions of how people as individuals feel empowered to make decisions for themselves. Experiments at self-managed firms point to interesting ways to think about what it means to democratise work, where democracy isn't indexed by the number of meetings you have, but maybe the number of meetings you don't have to have because there's more trust and informal connections between workers to figure out how to solve problems.

 

[.cdw-name]Amelia Horgan[.cdw-name]

How does the planning debate relate to democracy at work? What kind of tools would be needed to plan work better at the firm level and the societal level?  

[.cdw-name]Aaron Benanav[.cdw-name]

There is a really strong connection between the two. A world where people had more exit options via a universal basic income or universal basic services, and where they didn’t need to work, would be a world in which people would need to be motivated differently to do the kinds of work that are necessary for society. What kinds of motivations are available to people to work when they don't have to? People who enjoy work the most generally find the work to be intrinsically rewarding. There is a common misconception that inherently rewarding work should be fun or enjoyable, but a lot of this work is boring and difficult. I believe that the key to making work feel worthwhile has less to do with the content of that work, although that matters too, and more to do with the form in which the work is carried out. If we feel like we have a lot of control over how we do the work, if we get to use skills we have developed, and, crucially, if we feel the work is important, either to us or to other people, then we will find a lot of reasons to work even if we don’t need to work in order to survive. I think that you would find that you can't make work inherently worthwhile in these ways, for the vast majority of people, without democratising work in a very substantial way.

Democratising work would enable people, crucially, to be able to propose changes in how they work. The changes people would propose would of course not only be geared towards making that work more efficient, but also making the work more meaningful, enhancing workers' autonomy, or finding ways that people can work together that don't seem to include as much drudgery or meaningless engagement and effort (workers would also propose ways to transform work to make it more environmentally sustainable or to promote meaningful connections in a wider community). There are real limits to how many changes of this sort can be implemented in a capitalist society. Capitalist societies do make space for some workers to have a greater degree of control over how their work is done, but only in situations in which workers are in very high demand, and when it's very hard to control them. Take, for example, computer programmers in Silicon Valley, or professors in the golden age of the growth of the university, or those very many famous examples for Sweden in the late 80s when they had a very tight labour market and lots of problems with absenteeism.

There are these moments when capitalist firms feel like they're facing real problems retaining workers, especially skilled workers, and then firms focus on improving work quality. But outside of those very rare conditions, capitalist firms generally can't provide and don't provide those kinds of environments. A lot of the reason for that is that those firms have to promote efficiency, above all else — either because of competition or shareholder demands. They don't want workers’ demands for better quality work or safer work to stand in the way of the broadest possible freedom for managers to organise work in the most efficient and hence the most profitable way. 

There are really strong limits to how far anything like democratising work can go in a capitalist society, but even in a capitalist society, we do see exceptional moments, glimpses of what that democratic work could achieve in terms of making work more autonomous and more meaningful to workers and giving them a much greater say over the wider range of ways that work. For work to be transformed, many more criteria would have to be integrated into decisions about investment, that is, about the transformation of the work process, besides efficiency. But that’s very difficult to do in a wide and sustained way in capitalism, which is alarming because it's also destroying the planet.

[.cdw-name]Amelia Horgan[.cdw-name]

It’s a really interesting set of questions — what happens if you remove the wage and what sort of coordination would take place without it — that can’t be answered by wishing away the difficulties and imagining sunny uplands of freedom.

[.cdw-name]Aaron Benanav[.cdw-name]

There are two big pitfalls of the left in the past two hundred years. One is the idea that making work better means making it more playful. I think that's a total dead end. There are ways you can make this concept viable if you expand your definition of play, because “play is also serious”, but that's not what most people mean by play. And I think the other one is where you think democratising work means collective decision-making processes where everyone votes on everything in large meetings with open and endless debates. Of course, that's an element of a democratic work environment. But if that's all it means, then you do end up with the rule of the loudest and longest talkers and the people who are willing to stay at the meaning the longest. In reality, you need to think about how workers with very different concerns and problems and limited resources ultimately can achieve some kind of agreement about what to do that isn't just endless and long meetings. That’s a big part of this book I’m working on.

[.cdw-name]Amelia Horgan[.cdw-name]

Could you tell us about your next book project?

[.cdw-name]Aaron Benanav[.cdw-name]

I'm writing a book about post-scarcity economics. We've let economists tell us what scarcity means. They define it as the confrontation between our limited resources and insatiable human needs and desires. This view leads, I argue, to this very technological vision of how to overcome our problems and what humanity needs. There’s another way to think about scarcity, which is older than the economics definition: a period of scarcity is one in which you don't have enough to meet your needs. Capitalism generalises this experience of insecurity by creating a world where the vast majority of people feel very insecure about their ability to make ends meet because they are dependent on the market and unsure about their ability to get or keep jobs that enable them to survive.

The book is about imagining what a world would look like with an end to scarcity in the second sense, a world where people feel they have this very strong material floor under them such that they never have to worry about their survival. It’s a beautiful idea that's in a lot of science fiction and accounts of futures where we don't have to worry anymore about our survival. 

The book is about how society would have to change to make that vision possible, because in many ways, as people quickly realise, post-scarcity would be incompatible with the way that we currently organise our society. We would have to think about these foundational questions: about what happiness is, what motivates people, and how people might want to work in a world where they are no longer under the whip of economic necessity. The book touches on a lot of these questions about workplace democracy and also how that would have to look at the level of society as a whole. People would want to transform work in ways that are not only about increasing efficiency — and hence expanding our access to goods and services. That is of course crucial, especially in a world where so many people still live in poverty, but it isn’t everything. People would also want to transform work to promote human connection, meaningful activities, increases in free time, ecological sustainability, social justice, and more. 

Those questions have trouble entering into the conversation about how we transform the economy because decision making about the economy has been restricted to a very small number of wealthy people who make investment decisions around their own profitability concerns. That decision-making structure is fundamentally incompatible with a world where people are no longer experiencing scarcity in the sense I’ve outlined. But transforming decision-making structures to involve more people and more investment criteria, besides efficiency, would be extremely complex. It wouldn’t just work out spontaneously, so how would we do it? The book I’m writing offers a proposal that does just that.

 

[.cdw-name]Amelia Horgan[.cdw-name]

How do you think the return of state industrial strategy to the US will pan out? Can it restore manufacturing profitability?

[.cdw-name]Aaron Benanav[.cdw-name]

There are many reasons to be pessimistic about it working out. One is that the US has traditionally had a lot of trouble doing industrial policy, except when it comes to broad-based funding for military and medical advances. Industrial policy requires taking risks, and that means a lot of funded projects will fail. Doing strong industrial policy, therefore, requires a high degree of elite unity, so these failures are not politicised. In the US, such failures are routinely politicised, so parties are generally unwilling to take a risk on doing industrial policy in a concerted way. In addition, there are a lot of reasons to worry that this particular effort at green industrial policy is not really geared to solve the climate crisis because it’s not so scientifically grounded. On the other hand, maybe it will start us down the road towards a green transformation, it could be a first step in that direction. If these policies do produce a temporary economic boom, that will change the terrain on which these struggles for the democratisation of work and a true programme of sustainable development will take place.

It might raise the question of what it would mean to move from the current kind of technocratic model of building a more sustainable future to a democratic model where people have more of a say in what happens, but I remain pessimistic about its chances of success.

 

[.cdw-name]Amelia Horgan[.cdw-name]
Do you have a favourite book or article on or around the history of work or workers’ power or questions of democratising work that you think people do not read enough that they should read more?

[.cdw-name]Aaron Benanav[.cdw-name]

Claus Offe’s “Two Logics of Collective Action”. It explains both the wide range of qualitative demands that workers have and why facing up against the very powerful and unified enemy that is capital, unions and other workers’ organisations are pushed to try to force workers to have a more unified set of proposals, and how that leads to a real simplification in the kinds of demands that workers make. What would it mean for the question of democratising work to not have the external pressure of the unified and unifying enemy in capital? What would it mean to be outside of that?

This is some text inside of a div block.
Date
Authors