U.S. AI-labor protests could eventually resemble the French Yellow-Vest protests

When millions recently took to the streets over economic insecurity.

Left: AP Photo of French Yellow-Vest protests; Right: AI-generated protest photo

I do not yet have a well-formed opinion on the net impact of AI on job loss across different time scales. Will it be a large, net negative, or will it be close to net neutral, similar to previous technology cycles? “Expert” estimates are all over the place, and while there is little job loss directly attributable to AI right now, circumstantial evidence is accumulating.

That said, I believe it is clear that even in a net-positive scenario, many jobs will be displaced as new ones are created. For example, the World Economic Forum predicts by 2030 the creation of 170 million new jobs worldwide.

However, this growth is expected to be offset by the displacement of the equivalent of 8% (or 92 million) of current jobs, resulting in net growth of 7% of total employment, or 78 million jobs.

Supposing that this is true and a similar scenario unfolds in the U.S., I still think this amount of displacement is likely to have significant negative consequences for the displaced. That is, the displaced individuals are unlikely to be the same individuals who secure the latest jobs, and this will leave many with worse jobs, or in a place with no job at all. For example, are cashiers and truck drivers going to get new, fancy AI jobs in another industry, without significant help to do so? I highly doubt it.

So, recognizing that significant job displacement is on the horizon for many industries, regardless of how the overall total nets out, it would be ideal to provide affected individuals with a softer landing than we’ve managed to do in the past. Unfortunately, the likely policy outcome is that we do absolutely nothing until there is a considerable backlash to AI, and even then, we probably do nothing. The backlash would have to be big enough that politicians cannot ignore it, or it is seen as a political opportunity, such that it becomes a platform for elections.

In a recent post titled Will the AI backlash spill into the streets?, I made the case that, among the numerous anticipated AI harms to society, job displacement stands alone with the potential to significantly spill over into the streets in the form of sustained protests. I’m not saying that it is likely to happen, and I hope it doesn’t, but I also think we could increase the probability of preventing it if we had a clearer picture of what we’re trying to avoid. At the end of that last post, I pondered about historical analogs:

Recent protest movements seem more one-sided politically (e.g., climate change, Occupy Wall Street, Tea Party, etc.). Mid-century protests were arguably similar (civil rights movement, Vietnam War protests, etc.), though they were sustained for much longer and ultimately swayed public opinion and accelerated change. A better parallel here, though, would be something that is clearly bipartisan from the start, more squarely on an economic issue, and resulted in swift reforms.

This line of reasoning led me to the Yellow Vests protests in France from 2018 to 2020, which I’ve come to believe could provide a decent historical analog of what could happen if AI job displacement reaches a critical mass in the coming years. It’s recent, directly related to economic insecurity, and created policy changes, though significant disruption and violence as well. Ideally, we’d get the policy changes without the widespread disruption and violence.

Now I am not French, and I did not personally live through this protest movement. I do remember watching and reading coverage about it at the time, but that was the extent of my knowledge before I delved more deeply into the subject over the past week. (Below, facts are from Wikipedia unless otherwise noted.)

What were the yellow vest protests about?

They began as a reaction to a fuel tax increase, but quickly expanded to a broader reaction against economic insecurity. The movement never had clear leaders, nor was it tied to a particular political party. One widely circulated list of 42 demands compiled from online surveys went viral, including many specific economic demands such as rolling back the fuel tax, implementing minimum pensions, indexing wages to inflation, and providing jobs for the unemployed. It was somewhat disjointed, as it also demanded education and immigration reforms, among others; however, the central theme centered on economics and jobs.

How many people participated?

Approximately 3 million unique protesters participated, which is roughly 4% of France’s population of nearly 69 million. The U.S. has a population of around 341 million, so an equivalent number of unique protesters would be approximately 14 million. That’s the same order of magnitude as the George Floyd protests.

The peak protest time was Nov-Dec 2018, with the highest single-day attendance of about 287,000; the U.S. equivalent would be about 1.5 million. Protests occurred across the entire country. After the initial period, there were still continual weekend demonstrations for about a year and a half in total, until the pandemic essentially brought them to an end.

Did the protests turn violent?

Unfortunately, yes. At least a dozen people died in the protests, with another five people losing their hands as a result of police grenades, and a reported twenty-three people losing their eyesight. More minor injuries were in the 1,000s for both protestors and police as a result of clashes.

How much public support did it have?

Very high. Public support for the movement reached a high of 75% in the initial phase, and then it declined over time.

What was the government’s reaction?

On December 10, 2018, about a month in, French President Macron gave a speech to the public, pledging a €100 increase in the monthly minimum wage, among other reforms. The speech was viewed live by more than 23 million people, or approximately a third of the entire population. The U.S. equivalent would be around 80-85 million. Concessions from the government ultimately totaled about €17 billion, which, converted to USD and scaled up to the U.S. economy, would be about $150 billion.

What was the demographic makeup of the protestors?

An academic study of the protesters found that approximately 47% were first-time protesters. The median income of the protesters was about 30% less than the country’s median income. Participation cut across the political spectrum, and the researchers concluded:

In short, this is indeed a revolt of the ‘people’…in the sense of the working class and the lower-middle class, people on modest incomes. Consequently, in several ways the gilets jaunes movement presents a different kind of challenge from the social movements of recent decades. In addition to its size, the strong presence of employees, people of modest educational qualifications and first-time demonstrators, and, above all, the diversity of their relationship to politics and their declared party preferences, have made roundabouts and tollbooths meeting places for a France that is not used to taking over public spaces and speaking out, as well as places for the exchange of ideas and the construction of collectives in forms rarely seen in previous mobilizations.

Why did they wear yellow vests?

As noted, the movement originated as a response to a proposed fuel tax increase. Independently, a different French law requires motorists to have a yellow vest in their car to wear in case of an emergency, so many motorists had them readily available. A petition against the tax went viral, and then some associated viral videos called to “block all roads” and included the idea of using the yellow vests.

What are some parallels to AI?

First, while the backlash to AI is just getting started, as I noted in my previous post, the ingredients are there for a potential future movement that could match a similar broad-based revolt that transcends political parties, if significant negative job impacts accumulate over several years:

1. Cuts span industries, so outrage lands on both parties.

2. Every income bracket—from cashiers to coders—takes a hit.

3. Sudden, deep job cuts risk recession and years of high unemployment.

To be clear, we’re not there yet, and may never get there. Job losses may not materialize. Or, they may unfold over a much longer period of time that doesn’t lend itself to banding together across industries. For example, current AI labor organizing remains limited to specific sectors like entertainment and dockworking. But, if AI-driven job displacement accelerates across multiple industries simultaneously—affecting many millions of workers over the next 3-5 years—the conditions could ripen for a Yellow Vest-style eruption.

Second, as the Yellow Vest movement showed, it doesn’t actually have to be sparked by a sharp increase in job losses. Instead, if there is enough downward pressure on wages, unrest resulting from that pressure can build up until a critical mass is reached. In other words, that could still happen even if AI diffusion takes many years to touch many industries, as long as the impacts and resentment are sustained.

Third, if a critical mass is reached, that’s essentially a powder keg waiting to explode, which means any event could be the proximate cause that sparks it. Therefore, like the Yellow Vests movement, I could see a similar AI movement happening in a decentralized fashion. That is, one that begins with online viral calls to action, which then spills out into the streets.

What’s different?

One difference is that the Yellow-Vest movement was primarily rooted in the lower income brackets, whereas AI has the potential to draw in affected people from across the income brackets, as noted above.

Another difference is the Yellow-Vest spark occurred in response to an immediate economic pain from the fuel tax increase. It’s not clear exactly what that equivalent would be for AI, though I suspect if job displacement is vast enough, some match to light the powder keg will naturally emerge, like a layoff announcement.

And then there is the France of it all. France already has a much greater social safety net, and obviously a different history than the U.S., including with protests. For example, their unique system may increase the likelihood that they will take action in response to protests.

What could be the symbol equivalent of the yellow vest?

I don’t know, but I think that worked because it was readily available, so I could see everyone wearing a specific color, like black, as pictured above.

Could it similarly turn violent?

I hope not, but I can’t see why that wouldn’t be the case, especially in the U.S., considering historical violence in similarly-sized past protests, most recently the George Floyd protests. Add to that the fact that we have way more guns and that immigration protests are turning violent right now.

How can we avoid this?

To me, that’s the key question. Ideally, we can prevent violent protests by enacting necessary reforms ahead of time (soonish), rather than waiting until mass protests occur. That generally doesn’t happen for a couple of reasons. First, we tend to do things only under immense pressure. Second, the magnitude of the reforms to be enacted is not yet apparent. However, I’d like to work towards changing that, both to increase pressure now (as in this post) and to propose a list of sensible reforms. If you have particular ideas for that, I’d love to hear them, such as enhanced unemployment benefits, vouchers for private training programs, relocation assistance, etc. There are already numerous ideas circulating, and I plan to explore them soon.