My new book #WhatHappenedToLiberalDemocracy is out on August 11. Here is a brief message about my motivation for writing this book and its central thesis. Pre-order links are available here: shapingwork.mit.edu/what-hap…. I hope you will join me in the discussion.

An excellent analysis of where Europe is right now. It’s Not Just France. Europe’s Political Center Is Crumbling. nytimes.com/2026/10/20/world… via @NYTimes

A badge of honor

This is interesting.

Thanks to @joshgans for collecting my questions on AI and his responses in a single substack post.

This would be funny— if it weren’t our tragic reality

Another thoughtful response from @joshgans

As usual, @Yoshua_Bengio is raising central issues and speaking with great moral and intellectual clarity. Getting the direction of AI right is a responsibility for all of us.

Seventh and last question on AI. Can AI really help us create as many jobs as it will replace? The AI industry offers two narratives at once. The first emphasizes that because of AI’s capabilities, there will be a lot of job loss, and soon. Anthropic’s co-founder and CEO Dario Amodei is particularly vocal on this: cnbc.com/2026/01/27/dario-am… The second claims that AI will also enable lots of new jobs, so we should not fear joblessness, higher unemployment, and humans being sidelined (see, for example, Jensen Huang, fortune.com/2026/09/01/nvidi… or Marc Andreessen, digitaltoday.co.kr/en/view/4…). A lot hinges on this question. Sizable job displacement is costly for workers and could destabilize our already ailing democracy. Displacements from robots and cheap Chinese imports in the 1990s and 2000s in the United States, which were smaller than the projected job losses from AI, contributed significantly to declining trust in government and institutions (see my book What Happened to Liberal Democracy, penguinrandomhouse.com/books…, on this). Our democracy may not be able to survive another massive job dislocation. I am worried about job loss, for two fundamental reasons. AI is currently being developed as an automation technology and in addition, I do not find the arguments about large numbers of jobs being created by AI particularly realistic. (I’ll come back below to how AI could create jobs if it weren’t developed as an automation technology). Proponents of the job creation side fall back on four general arguments, none of which is convincing in my assessment. First, new technologies have always created lots of jobs. True, but that does not guarantee that this time will be the same. And more importantly, this account neglects the richer details about different historical episodes. In a nutshell, automation technologies by themselves need not create jobs (though they can, depending on circumstances and how productive they are). They involve capital substituting for tasks previously performed by workers. The way they would create jobs is that the displaced workers would go and take non-automated jobs in other parts of the economy. This can happen, but it also depresses wages (or at the very least the labor share), and consequently, some of the workers may be discouraged from looking for work or may not have the skills that remaining non-automated tasks require. This, in any case, is what task-based models predict. See, for example: aeaweb.org/articles?id=10.12… and onlinelibrary.wiley.com/doi/…. The empirical evidence also casts doubt on plentiful job creation taking place automatically. For example, the introduction of robots led to many fewer jobs in local economies that depended on blue-collar jobs in robot-heavy industries (and this effect was quantitatively large; one more robot per thousand industrial workers reduced the employment population ratio in the local economy by about 2 percentage points and local wages by 0.42% between 1990 and 2007). See journals.uchicago.edu/doi/fu…. Second, AI will increase productivity and higher productivity will lead to more jobs. But this argument suffers from the same weakness as the previous one: higher productivity will trigger job creation only in remaining non-automated tasks (unless there are new tasks, which I discuss below). If wages are depressed and workers don’t have the skills for these non-automated tasks, even with higher productivity, there will not be AI-driven job creation. Third, AI will need a lot of supporting roles, such as people to train the AI models. This one is the weakest argument of all. If AI is indeed an automation technology, then the whole point is to save labor costs. By the same logic, the labor needed to train AI must be much less than the labor AI ultimately saves. For this reason, I see essentially no chance that there training and maintaining AI will create anywhere near as many jobs as AI replaces. Fourth, AI is already complementing workers. This argument is based on some evidence that many workers are using AI in their jobs – see, for example, thefp.com/p/ai-jobs-unemploy…. However, the mere fact that workers are using AI in their jobs doesn’t establish that new jobs will be plentiful. The theory here is again quite clear (see the detailed treatment of this issue here, nber.org/papers/w32872). If AI simply makes workers more productive in the tasks that they used to perform already, it will drive down task prices. If you are the only worker using AI, this will benefit you. If all workers like you start using AI, then the skills on which you depend become commodified. This is what happened, for example, to London black cab drivers, whose highly-valuable knowledge of London streets was commodified when GPS technology enabled minicabs, and then later Uber drivers, to perform the same tasks as effectively. To put it simply, it isn’t enough for AI to make workers more productive at the tasks that they are performing already. This brings me to how AI could create a lot of jobs. Again, the task-based framework provides the answer here: new technologies create jobs, and in fact well-paying jobs, if they enable new tasks (either in the same industry or in new industries). I have argued elsewhere that AI has this potential as a powerful information technology. See, for example, brookings.edu/articles/build…. But this isn’t currently the direction in which AI is being developed. Hence, on its current trajectory I remain quite concerned about future job losses (and wage declines) from AI. Of course, things may turn out differently. But if they do, it will be because AI becomes more pro-worker. I see this as unlikely, but it is useful to still consider the scenarios that would bring such an outcome. These are: (1) current AI technologies may become turn out pro-worker than I am anticipating; (2) in the future the industry will automatically and by itself change and focus on pro-worker applications; (3) based on AI-driven science and research, new sectors will be created where there will be need for new workers. I’m not optimistic about any of these possibilities, and therefore my conclusion is that AI is likely to lead to more jobs being lost than gained. (But I do think there are policies that would encourage pro-worker AI; see, for example, nature.com/articles/d41586-0…). That being said, I don’t think the job loss will happen very quickly, because as I discussed in an earlier post (third question), the diffusion of AI is relatively slow. But this also implies the productivity gains from AI will be slow in coming.

Sixth question on AI. Does AI need to be a zero-sum race between the US and China? Many influential voices in tech as well as the current administration argue that it does. (See, for example, darioamodei.com/post/on-deep…, washingtonpost.com/opinions/…, forourposterity.com/situatio… and this from the White House, whitehouse.gov/wp-content/up…). There are many aspects of these arguments, but I would like to focus on three that have received less attention than they deserve. The first aspect is that this framing typically assumes that there is a single direction of progress for AI, and the US and China are racing along this dimension toward superintelligence or at the very least some sort of powerful AI that will confer huge military and strategic advantages to the winner. If, on the other hand, the US and China are developing these technologies in different directions, there is much more room for coordination and even for each country learning from the other’s best practices. My reading of the evidence is that there are huge differences in the American and Chinese approaches. The US is focused on foundation models, such as frontier large language models. China is also investing – and arguably catching up – in these models, but there is much greater emphasis in China on applications of AI (with the government setting targets for the spread of AI across industries, english.www.gov.cn/policies/…). This reflects several factors, in my assessment. First, Chinese AI investments started with government demands for surveillance technologies and government-supplied data to AI companies (e.g., academic.oup.com/restud/arti…). Second, China is facing a huge demographic transformation, which will create scarcity of labor in the near future. This encourages companies to integrate robotics and AI much more actively into the production process (and China accounts for 59% of all robot installations in the world, ifr.org/ifr-press-releases/n…). Third, applications and integration of these technologies into the production process requires a lot of engineering. China has a large and very well-trained supply of engineers (whereas the US has far fewer engineers and a lot of engineering talent is currently going to Silicon Valley, which hasn’t focused on applications). Fourth, China was relatively late in catching the “artificial general intelligence,” AGI, bug, although now there is more emphasis on this in China as well. Whatever the exact causes, this difference in approach implies that there is plenty for the US to learn from China as well as the other way around. The second aspect is that the zero-sum framing ignores or downplays the crucial need for global coordination. This framing claims that the strategic and military advantage that powerful AI will create is all that matters. But the race to AGI also poses huge risks (and this is true regardless of whether this race is truly advancing toward superintelligence or just producing distorted intelligence, as I argued in my first post in this series). And these risks are global. If the US and China are racing toward AGI with no guardrails and coordination, even the follower (probably China) will have models capable of creating havoc. In this case, racing without any global coordination is a recipe for disaster. The third aspect of the problem is that this framing narrows the scope for any kind of coordination on other global issues between the US and China – including climate change, pandemics, nuclear nonproliferation and world peace. Tellingly, climate cooperation between the US and China has now largely stalled). Let me add that I’m saying all of this not as an admirer of the Chinese system. On the contrary, I have been a vocal critic of both the Chinese Communist Party and the Chinese development strategy since Why Nations Fail (penguinrandomhouse.com/books…) and I have argued that the Chinese success has some of its roots in severe labor repression (project-syndicate.org/commen…). Indeed, this argument does not ignore the fact that China has an authoritarian, repressive regime, and it is using AI for surveillance and to stamp out dissent. Global coordination and efforts to set up reasonable rules that apply to all parties do not require the counterparty to be fully democratic or ruled by angels. In fact, sometimes it is more important to set up such a global framework when dealing with dangerous adversaries (as the arms-control treaties that the US and the Soviets signed demonstrate). In sum, my assessment is that it is a colossal mistake to frame the AI race as a zero-sum one between the US and China. It poisons the much-needed dialogue and coordination between the US and China on broader issues; it makes the AI race more reckless; and it discourages the building of reasonable guardrails and regulations. The last point is particularly relevant, and makes me sometimes say: if China did not exist, Silicon Valley would have to invent it.

And this

Dear followers, I’m delighted to share this interview with 60 minutes on AI and jobs. Thank you

And please see this somewhat more optimistic take on my third question from @joshgans

I think this is well worth reading.

Reactions on my second question from @joshgans

Fifth question on AI. AI is currently viewed as a centralizing technology – with ever-larger models, trained on much of the world’s knowledge, running on massive, concentrated compute, overseen by a few companies, and aiming to control and deploy all of that knowledge But many early computer scientists and pioneers thought of digital technologies as powerfully decentralizing, as Simon Johnson and I discuss in our book Power and Progress (hachettebookgroup.com/titles…). One such pioneer, Lee Felsenstein, wrote “The industrial approach is grim and doesn’t work: the design motto is Designed by Geniuses for Use by Idiots.” In the words of one of his collaborators, Bob Marsh, “We wanted to make the microcomputer accessible to all human beings.” And this wasn’t just an aspiration; they believed it was both feasible and natural that computers would empower individuals. The first newsletter of People’s Computer Company also captures this optimism: “Computers are mostly; used against people instead of for people; used to control people instead of to free them; time to change all that – we need a…PEOPLE’S COMPUTER COMPANY”. Should we accept centralizing AI? Or is there a way in which the promise of artificial intelligence can be realized while data, capabilities and control are much more decentralized both across domains and across people? Wouldn’t decentralized AI be more consistent with human agency and control? (Some of the ways in which greater decentralization can happen are discussed in my first post in this series.)

Fourth question on AI. One of the great promises of AI is the discovery of new drugs and cures, so that we can live longer and healthier lives. If and when this becomes a reality, it would indeed be a big achievement. These potential health benefits are often invoked to justify current investments (and lack of regulations). Some even argue that slowing down AI would harm humanity by delaying these benefits. (See a16z.com/the-techno-optimist…). This raises another uncomfortable question: can we justify rapid and large AI investments based on health benefits? Here is why I think this is an uncomfortable question. First, despite significant effort, AI has so far produced few gains in drug discovery or new cures. A fascinating new paper by Ryan Hill and Carolyn Stein documents significant (downstream) scientific work on proteins whose structures AlphaFold predicted. But the authors conclude “we find no evidence so far that more applied, early-stage drug development is targeting these proteins.” (See their paper here papers.ssrn.com/sol3/papers.…). This thought-provoking post by Daphne Koller explains the difficulties that AI is currently facing in drug discovery. Koller emphasizes that AI-based research is focused on the last layer of drug development (creating new molecular entities for already well-understood therapeutic modalities), rather than targeting new disease mechanisms (see deepphenotype.substack.com/p…). Understanding disease mechanisms is harder because there is much less data on it, and this endeavor requires more innovative approaches. It remains an open question whether current AI models are capable of doing this. Second and more importantly, if you want to improve life expectancy and health in the United States, there is plenty of low hanging fruit and diverting some of the huge investments going to AI for this purpose would likely do much more for health outcomes. The United States has the lowest life expectancy at birth among large rich economies. Americans live about five years less, in expectation, than the citizens of Switzerland, Sweden, Japan, Italy, South Korea and several other industrialized countries, and most of the gap comes from deaths before age of 70 due to chronic diseases, overdoses and other preventable causes (see, for example, healthsystemtracker.org/char…). These largely reflect failures in US public health: preventable problems, unhealthy diets, insufficient immunization rates, poor access to primary care; and poor information about health. The country spends much more than other peer economies on healthcare, but too little and too ineffectively on public health (here is a reference to my own work on this: aeaweb.org/articles?id=10.12…). Could we, and should we, divert some of the massive investments in AI towards public health if what we want is better health and longer lives for Americans? (And yes, new drugs will benefit citizens of other countries as well, but they can also invest more in public health and the bulk of global AI spending is currently in the United States).