From the book

The question “Is AI good or bad?” is too small, too clean, and far too useful to the people selling the answer.

Good or bad is the language of product reviews, not of political economy. It asks us to judge the tool while leaving the system around the tool politely and very conveniently underexamined. It invites a familiar performance: one side points at productivity, innovation, medical promise, accessibility, creativity, and convenience. The other side points at extraction, exploitation, surveillance, environmental cost, monopoly power, and harm. As a result, everyone gathers evidence and becomes more certain, but the machine keeps running as is: The show must go on.

The problem is not that the question has no answer, but that it is designed to avoid the harder ones. We know that every question or problem statement contains a larger issue that must be addressed to reach the root cause. The bigger questions in this case are:

Who owns the infrastructure? Who supplied the data? Whose labor was hidden? Whose work was absorbed? Whose body became material? Whose community absorbs the water use, energy demand, mining, waste, or land pressure? Whose job becomes measurable? Whose expertise becomes capturable? Whose life becomes a risk score, a summary, a flagged case, a denied service, a suspicious pattern, a training example? Who gets relief at the interface, and who becomes part of the boiler room that makes the interface feel clean?

The public debate often behaves as if we are standing in front of a new machine and deciding, calmly, whether to turn it on. In fact, we are not. We are standing inside a world already organized by ownership, class, empire, patriarchy, race, borders, debt, surveillance, extractive supply chains, broken public services, and work that has been made unbearable for too many people. AI did not arrive into a neutral world. It arrived into this one. That matters because technologies do not only do what their designers imagine. They do what institutions, markets, states, employers, platforms, and desperate people find useful. A tool built inside domination will not automatically become liberation just because the demo was inspiring.

The bargain we are offered is simple enough: accept the machine as progress, and manage your discomfort privately. Use the tools to stay relevant, be responsible, learn the new skills, do not fall behind. Most importantly: Do not be difficult. But also: Do not be naive. The future is coming, and serious people adapt.

That bargain is everywhere now. It appears in workplace training, product launches, conference keynotes, investment decks, policy language, public sector modernization plans, university strategy documents, and the cheerful little prompts that invite us to let the system help. It is rarely stated bluntly because that would spoil the mood. But the shape is clear and unfortunately, oh-so familiar: The infrastructure is privately owned and costs are socially distributed, the benefits are uneven and risks move downward, all while the guilt moves toward the user, and profits move upward.

There is a specific kind of exhaustion that comes from seeing the contradiction and still living inside it. Most people using AI are not fools. They know the sales pitch is too clean. They know the technology is unreliable, overhyped, and often sold by people who confuse market valuation with moral authority. They also know the tools sometimes help. They use them to draft, translate, summarize, explain, code, organize, search, plan, and get through work that was already too much. They feel the relief and still distrust the system producing it. They benefit from the interface and suspect the infrastructure. They understand that refusing individually will probably not weaken the machine, and participating does not require believing the machine’s mythology.

That position has a name: Ambivalence. Not indecision, nor a polite middle between two equally reasonable sides. Ambivalence means holding a contradiction long enough to stop it being laundered into a simpler story. It means refusing the emotional convenience of purity, when purity is not available, and refusing the professional convenience of enthusiasm when enthusiasm is demanded as proof of competence. It means saying: yes, the tool can help, and yes, the system around it is dangerous. Both claims are true at the very same time. Neither cancels the other.

Corporate hype offers one exit: stop worrying and optimize. The system is inevitable, the market has spoken, the tools are improving, the serious people are already moving, and hesitation is just fear that one needs to overcome. This exit flatters the ambitious and disciplines the doubtful. It turns political questions into adoption problems. It tells workers to upskill, institutions to modernize, and everyone else to trust that governance will arrive shortly after the next funding round.

Moral purity offers the other exit: refuse everything, stand outside, remain clean. This exit has emotional appeal, especially when the harms are real and obscene. But most people do not live outside. They work inside institutions, depend on platforms, serve clients, teach students, care for families, write code, manage workloads, navigate bureaucracy, and try to survive an economy that rarely rewards principled isolation. A politics that only offers purity will lose many of the people who need language for the contradiction they are already living.