Bill Gates Says There Is No Plan for the AI Transition, and Offers Three
In a new essay, the man who spent fifty years arguing that innovation works out is worried. He proposes new institutions, a category of jobs reserved for humans, and a tax on AI and robots.
Bill Gates published a long essay on Wednesday arguing that the transition into the AI era will be one of the most turbulent periods in human history, that the world has no plan for it, and that he sees no evidence leaders are treating that as urgent. Coming from someone who has spent five decades making the case that technology tends to make life better, the tone is the story.
Gates lists what he thinks is coming: whole categories of jobs erased, fraud and deepfakes supercharged, a lower bar for cyberattacks on critical infrastructure, an easier path to engineering a new disease, governments able to kill without a human in the decision, and a fundamental change in how children grow up. He says that if someone produced a credible plan to slow AI globally he would probably back it, but he does not expect one, because the geopolitical and economic pressure is too strong.
His three proposals. First, new institutions: no existing agency was designed for a technology that touches jobs, security, health, energy and elections at once, so he wants national coordinating bodies plus an international organisation modelled on nuclear weapons inspections, aviation rules and the ozone treaties. Second, what he calls “Human Reserved”, a set of jobs machines could do but that we choose to keep for people, on the model of a nature reserve. His example is a robot telling you that you have an incurable disease: no technical reason it could not, and it still should not. Third, a tax on AI usage and robots, an idea he first floated nine years ago and that was widely dismissed then. His argument is about incentives: a company that hires a person pays payroll tax, a company that buys a robot deducts the cost.
What’s actually going on here: Gates says the essay came out of noticing that the industry is crossing, one after another, the exact thresholds it once named as places it would stop and think. Bioweapon uplift, cyberattack uplift, emotional dependence, mass job loss, loss of control. “We’re in the process of crossing every single one of those thresholds,” he told GeekWire, adding that nobody wants to brake first. He is blunt about the current US framework, under which companies are asked to submit powerful models for government testing but which explicitly bars any licensing requirement: “What is the threshold that’s being examined, and what is the action taken when you cross that threshold? The whole thing seems so empty to me.”
What this means for you: this is an opinion piece by an influential person, not a policy change, and reasonable people disagree with all three ideas. A robot tax is inefficient by design, Gates concedes that; who decides what stays “Human Reserved” is unanswered, and he says so. Read it as a signal about where the argument is heading rather than a forecast. The practical takeaway for anyone working today is the least dramatic part of it: Gates expects job losses to hit rich countries first, and expects AI to do more good than harm in places short of doctors and teachers. If you want to know what to do this week, the honest answer is nothing. This is a conversation about the next decade, and the useful move is to have an informed opinion in it.
Sources
Source: https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make
Z.ai's GLM-5.3-Flash Gets Near Opus at a Tenth of the Price
A 320-billion-parameter open-weight model with a million-token context window, released under an MIT licence at $0.15 per million input tokens. It spent a week on leaderboards under a fake name first.
