For years Bill Gates argued that AI would free people from labour and speed the discoveries that fight disease and climate change. He now expects widespread and permanent unemployment and serious social upheaval, with white-collar and blue-collar work alike at risk.

He makes the case in a nearly 6,000-word essay published Wednesday, his first extensive writing on the topic in three years. Humanity is nowhere near ready for what AI will deliver, he writes, and nothing is being done about it.

He proposes taxing AI tokens and robots, both to discourage companies from replacing workers and to fund social safety nets, and he wants certain jobs reserved for people alone — human reserved, he calls it, on the model of land that could be built on and is left alone because what would be lost is worth more.

The scale of what is coming is underestimated, he argues, partly because comparisons to earlier technologies such as the PC mislead, partly because the failures visible in today's models get fixed quickly. This shift, he writes, is genuinely different from the ones before it.

For the first time in his life he finds himself wanting a technology to move ahead more slowly: if a credible plan existed to slow AI's advance globally, he would likely support it.

No existing institution is built to govern this, he argues: AI cuts across taxes and employment, health, schooling and national security at once, and no conventional body can take in the whole picture. He wants a new global organization for risks that spill across borders, modelled loosely on the nuclear accords — without saying which risks he means, how it would be built, or how the United States and China, whose cooperation he concedes it needs, would be brought to the table.

Much of it is familiar ground, and it is a bleak turn from the optimism of his 2023 writing on AI — though he maintains it is not too late for the benefits to outweigh the harms.

After a conspicuous silence, the essay amounts to a bid to win back a leading part in steering where AI goes. Gates says he will raise it with legislators and world leaders wherever he can, and his team is trying to arrange a November sit-down with Chinese president Xi Jinping on reducing AI's dangers — the first since Xi received him three years ago as the first foreign businessman after the pandemic.

Cover image: “Authorized personnel only sign hanging on chains” by Dave Meckler, Unsplash, Unsplash License, resized, re-encoded.