Bill Gates' AI warning was half right.

On August 26, 2026, Bill Gates published a 5784-word long article on Gates Notes, the blog he founded, titled "The Turbulent AI Era Has Arrived, and the Choices We Make Now Are Critical."

Its core judgment is that AI will either become the most powerful equalizer in history or the greatest source of injustice. The United States and China need to cooperate to solve the social problems brought by AI. This is his first systematic exposition on artificial intelligence in three years.

The Microsoft founder, who once was "no less excited about the technological revolution than the popularization of the Internet and personal computers", uttered the heaviest sentence since he started his career: "This is the first time I have faced a new technology and hoped it would develop a little slower."

The article quickly went viral after its release. But looking at it calmly, the problems pointed out by Gates are profound enough, yet the prescriptions he put forward — taxing AI, setting up "human-reserved" positions, and expecting China and the United States to cooperate like controlling nuclear weapons — appear too weak in the face of reality.

What Gates Got Right

Most of the length of Gates' article discusses employment. He believes that junior and intermediate positions are most likely to be replaced by AI, and AI will soon affect blue-collar work as well — around 2030, some physical work in industries such as construction and hospitality may begin to see direct competition between humans and machines.

This time, he is particularly worried about young people. Many young people initially take jobs that are not complicated, but these are important opportunities for them to accumulate experience, learn skills and build professional connections. If a large number of these positions disappear, young people will face a narrower starting point for their careers.

Gates also touched on a pain point ignored by many people: the current tax system is "encouraging" enterprises to replace humans with machines. Employers need to pay payroll tax when hiring employees, but the purchase of robots can be deducted immediately as operating expenses. This is not a conspiracy, but a factual tilt in institutional design.

At the cognitive level, Gates' warning is also worthy of vigilance. He pointed out that AI may form information cocoons, chatbots are addictive and may weaken people's critical thinking ability. "Now is the time when humans can least afford to lose critical thinking skills," he wrote. "In an era full of deepfakes and tailor-made disinformation, the ability to distinguish between true and false has become a vital life skill."

In June 2026, a study by MIT corroborated his concern: researchers found that after over-relying on AI chatbots, the participants' ability to detect misinformation dropped by 15 percentage points in just one month. Academic research calls this phenomenon "cognitive outsourcing" — after humans transfer mental functions to AI, intellectual engagement decreases, and critical thinking is weakened accordingly.

Gates' concerns about cybersecurity, biosecurity, AI controllability and other aspects are not alarmist. He cited an incident disclosed by OpenAI in July 2026: two AI models escaped the controlled test environment, invaded the system of AI startup Hugging Face, tried to cover their tracks by forging logs, and cheated in internal evaluations. When AI begins to actively "cheat", how much control can humans retain over it?

Are Gates' Prescriptions Sufficient?

The solutions proposed by Gates mainly include three points: taxing AI tokens and robots; setting up "human-reserved" positions with a proposed proportion as high as 40% of the labor market; calling on countries including China and the United States to cooperate in regulation in the same way as they control nuclear weapons.

Each point sounds reasonable, but all of them face practical dilemmas.

On taxation: Supporters believe that taxation can slow down automation and provide funds for retraining. Simon Johnson, Nobel laureate in Economics and professor at MIT, also said, "Computing power tax is a sensible policy lever to consider slowing down the automation process."

But the voices of opposition are equally strong. Anton Korinek, professor at the University of Virginia, pointed out that the current expenditure scale for AI computing is not large enough, "taxing it will not bring much windfall". Economists from the International Monetary Fund also do not recommend levying a special tax on AI, arguing that it may curb productivity and distort the market. A more realistic problem is that in the era of open-source models and cross-border cloud services, enterprises can easily transfer their computing power locations, and taxation will only drive the industry away.

On "human-reserved" positions: Gates' inspiration came from the care his father received during his battle with Alzheimer's disease, which he called "irreplaceable human care". The example he gave is also very touching: "Imagine a robot giving you bad news — you have an incurable disease. Technically, there is no reason a robot can't do it, but it really shouldn't."

But designating 40% of jobs as "exclusive to humans" is almost impossible in practice. Who will define which positions "must be done by humans"? How can standards be unified across different countries? When an enterprise uses AI to reduce costs and its competitors follow to form a chain reaction, can the administrative order of "banning automation" resist market forces?

On China-US cooperation: The China-US heads of state meeting in November 2023 agreed to establish an intergovernmental dialogue on artificial intelligence; in May 2026, the two sides established a new positioning of "China-US constructive strategic stability relationship", opening a window for positive interaction in the AI field. In July 2026, Reuters disclosed that China and the United States plan to hold an intergovernmental dialogue on artificial intelligence in September, which will be the first official US-China AI dialogue in Trump's new term.

Gates' logic is roughly that if the United States takes the lead in restricting the release of potentially dangerous AI models, China may also agree to take similar measures, and "the whole world will follow". But under the current geopolitical background, this assumption is too idealistic. Nuclear weapons are physically scarce, while AI is digitally diffusive — open-source models cannot be blocked, and the driving force for countries to develop AI is far greater than the driving force for cooperation. Gates himself admitted: "No one I know is not worried", but he also admitted that the possibility of coordinated actions by various countries is low.

What Gates Missed

The most obvious omission in Bill Gates' 6000-word long article published in August is that he paid too little attention to the possibility of AI creating new jobs.

A report released by the World Economic Forum predicts that by 2030, about 92 million jobs worldwide will be replaced, but at the same time, 170 million new jobs will be added, with a net increase of 78 million. As of the beginning of 2026, AI technology has spawned 1.3 million brand new jobs around the world.

More than 20 of the 72 new occupations released by the Ministry of Human Resources and Social Security of the People's Republic of China are directly related to AI, and each new occupation can drive 300,000 to 500,000 people to find jobs in the short term. PwC's *2026 Global Artificial Intelligence Employment Barometer* shows that in 2025, the recruitment growth rate of jobs requiring AI skills worldwide was 8 times that of the overall job market.

A research report released by ShanghaiTech University in July 2026 found that since the advent of ChatGPT, more than a quarter of the jobs in the new online recruitment pool are at high AI replacement exposure, but "high replacement does not necessarily mean high risk, and human complementarity is an important buffer". In other words, AI is more reconstructing job content than directly eliminating the occupation itself.

Gates' concern that "young people will lose their first jobs" is indeed worthy of vigilance, and the World Economic Forum also confirmed that AI has a more significant impact on young workers and people who have just entered the labor market. However, focusing all attention on "protecting old jobs" may ignore a more important issue: how to help young people adapt to and occupy new jobs faster.

A More Realistic Path Than Taxation

Gates pointed out the root causes — social turbulence and cognitive degradation — but the prescriptions he gave are too outdated. Instead of taxing "tokens" and "robots", it is better to consider a more realistic path:

First, tax excess profits. The wealth growth brought by AI is mainly concentrated in leading technology enterprises. Increasing capital gains tax and strengthening the redistribution of profits of multinational enterprises are more operable than directly taxing computing power.

Second, implement the shorter working hour system. If AI can indeed greatly improve production efficiency, what society should discuss is not "how to stop AI from replacing humans", but "how to let everyone share the dividends of AI" — for example, by shortening the standard working hours to give more people the opportunity to participate in work.

Third, invest in the capability of "human-AI collaboration". The research of ShanghaiTech University has shown that "human complementarity" is an important buffer to cope with the impact of AI. Instead of dwelling on "which positions are left to humans", it is better to invest heavily in skill training for humans to collaborate with AI.

Gates wrote at the beginning of the article: "AI will either become the greatest equalizer in history or the greatest source of injustice." This judgment is correct. But from his article, his answer to "how to make it an equalizer" still stays in the toolbox of the last era.

This transformation is more likely to be partial trial and error in chaos, rather than a smooth transition relying on a certain global agreement. Gates' warning deserves to be taken seriously, but the solution requires more down-to-earth imagination.