Microsoft co-founder Bill Gates wrote in a nearly 6,000-word essay that even in the best case, the transition to the AI era will be "one of the most tumultuous periods in human history." He believes that there are currently no plans in place to buffer this transformation.

Gates stated that AI will reshape employment, government operations, and interpersonal relationships within the next decade, impacting almost all fields including healthcare and education. White-collar jobs are already experiencing initial shocks, particularly in customer service, sales, software engineering, and legal assistance. By the end of this decade, intelligent robots will compete directly with human workers in hotel services and construction. He emphasized that this employment shock will be completed within ten years, unlike previous transitions that spanned several generations, leaving society without a generational buffer.

He expressed particular concern for young people, believing they will enter a job market with significantly fewer entry-level positions. He also warned that AI will make it easier to access and utilize information related to weapons and cyberattacks that previously existed online, exacerbating fraud, deepfakes, misinformation, and cyberattacks, allowing criminals to target individuals, businesses, and governments more precisely. He is also worried that some governments may misuse powerful AI systems in dangerous ways.

On the policy front, Gates proposed three specific measures. First, to tax tokens used by AI companies, leveraging the fact that AI companies already measure and bill for tokens. Second, to tax robots that replace human workers. He pointed out that employers currently pay taxes on workers' wages, while robots can be deducted as a business expense, effectively subsidizing automation. Third, to establish a "human reserve area," reserving caregiving, mental health services, and some medical positions for humans. The tax revenue would be used for retraining and supporting displaced workers.

Regarding slowing AI development, he stated that he would likely support a credible global slowdown plan if proposed, but geopolitical and economic incentives are driving all parties to move forward at full speed, so this will not happen. He advocates for managing this transition through democratic processes, involving elected officials, policymakers, educators, healthcare professionals, local officials, community and religious leaders, and specifically calls for cooperation between the U.S. and China. He also wrote that AI could either become the greatest equalizer in human history or the source of the most severe injustices.

This article has also drawn criticism. Oren Etzioni, a computer science professor at the University of Washington and founding CEO of the Allen Institute for AI, believes that Gates' diagnosis of the problem is accurate, but the token tax "is like taxing keystrokes on a typewriter," measuring usage rather than job replacement. He also pointed out that unilateral taxation by the U.S. could push AI usage overseas, weakening its competitive advantage against China, and this contradicts Gates' own statement that "we have no room for a slow approach."

In terms of market mechanisms, this is a viewpoint-driven discussion that does not directly correspond to transactions. The impact is mainly transmitted through policy expectations. Companies that charge based on tokens and cloud providers may face pressure; once token taxes enter legislative discussions, their gross margins and pricing models will need to be reassessed. Similarly, outsourcing, customer service, and legal process service providers that rely on entry-level white-collar jobs, as well as young people just entering the labor market, may also be under pressure. Industries such as caregiving and mental health, which are included in the "human reserve area," as well as retraining and educational service institutions, may benefit. Financially, this type of discussion will make the market more focused on the policy risks of AI capital expenditures, rather than just computing power demand.

Source: Public Information