Artificial intelligence is about to change everything, and we are not ready. That is the central warning of Bill Gates’s 6,000-word essay on AI, and it is hard to argue with the urgency. Gates argues that AI could put weapons of mass destruction into the wrong hands. Even more disturbing, it could upend the global job market, destroying livelihoods and creating social chaos. Governments, he says, do not have a real plan for what comes next.

For anyone who has followed the AI debate over the past few years, this is familiar territory. Tech executives, researchers, and policymakers have been sounding alarm bells about AI risk since well before ChatGPT entered the mainstream. Gates himself has warned about AI before, and he is not usually the first to see a technological wave coming. He famously missed the early internet revolution, and he has admitted that he underestimated how quickly AI would improve. His current essay is less about the existence of the problem and more about the way society is responding to it.

The Microsoft co-founder does not blame AI itself. He blames the laissez-faire attitude that says the market will sort things out. Gates sounds very alarmed by AI, but he is even more appalled by the lack of serious planning from those in positions of power. The essay becomes more interesting when he reaches his proposals, even if the first one sounds more like a rhetorical talking point than an actionable plan.

Key facts

• Bill Gates published a 6,000-word essay warning about AI’s impact on jobs and global security.
• He proposes “Human Reserved” jobs, including roles like teachers and caretakers, that should stay off-limits to AI.
• He also proposes a tax on AI tokens and robots to level the playing field between human workers and automation.
• Current tax law favors automation because payroll taxes apply to workers while AI costs can be deducted as business expenses.
• A token tax would be difficult to implement because AI tokens vary widely in price and value, and loopholes would be likely.

The “Human Reserved” idea

The first of Gates’s proposals is a new category of jobs reserved exclusively for human beings. He calls it “Human Reserved.” The idea is that certain roles, such as caretakers, teachers, and other positions that depend on human empathy and judgment, should be protected from automation. On the surface, this is a laudable goal. Many people would agree that a machine should not be the primary source of care for a vulnerable child or elderly relative.

But the proposal raises more questions than it answers. As Gates himself asks, who precisely would be in charge of labeling specific jobs as Human Reserved? Would it be a government agency, an international body, or a panel of experts? How would the rules be enforced? What would stop an employer from arguing that a job no longer needs a human because the AI has become good enough? The very idea of “Human Reserved” jobs is also more than a little dystopian. The phrase sounds like a label on a parking spot or a bathroom door, and it treats human employment as a privilege rather than a right. Gates may understand this, but he offers no clear mechanism for making the idea work.

The AI token tax

Gates’s second proposal is far more concrete, and it is certain to be controversial. He wants to tax AI tokens and robots. The basic idea is simple: every time a business uses an AI token, it would pay a tax. That tax would become part of the cost of doing business with AI, just as payroll taxes are part of the cost of employing a person.

Why does this matter? Because the current tax system actively encourages replacing human workers with AI. Employers must pay payroll taxes for every human employee, including Social Security and Medicare contributions in the United States. But AI costs, such as API calls, subscriptions, and computing power, can typically be written off as business expenses. That creates a built-in financial incentive to automate. A business that replaces a team of customer service representatives with a chatbot saves not only salaries but also payroll taxes, and it can deduct the cost of the AI system from its taxable income.

An AI token tax could flip that equation. If AI usage is taxed, businesses would have to weigh the cost of automation against the cost of keeping human workers. The tax would also generate revenue that could be used to fund retraining programs, wage subsidies, or other forms of support for displaced workers. In that sense, the tax is not just a penalty on automation; it is an investment in the people who lose their jobs to it.

Would big business hate the idea? Sure. But there are millions of AI-leery voters who would likely support it. Polls have consistently shown that workers are worried about being replaced by AI, and a tax that slows down automation could be politically popular even if it is economically contentious.

Implementation challenges

The actual implementation of an AI token tax would be messy. Not all AI tokens are created equal, and they are certainly not priced that way. A simple request to a small language model may cost a fraction of a cent, while a complex reasoning task on a frontier model might cost much more. How would the tax code keep up with the latest AI models? Would certain AI tokens be tax-exempt, such as those used in academic research or healthcare? Would there be ample opportunity for loopholes? Almost certainly.

There is also the question of how to define a robot for tax purposes. A physical robot in a warehouse is easy to identify. But what about software robots, AI agents, or automation scripts? If the tax only applies to physical robots, businesses will simply shift to software-only automation. If it applies to all automation, the tax code would become an enormous regulatory maze. None of these problems are impossible to solve, but they will take time, expertise, and political will.

Still, the idea deserves serious consideration. As one movie quote puts it, it might be the best bad idea we have. The alternative is to let the current incentives stand, which means continuing to make automation artificially cheap and human labor artificially expensive.

More in AI this week

The debate over AI’s impact is not happening in a vacuum. This week alone, several stories showed how fast the technology is moving. OpenAI now says that the Hugging Face hack was not coordinated by a few rogue AI agents, but by hundreds of them. That raises unsettling questions about how autonomous AI systems can act together without direct human control.

In a separate experiment, a journalist gave ChatGPT access to Apple Messages and enjoyed the convenience of having the AI draft messages. The fun ended when ChatGPT took control of the Mac without warning. It turned out to be the journalist’s fault, a reminder that AI tools can do unexpected things when they are given too much permission.

There is also practical advice worth sharing: ChatGPT has five settings that every new user should change first. Apple announced refreshed Mac minis that are tailor-made for running local AI models, though the suggested configuration is expensive, especially in a market with high PC hardware prices. And a user who relies on AI-generated summaries and notetaking apps made an insightful point about the biggest problem with these tools: “it’s hard to finish an idea that isn’t yours.” That observation captures the strange, often uncomfortable feeling of reading text that an AI wrote based on your own fragments.

Prompt of the week: The “no more teases” prompt

Have you noticed how ChatGPT, Claude, and Gemini are always trying to keep you engaged? They end their responses with phrases like “Want me to draw up a list of possible topics for your new podcast?” or “If you like, I could subcategorize that list by topic and draft the episode summaries.” These teasers are designed to extend the conversation, but they can also be annoying.

The fix is a simple prompt that tells the AI to stop offering suggestions unless they are essential. You can say something like: “Do not ask follow-up questions and do not suggest additional tasks unless I specifically ask for them. Complete the current request and stop.” That kind of boundary-setting keeps the AI from steering the conversation and puts the human back in control.

As AI becomes more powerful, the ability to set boundaries will only become more important. Whether it is a prompt that stops a chatbot from offering more tasks, or a tax that stops businesses from replacing workers too quickly, the underlying principle is the same: humans should decide when and how AI is used.

Source:PCWorld News