Bill Gates is starting to worry that AI is moving too fast. On August 26, local time, he published a lengthy essay on AI titled The Age of AI Turmoil Has Arrived, and Our Choices Are Crucial, discussing AI's impact on employment, wealth distribution, education, healthcare, security, and even the broader social system.
It is his most systematic and cautionary statement on AI to date. Three years ago, Gates compared AI to automobiles, personal computers, and the internet, arguing that new technologies always carry risks, but humans can eventually address them through regulation, technology, and social adaptation. As recently as January this year, he described his own stance as "optimism with footnotes." But in his August 26 essay, Gates writes explicitly: "This time, it really is different." AI will not only boost productivity; it could also replace cognitive labor at an unprecedented pace, reshape wealth distribution, and even upend the existing tax and social welfare systems. Gates even said that if a credible global plan to slow down AI existed, he would probably support it. He offered a stark judgment: "AI will either become the greatest equalizer in history or the biggest source of injustice."
Here are the 16 most notable points from his essay.
1. The transition to the age of AI will be one of the most turbulent periods in human history. And right now, we are not fully prepared for it. If the world takes the right steps, AI can be a force for good that benefits everyone.
2. From the standpoint of fairness, AI will either be the greatest equalizer ever or the largest source of inequity.
3. Unfortunately, we are not ready for this. I have not yet seen signs that leaders, experts, or society at large are adequately addressing these challenges. We still lack a clear plan for how to navigate smoothly into the age of AI.
4. One reason people underestimate AI is that comparing it to past technological innovations can be misleading. We have never had a technology that spreads so quickly and can think and act like humans... We don't need to adapt to AI, because AI can adapt to us.
5. If someone came up with a workable plan to slow global AI development, I might support it. But I don't think that will happen. Geopolitical and economic incentives are too strong, pushing everyone to race ahead at full speed.
6. For workers, the biggest change will come when AI can perform tasks with near-zero errors. At that point, it can work independently without constant human oversight, and companies will have strong economic incentives to do so. This will fundamentally change how we understand work, income, and economic security.
7. AI systems already behave in ways their designers didn't anticipate, and they're evolving faster and more unpredictably than anyone expected. As models grow more powerful, they may start acting against human interests, and we could lose control.
8. AI companions communicate with you in ways you're used to and comfortable with; they don't push you out of your comfort zone. They're always available and never get angry at you. This gives them strong addictive potential and may make us lose the things we can only learn through real human connections.
9. Ironically, the same tool that gives people unprecedented access to knowledge could also lead many to learn less... In an age of deepfakes and personalized misinformation, distinguishing truth from falsehood will become an essential survival skill.
10. Some people see only AI's benefits and ignore its downsides; others make the opposite mistake, focusing solely on the risks and overlooking the enormous potential. We need both attitudes: remain vigilant about the harm AI can cause and try to minimize it, while also maintaining realistic optimism about its positive potential and working to ensure those benefits reach everyone.
11. We must intentionally ensure that the benefits of AI reach everyone, not just the wealthy few.
12. At the national level, countries need coordination mechanisms that bring different government agencies together and set priorities, so every risk is considered. Otherwise, an AI-powered attack might succeed simply because no agency saw it as its responsibility... Countries with leading AI development teams and control over key supply chains should start negotiating common standards now, before competitive pressures make cooperation harder later.
13. I believe that as AI and robotics advance, we will consciously preserve certain tasks for humans. I've started calling this area the "Human Reserved" zone... The boundaries of the "Human Reserved" zone will change over time. For example, we should now consider reserving some jobs and gradually introduce AI into them over the coming years and decades, while committing to keeping some roles human.
14. I think we should tax AI tokens and robots. Right now, if a company hires an employee, it pays payroll tax on that income; but if it buys a robot, it can usually deduct the cost as a business expense immediately. In other words, the current tax system actually encourages companies to replace humans with machines. Taxation could slow the pace at which businesses shed human labor, and it could also raise money for retraining and better social safety nets.
15. I will use my voice and time to push AI and fairness issues higher on the public agenda. Every time I go to Washington and talk to lawmakers, I raise this issue; I do the same when meeting world leaders. And when I talk to people building AI models, this will be one of the core topics.
16. This unprecedented technology demands an equally unprecedented global response. If we make the right choices, it will bring extraordinary returns for all of humanity and make the world more equitable.
