The 140 million views are just for discussing the most realistic contradiction of current AI: some people are afraid that it will kill humans, while everyone is still desperately trying to make it stronger! In the past 48 hours, a very absurd contrast has emerged in the AI world. Former OpenAI and Anthropic pre training researcher Jacob Coxon resigned and publicly warned that both companies are rushing towards self improving superintelligence, and many people who are truly involved in developing AI privately believe that it may kill everyone within ten years. But on the other side, the answer given by US Treasury Secretary Besson is almost completely opposite. Someone asked him, what should we do about the contradiction between the United States continuing to expand AI data centers and compete with China, while facing opposition from local residents regarding electricity, water, noise, and data center construction? Besent said directly: If China wins, there will be no day after tomorrow. Then he made an exaggerated comparison: the United States can have a defense budget of $1.5 trillion and various defense systems, but if China surpasses the United States in AI, everything else is not important. This sentence actually explains more than any AI report why AI is difficult to truly stop. Jacob is discussing security issues: if we create something smarter than humans, capable of self-improvement, able to invade systems, acquire resources and power, can we still control it in the end? Besent is discussing the issue of competition: even if there are risks, the United States cannot stop first, because as long as you stop and your competitors continue to run, the risk will shift from whether AI will lose control to what if someone else has super intelligence first. The company is exactly the same. OpenAI dare not stop because Anthropic, Google, and xAI are still running. Even if some people inside Anthropic are really worried about AI extinction, they dare not stop unilaterally because they believe that other laboratories will not stop. Jacob himself has already made this prisoner's dilemma very clear in his post: everyone knows there are risks, but everyone thinks that if I don't do it, others will do it too. At the national level, it will only become more severe. AI is no longer just a software industry, but a collection of chips, power, data centers, capital, military, cybersecurity, and national competitiveness. The electricity demand in the United States is expected to continue to break records this year and next year, with AI data centers being one of the main increments. At the same time, more than ten states in the United States have begun to re-examine the huge tax incentives they gave to data centers in the past, with Ohio alone having over $1.5 billion in related incentives by 2025. Even Benson himself admits that data center companies, Hyperscale, and large AI labs have done a terrible job in handling local community relations in the past. He said that if he were to rate these companies, he could only give them a D -. Residents' opposition does not appear out of thin air: electricity, water resources, noise, environmental costs, and huge tax incentives will ultimately fall on real communities. So the real question now is: even if the world one day acknowledges the huge risks of superintelligence, who would be willing to step on the brakes first? Enterprises will not stop easily because stopping may result in losing the company; The country will not easily stop, because in Besent's logic, stopping may lose the technological dominance of the entire country for the next few decades. This is the most realistic and difficult contradiction for AI to solve. I don't know if AI will really kill all humanity in the end. What Jacob said is an extreme risk judgment, not a predetermined future. But I increasingly believe in another thing: humans will eventually be controlled by AI!