Elon Musk recently shared two key predictions during an interview with the Economist’s editor-in-chief, Zanny Minton Beddoes. First, he believes artificial intelligence could surpass the total intelligence of all humans within roughly five years. Second, he anticipates that Chinese companies are likely to become global leaders in the AI race.
Musk remains one of the most aggressive forecasters in Silicon Valley regarding AI's pace. He stated that within about five years, AI will be capable of nearly everything, potentially exceeding the combined intelligence of all humanity. “At that point, AI will be able to do almost anything, except perhaps be human,” he said. He predicts that by around 2036, barring major disruptions like a world war, humanity could enter an era of great abundance, where almost any goods and services imaginable are available to everyone.
Just three years ago, Musk was a vocal opponent of rapid AI development. In 2023, he signed a public letter calling for a pause in training powerful AI systems, and last year he estimated a 10% to 20% chance that killer robots could eventually destroy humanity. Beddoes questioned whether Musk had shifted from an AI risk pessimist to an optimist. Musk denied a fundamental change in his views, still acknowledging the risks of AI and robotics. However, he now leans toward a more optimistic outlook because he sees no way to stop the powerful momentum of AI and robotics. “Even if there were a stop button, perhaps we shouldn’t press it,” he said.
Musk stated that all paths lead to AI acceleration, and humanity’s role is not to halt development but to minimize the probability of negative outcomes. The impact on employment will be the first area where AI’s effects are felt. Musk noted that AI’s ability to write software already surpasses at least 90% of professional software engineers, and will soon exceed 99%, eventually reaching a point where humans cannot compete. Digital work will be replaced first, followed by physical labor like assembly, which will be taken over by humanoid robots.
However, Musk envisions that the disappearance of jobs does not necessarily mean widespread poverty. As the output of goods and services increases rapidly, society could move from “universal basic income” to “universal high income.” Work, he suggested, would become like gardening—people would engage in it out of interest, social interaction, and a sense of achievement, rather than for survival. The challenge lies in the potentially rough transition period between widespread job displacement and material abundance. Musk acknowledged that this process will not be smooth.
Turning to China’s competitive advantages in AI, Musk commented on the latest Kimi K3 model from Moonshot AI. He believes that Anthropic’s Fable is currently the smarter model, but Kimi is steadily closing the gap. “If Chinese AI companies obtain a large amount of computing resources, they are very likely to become leaders. And at some point in the future, they likely will have a lot of computing resources,” he said. Based on this, Musk views Chinese AI companies as strong candidates for future leadership.
Currently, the main constraints on AI include electricity and AI chips. Musk stated that while China’s primary limitation is chips, elsewhere in the world, the real bottleneck for AI is electricity and cooling. “AI chips consume a lot of power. People talk about the water needs of AI data centers, but in reality, AI consumes very little water; the huge demand is for electricity.” This is a reason Musk is considering building orbital data centers. He noted that the production speed of AI chips has already outpaced the rate at which new electricity supplies are coming online. The deciding factor for whether an AI chip can be used is having enough power, cooling capacity, and electrical equipment.
In his view, China’s clear advantage is electricity. China’s power generation already exceeds that of the United States, Europe, and India combined, and he believes it could eventually reach four times that of the US. Regarding US efforts to curb China’s AI development through export controls, Musk believes the effectiveness is limited. The US government might prevent US companies from using Chinese models, but it cannot stop other countries and regions from adopting them, nor can it prevent China from becoming a leader in AI. He further predicted that China is “closer to solving the lithography problem than most people realize.” Once China solves this, it will be able to mass-produce AI chips, and the constraints from current export controls could weaken. “China is already very strong in digital AI and robotics, and has vastly more electricity supply than the US. If China solves the problem of advanced chips and lithography, the competitive landscape in AI could change further,” Musk concluded.
Comments