When Discussing AI, Don’t Always Use the Industrial Revolutions as a Shield
Whenever the topic of whether AI will take away jobs comes up, there’s always someone who steps forward to say, “Wasn’t it the same with the first three industrial revolutions? Old jobs disappeared, and new ones naturally emerged—there’s no need to worry.”
People who say this are either malicious or ignorant. They deliberately ignore how many people died during the transitions of the first three industrial revolutions, how many years of colonial expansion took place, and how many lives were lost in World War I and World War II. To people living centuries from now, a few decades will be nothing more than a line in a history book, but for those who lived through those times, it was a lifetime of upheaval.
Now, as we stand on the threshold of the AI revolution, we are no longer bystanders—we are participants. Stop constantly citing historical precedents. If we were to strictly follow historical patterns, China would have been a powerful nation for five thousand years—so does that mean we can just sit back and keep winning? Obviously not. So why assume that just because the Industrial Revolution created new jobs, the AI Revolution will inevitably follow the same path?
One more point must be made clear: the AI race is now the central battleground in the US-China strategic competition. AI certainly has its issues, but if we lose this race, the problems we face will only become more severe. The wheels of technological progress cannot be stopped; avoiding the issue is futile. It’s better to figure out exactly what will happen.
The Most Frightening Aspect of AI Isn’t Its Learning Ability—It’s Its Ability to Inherit
Everyone is talking about how powerful AI’s learning ability is—it can master a human’s decade-long curriculum in just a few months. Whether it’s law, finance, or medicine, any discipline that relies on experience, AI picks it up very quickly. But that isn’t the scariest part.
When it comes to learning efficiency, humans are leagues ahead of AI. Humans can work a full day on just a few steamed buns. Back in the days of the “Two Bombs, One Satellite” program, scientists could develop cutting-edge technology even on empty stomachs. The human brain consumes only 20–30% of the body’s energy yet can perform such complex tasks. What about today’s AI? It requires stacking countless servers, consuming vast amounts of electricity, and feeding it endless data. If it were to truly achieve human-like adaptability—learning from one example to apply it to many—the cost would be unimaginably high.
The truly terrifying aspect of AI is its ability to replicate. Once a model is trained, you simply copy its parameters and deploy it on as many servers as you like—and it iterates incredibly quickly. It takes humans two or three decades to train a single skilled engineer, but AI can replicate ten million instances of that same capability in just a few minutes. It is this disparity in the efficiency of generational transfer that poses the most fundamental threat to human society.