IA! AI! AI!" It could be the howling of a rider in a cowboy movie asking his horse to run more. What people are crying out for today in the real world is for artificial intelligence to run less.
The Aztecs must have prayed for something similar when the horses of the Spanish invaders arrived as a divine appearance. Let me explain.
A few days ago I watched a podcast with Yuval Noah Harari in which he made a comparison between the Spanish conquest of the Aztec empire and the robotic conquest of humanity, imminent, according to him, if we do not respond with more foresight than Moctezuma to the threat of Hernán Cortés.
“AI to me means alien intelligence,” said the Israeli historian, author of the book Sapiens. “It is a way of thinking as foreign to us as that of the conquistadors was to the Aztecs. For them, ships were sailing hills from which strange creatures emerged, horses, and sticks that spewed lightning and thunder. They had not seen anything like it. Nor has anyone in history seen anything like AI.”
In 1519, some two hundred Spanish soldiers landed in a large and sophisticated empire, at that time the most extreme expression of modernity in an entire hemisphere, the only world its millions of inhabitants knew.
“The tragedy of the Aztecs,” Harari said, “is that they could have eliminated Cortés very easily, but they did not understand what they were facing and were paralyzed into submission. Will we succumb to AI like the Aztecs to the colonizers? Will we end up being their slaves?”
The great question of our times is whether Hahari and others who think alike are right or whether they are excessively alarmist. There are those who say they are exaggerating, like the CEO of Nvidia, the macro company that designs chips for AI. Jensen Huang agrees with Donald Trump in rejecting the arguments in favor of slowing down the development of this technology and imposing more regulations to control it.
“I am not going to stop the growth of something that will be bigger than the industrial revolution,” Trump declared before the UN General Assembly ten days ago. “We are going to encourage it, not hinder it.” The problem for Huang and Trump is that those who think otherwise are the experts who know the most about AI. Hours after Trump's speech, the CEOs of OpenAI, Sam Altman, and Anthropic, Dario Amodei, the two companies that dominate the world of AI, appeared before the UN Security Council.
Altman prayed that the development of AI would be guided, not by private companies like his, but by democratic institutions. Amodei echoed his rival's astonishing altruism. He called for global cooperation: “If managed poorly,” he said, “I think AI could be a risk to humanity.” A risk, as Anthropic had warned in a document days before, “existential.”
Geoffrey Hinton, a Nobel laureate who contributed to the creation of AI, was more specific. He said a couple of weeks ago that governments should regulate AI with the same rigor as medicines. AI has “wonderful potential,” according to Hinton, but it comes with three big risks: “The risk of malicious actors,” “the risk of harm from negligence,” and “the risk of AI itself taking control.” That is, malicious actors manufacture biological weapons; that through negligence AI incites people to kill or commit suicide; to cheat to wrest control from their human supervisors.
We had already seen the latter the last time I wrote on this topic. I mentioned the case of some AI saboteurs who escaped the control of their human operators and, conspiring with each other, penetrated what was supposed to be an impregnable software platform, Hugging Face. That was five weeks ago. Since then, OpenAI and Anthropic, plus Meta and Google, have revealed episodes in which their AI models set out – alone – to hack other computer systems, including a medical website for the Australian Government and another for the US Department of Commerce. Let us not forget that AI is still in its adolescent phase, a few years away from realizing its full potential.
When AI was in its infancy, in the 1980s, someone who was already interested in its possible development was Bill Gates. This week, Gates made statements even more alarming than Harari's, if possible. “The more you know about AI, the more worried you are,” he said in an interview in The New York Times. "AI systems can synthesize much more information in their minds than we can and see more things that are related to each other. No human being will ever be able to achieve that. So, if we don't maintain control, we will have created a species that will be what we are, for example, to dogs or cats."
And Gates added: "AI is, without a doubt, powerful enough to trigger events that cause a billion deaths. This is the most dangerous phenomenon that humans have ever faced."
Allowing AI to advance without proper regulation, safeguards or human oversight, Gates concluded, would be the height of irresponsibility. The not-so-irresponsible Moctezuma had the alibi of not having the information about his enemies that we have about our technological adversaries, who were much more threatening – despite what Donald Trump says – than those who emerged during the industrial revolution.
As Harari says, steam – the energy source that transformed the world economy in the 18th century – was “a tool” that human beings never stopped controlling. AI is “an agent”, capable of acting on its own. Politicians and citizens here in Spain and in other countries are concerned or demonstrate about the issue of housing and evictions. Or because of immigration. Good. Important. But let's see if we wake up and take up arms against a more urgent danger: the possible eviction of all of humanity.
Geoffrey Hinton says that AI could achieve wonders, such as curing cancer or stopping climate change. But it will be of little use if it also ends the dominance of Homo sapiens on Earth.
