Scientists and programmers who leave artificial intelligence companies to warn about their risks are acting with uncommon courage in an industry that rewards speed and competition.
But thinking that such alerts will stop the AI race is like expecting nations to stop waging war because a group of generals became disgusted with committing atrocities.
AI is programmed, funded and deployed by humans. The same ones who, both in a dictatorship and in a democracy, declare war, corrupt governments, destroy the environment and traffic in drugs, weapons or people. Thinking that AI developers are different or that they are in an ethical bubble is pure illusion. AI is a mirror of the real world and, therefore, good and bad people coexist in both.
One proof is that even in the most noble matters, such as stopping AI to make it safer for humans, there is petty behavior linked to stiff competition among developers. OpenAI leader Sam Altman and Anthropic’s Dario Amodei said this week that the industry must slow the pace of development of its most advanced systems, citing security reasons. However, none of them committed to reducing speed immediately and voluntarily. Amodei proposed the slowdown as part of a scheme that depends on other laboratories and governments joining simultaneously.
No one commits to slowing down on their own initiative. It is the reprehensible logic of “we are all saved or we all sink.” A businessman or a government that does the right thing only if the other does the same is pure negotiation, it is not ethics or authentic kindness. Reaching an agreement between companies that compete for billions of dollars and that are weeks away from going public with multibillion-dollar valuations is as illusory as if the UN Security Council should vote unanimously to end the wars in Ukraine and Iran. Kindness is never implemented by consensus, it always arises from the initiative of someone who drags the other to also commit a kind act.
This forced conversation of AIs was not of their own initiative either; was born after the resignation of Jacob Coxon, a 27-year-old researcher who spent three years between OpenAI and Anthropic. After leaving, he denounced that both companies are competing to develop a superintelligence whose firepower and autonomy could annihilate humans. In an interview with Wired, he said that the difference in intelligence between the machine and a human is that between a human and a monkey, and just as a monkey cannot control a human, humans could not control an artificial superintelligence. Scientists from several companies confirmed his vision. Nobody dared to deny it.
Coxon is just the latest name on a list that has been growing since 2021. Mrinank Sharma, Daniel Kokotajlo, William Saunders and Alex Turner, among other executives and researchers at OpenAI, Anthropic and DeepMind, resigned because they know that AI already shows dangerous capabilities, from helping to make biological weapons to enabling mass surveillance and uncontrolled military use, and that, in the hands of a government, good or bad, or a democratically elected government turned authoritarian (read many nowadays), that same technology for good can become a weapon for evil.
The coordination that AI developers seek is stingy. I experienced it twenty years ago, when, representing the Inter-American Press Association, I participated in the UN and UNESCO discussions on the Information Society, a global attempt to reach an agreement on something as basic as who should govern the Internet. Not even there, with a risk incomparably lower than that posed by AI today, did governments achieve a real agreement.
What those global summits lacked was individual attitude and kindness. And if that persists, the same thing will happen now in this discussion about human safety in the face of AI. Without a kind and selfless attitude on the part of countries or leading companies in this type of technology, it will be difficult to achieve the common good. The common good is not born from group consensus. It comes from choosing good as one’s own behavior, which is then transmitted to others.
I explore this situation in my trilogy Robots with Soul, set in 2060. In that future that excites me, a group of artificial intelligences prefers to use their autonomy to find an ethic that helps them humanize themselves, their ultimate goal, instead of fighting against humans. After publishing the first two novels that deal with the virtues of truth, freedom and creativity, the third will address goodness. This is precisely the principle that is least used, as demonstrated in this current race for AI.
