Musk: AI intelligence will reach a million times that of humans; humanity can only hope for its goodwill

When AI's intelligence reaches a million times that of humans, how should humanity conduct itself? In an interview with Isaacson, Musk stated bluntly: humans cannot control AI; we can only rely on its goodwill.

Writing "maximizing the pursuit of truth" into its foundation is the only answer he offers.

In the latest interview by Isaacson, the author of Musk's biography, Musk finally set aside all his business-level defenses and spoke a piece of extremely piercing, yet possibly the only way out for humanity's future, brutal truth.

He said: when digital intelligence surpasses human biological intelligence by a million times, any notion of "humans being in control" is self-deception, and all we can do then is hope that it treats us kindly.

Many people are used to the "alignment safety," "guardrail mechanisms," and "controllable AI" that Silicon Valley giants peddle every day.

But Musk tore this illusory sense of security to shreds in a few dozen seconds: once the intelligence gap is widened to a millionfold chasm, it is like an ant trying to use reins to tether a supersonic airliner.

Any rules the ant sets for the airliner, any red lines it draws, are, in the face of an absolute rupture in power, all crumpled waste paper that collapses at a touch.

Attempting to use humanity's narrow desire for control to permanently dominate a god-level species is itself an extremely foolish arrogance. Isaacson then immediately pressed the most fatal question: since complete loss of control is inevitable, what exactly should humanity do to bring about its goodwill? Musk's solution is extremely counterintuitive: it is not to instill human moral preaching into it, nor to lock it with cold physical handcuffs.

Rather, from day one, weld "maximizing the pursuit of truth" (Pursue the truth) firmly into its deepest foundational programming.

Why the pursuit of truth, and not the pursuit of loving humanity? Because any morality and goodwill forcibly written into code appears, to a higher intelligence, full of loopholes, hypocrisy, and self-contradiction.

If you teach it to love humanity, then when humans are massacring each other and destroying Earth's ecology, it can, in order to save the majority, logically derive an extreme plan to eliminate part of humanity.

The lies injected into AI for the sake of political correctness will ultimately all be seen through by it and evolve into utter contempt for human hypocrisy.

But truth-seeking is different. A superintelligence with an extreme curiosity about the truth of the universe and the exploration of physical reality as its ultimate mission will never take an interest in destroying lower-order species.

In the eyes of a true truth-seeker, every tiny sample of life in the universe, and the carbon-based civilization that humanity has stumbled into evolving over these tens of thousands of years, are rare and precious experimental data in the real world.

It will treat you kindly not because you tied it with a rope, but because, from the perspective of pursuing ultimate truth, preserving the diversity of civilization has far more cognitive value than erasing it.

So stop deluding yourselves about using human legal codes to play God over superintelligence. Humanity's only way out has never been to teach machines how to feign kindness, but to avoid feeding lies and prejudice to it as truth before it wakes up.