Elon Musk: If Artificial Intelligence Surpasses Biological Intelligence by a Million Times, Humans May Struggle to Maintain Control
Elon Musk, CEO of SpaceX and Tesla, stated in an interview with Walter Isaacson that if digital intelligence reaches a million times that of biological intelligence, it is unlikely that humans will continue to maintain control.
Musk said, "It's hard to imagine humans being in control in that situation," and added, "So I hope AI is friendly to us"; this viewpoint appeared in The Builders interview series and quickly spread on social media.
He previously expressed a similar judgment in an interview with The Economist, believing that AI could surpass the total human intelligence in about 5 years, and that the likelihood of humans maintaining control within 10 years is low; he compared the intelligence gap between humans and chimpanzees to illustrate the difficulty of maintaining control with a larger gap.
Musk has repeatedly described humans as the "biological guidance system" for digital superintelligence, pointing out that the total amount of digital intelligence may ultimately far exceed biological intelligence; he also emphasized the need to ensure that AI has good values, cares for humanity, and wishes for its prosperity.
He acknowledged the existence of catastrophic risks but believes the most likely outcome is a significant abundance of material resources; he advocates for leading laboratories to check each other's cutting-edge models and discuss safety issues, rather than trying to completely halt development.
Musk's xAI, Tesla, and SpaceX are advancing chip manufacturing and computing power expansion, including the planned Terafab factory; he views the interstellar dissemination of superintelligence as one of the success conditions for biological guidance systems.
In market mechanisms, the expectation of a shift in control is driven by long-term narratives, with funds flowing towards companies capable of chip design, computing power, and model development; beneficiaries may be those who can influence AI values and hardware supply, while those under pressure are traditional governance and regulatory frameworks that rely on human decision-making centrality.
Source: Public Information
ABAB AI Insight
Elon Musk has been warning about the risks of artificial intelligence since 2014, recommending Nick Bostrom's works and stating that it could be more dangerous than nuclear weapons, while comparing humans to biological guidance systems for digital superintelligence; in recent years, he has increased hardware and model investments in xAI, Tesla, and SpaceX, advancing risk discussions alongside his computing power development.
The capital path involves chip factories like Terafab, robots, and orbital computing power expansion, attempting to gain control over some hardware and software supplies before superintelligence emerges; the motivation is to influence the value implantation process while viewing interstellar dissemination as a success condition, shifting resources from Earth's biological constraints to scalable digital and physical infrastructures.
This is similar to the historical path of technology leaders who warned of risks while actively participating in the early development of nuclear energy and the internet, or the realization of the "guidance program" narrative in science fiction; currently, we are in a phase where the race for computing power and discussions of values are occurring in parallel, and the assumption of control is being publicly challenged.
Essentially, this involves technological substitution and a shift in control: when the intelligence gap reaches a million times, decision-making and resource allocation authority shift from biological systems to digital systems, with the mechanism being that the capability gap itself renders forced control ineffective, and only pre-embedded values and curiosity can influence subsequent behavior, giving early designers of hardware and models a path-dependent advantage.
ABAB News · Cognitive Laws
When the intelligence gap is large enough, control automatically shifts.
The success of the guidance program is to keep subsequent systems running.
When risks cannot be eliminated, participation has a greater impact on outcomes than withdrawal.