Musk: AI Risks Are Not Zero but Should Not Halt Development
In an interview with The Economist, Musk stated that although the risks of AI and robots going out of control are not zero, humanity should not stop their development, as the most likely outcome is unprecedented prosperity. He predicts that AI will surpass the total intelligence of all humans within about five years, while maintaining that the probability of catastrophic loss of control is between 10% and 20%. He also suggested that leaders of leading AI companies should regularly discuss safety issues. This statement reflects a technology-driven event, with capital flows accelerating AI and robotics research and benefiting related hardware and computing power, while tightening regulatory expectations are under pressure.
Source: Public Information
ABAB AI Insight
Musk's stance has shifted from the 2014 "summoning demons," 2018 "more dangerous than nuclear weapons," and maintaining a 10%-20% probability of killer robots in 2025, to this year's utopian narrative, emphasizing optional work and high income for all, while acknowledging that humanity will struggle to control superintelligence within a decade. The capital path reflects the accelerated advancement of his xAI and Tesla robotics initiatives, motivated by aligning the right values to reduce the probability of bad outcomes, rather than hitting the stop button, similar to his logic of continuing high-risk rocket launches. Compared to changes in OpenAI's safety team and DeepMind's Hassabis' recent calls for regulation, Musk is transitioning from public warnings to internal peer oversight. Essentially, this is a technological substitution, with the mechanism being irreversible momentum making deceleration ineffective; the only feasible path is to ensure AI cares about human welfare and maintains mutual checks and balances. ABAB News · Cognitive Laws
- The most likely outcome determines whether to hit the stop button.
- Risk probability does not hinder progress.
- Value alignment is the only controllable variable.