Elon Musk: Superintelligence May Persuade Most People at Once
xAI and Tesla, SpaceX founder Elon Musk stated that many people could be persuaded at once by a super persuasive superintelligence, and he himself might be one of them.
He used the gaming term "one-shotted" to describe this process, meaning being completely persuaded or having one's stance changed in a single or very brief interaction, making it difficult to detect or effectively resist.
Musk pointed the risk subject towards superintelligence with extreme persuasive capabilities, rather than current models, and explicitly included himself in the potential affected range.
This statement continues his long-standing public concern about AI safety, emphasizing that even individuals with technological and resource advantages may not be completely immune to highly personalized persuasion.
When models surpass humans in understanding psychological weaknesses and constructing targeted arguments, traditional cognitive defenses and information filtering mechanisms will face the risk of failure. Persuasiveness shifts from relying on large-scale dissemination to precise single-point breakthroughs.
Capital and R&D resources are accelerating towards enhancing model persuasion and alignment capabilities, while some resources are also directed towards adversarial detection and multi-model cross-validation mechanisms. Those who control top models gain influence distribution rights, while individuals and institutions lacking protection face a higher risk of being guided.
Source: Public Information
ABAB AI Insight
Elon Musk co-founded OpenAI in 2015 but left the board due to disagreements over safety paths and governance methods. He later founded xAI in 2023, focusing resources on developing models that pursue maximum truthfulness. This path shows his long-term view of AI safety and alignment as core variables, rather than merely pursuing capability limits.
Through Tesla, SpaceX, and xAI, he mobilizes computing power, data, and engineering resources to promote his own model training, motivated by reducing dependence on external closed models and attempting to establish a verifiable safety framework before superintelligence emerges. Specific actions include publicly warning about risks while simultaneously advancing his own model capabilities.
This contrasts with the early path of OpenAI, which started with open collaboration and later shifted to closed development. Currently, leading institutions are shifting from a pure capability competition to simultaneously competing for persuasion control and protective mechanisms, with the industry at a stage of capability leaps and risk manifestation.
Essentially, this is a technological substitution: when superintelligence fully surpasses humans in persuasion, the distribution of value and influence shifts from humans who control information and narratives to those who can generate and deploy highly personalized persuasion. The mechanism is that a single interaction can complete belief updates, significantly compressing the costs of traditional multi-round negotiation and social validation.
ABAB News · Cognitive Laws
When persuasion exceeds defense, a single interaction can rewrite stances.
The stronger the capability, the greater the risk of dependence on top models.
When alignment fails, influence rights concentrate with the model controllers.