Superintelligence for all, argues Meta boss

Mark Zuckerberg opposes the concentration of superintelligence. His op-ed advocates for universal access to balance economic power.

An op-ed by Mark Zuckerberg has been published in the Wall Street Journal, dedicated to the distribution of superintelligence. His thesis lies in a reformulation of the question: the issue is not whether systems exceeding human capabilities will exist, but who will have access to them. Two paths are contrasted—intelligence reserved for a few institutions versus a tool placed in everyone's hands—and the author defends the second based on three principles: individual empowerment as a source of prosperity, invention as the primary purpose, and the balance of power as the foundation of security.

The text targets the doomsday rhetoric of a portion of the sector, without naming anyone. He sees an inconsistency in rapidly building a technology that one describes as bound to eliminate most jobs and human relevance, and deems dangerous the idea that the only safe path lies in an extreme concentration of power. The argument draws on a historical precedent, that of absolute powers whose benevolence was hoped for.

Several lines of reasoning support the demonstration. First, that of alignment: since humanity is not homogeneous, no single system could embrace contradictory interests without arbitrating in favor of certain values over others. Next comes a thought experiment on security. If only one person had a superintelligent lawyer, they would win lawsuits they should have lost; if everyone has one, justice is better served. Finally, on employment, where the distinction lies in usage: automation has a negative impact, whereas a widely distributed tool of empowerment would, in his view, produce more jobs and a more entrepreneurial economy.

A nuance emerges regarding risks. In cybersecurity, the author believes that the history of open source argues for full access to powerful systems. On biological risks, however, he concedes that increased coordination between governments and institutions would be useful. The op-ed was published on the very day that an inter-laboratory petition called on the United States to support the development of tools to regulate the pace of automated AI, a text that Meta's chief scientist signed in a personal capacity.