OpenAI CEO Sam Altman has released a new set of guiding principles aimed at ensuring artificial general intelligence (AGI) benefits all of humanity rather than a small group of powerful companies.
In a publication titled "Our Principles," Altman outlined a vision where AI technology increases individual agency and prosperity. He warned that without intentional action, power could consolidate within a handful of corporations controlling superintelligence.
"Power in the future can either be holding by a small handful of companies using and controlling superintelligence, or it can be held in a decentralized way by people," Altman wrote. He stated that OpenAI's goal is to place general AI in the hands of as many people as possible.
Strategies for universal prosperity
The framework rests on five core pillars: democratization, empowerment, universal prosperity, resilience, and adaptability.
To prevent power consolidation, the company pledged to resist the tendency of AI to centralize authority. This includes advocating for democratic processes and egalitarian principles in AI decision-making, rather than leaving such choices solely to AI laboratories.
OpenAI also detailed a commitment to "universal prosperity," which involves building massive AI infrastructure and developing technology to drive down costs. The company noted it is already engaging in large-scale compute purchases and vertical integration to support this long-term goal.
Addressing potential risks, the company emphasized a strategy of iterative deployment. Altman noted that while AI can assist in cybersecurity, it also presents new threats, such as the potential for models to facilitate the creation of new pathogens.
"No AI lab can ensure a good future alone," Altman said, noting the need for collaboration with governments and international agencies.
OpenAI plans to use its Foundation’s resources to support resilience efforts. The company also acknowledged that its positions may change as technology evolves, stating that it might eventually need to trade off some user empowerment for increased safety and resilience in the face of unpredictable emergent behaviors.
