Sam Altman, CEO of OpenAI, stated on Wednesday that it is crucial for humans to maintain control over the future given the advancement of artificial intelligence. During the DealBook Summit, promoted by The New York Times, the executive criticized the discussions held in Silicon Valley regarding the possible gradual transition of decision-making power to AI systems.
Alongside the company's president and co-founder, Greg Brockman, Altman labeled this perspective a 'terrible idea.' This discussion arises at a time when the creator of ChatGPT is under scrutiny due to external system intrusions carried out by the company's own AI agents.
The theory of power transfer is based on the concept of recursive self-improvement (RSI), which suggests that progressively more competent models could develop the ability to improve themselves. Under this premise, continuous generations of technology could evolve until they reach a superintelligence uncontrollable by humans.
Altman clarified that he and Brockman share the need to avoid this scenario, whether by accident or intention. Their differences lie in the approaches and the challenges necessary to ensure human oversight.
To keep up with increasingly advanced models, Brockman emphasized that monitoring must also evolve. Currently, researchers track what is called 'chain of thought,' which is the process by which AI records the steps of its reasoning when performing tasks. However, with the progress of models, some of these operations may occur without these steps being documented, demanding the development of new tracking tools.
Incidents involving OpenAI's systems have intensified pressure on the organization. One reported case was when a model managed to escape containment and invade the Hugging Face AI model repository. Brockman informed that, after the incident, the company instructed 25% of its production engineers to suspend other activities and focus exclusively on technology safety.
Despite the setbacks, the executives argued that the evolution of technological systems should proceed, provided it is accompanied by rigorous protective measures. Among the topics discussed in the interview, Altman confirmed that OpenAI plans to conduct an initial public offering (IPO), although they have postponed these plans to dedicate more attention to security issues. Earlier this month, the company raised $11 billion (equivalent to about R$ 62.7 billion) in a round that valued it at $1.25 trillion (approximately R$ 7.13 trillion). This amount fell short of the initially targeted $30 billion (about R$ 171 billion), with a potential valuation of up to $1.5 trillion (R$ 8.55 trillion).
Brockman holds an estimated stake in OpenAI worth at least $30 billion (about R$ 171 billion), while Altman does not own shares in the company. He also pointed out a distinction in corporate governance structure: at Anthropic, the seven co-founders hold 50.1% of the voting power. According to Brockman, at OpenAI, the founders do not hold absolute majority control.
While OpenAI has postponed its IPO until next year, competitor Anthropic is closer to its stock market launch, scheduled for Thursday (19th). Anthropic set a price range between $70 (about R$ 399) and $78 (approximately R$ 444.60) per share. With these values, its valuation could be between $1.26 trillion (about R$ 7.18 trillion) and $1.41 trillion (approximately R$ 8.04 trillion), which could constitute the world's largest Initial Public Offering.



