Sam Altman has not stopped expressing his concern about the evolution of AI in the last 48 hours, since a former Anthropic engineer voiced alarm that this technology could end Humanity. One day after explaining that it would not be appropriate for his company, OpenAI, to go public now, the company’s CEO admitted on his X account serious safety concerns about artificial intelligence.
Read more Someone stop us
Altman explained that “there are two ways AI progress could go very wrong and that we must avoid.” The first is that “we could lose control of the future to AI.” And he added: “this is unacceptable.”
“We are unequivocally on Humanity’s side, and AI must always serve people,” he added. According to the OpenAI leader, “to ensure that, we need ways to make sure alignment and safety techniques stay ahead of progress in model capabilities.”
“If a very powerful AI is used by a person or company, the results could be extremely dystopian.”
The second threat is that “we could end up in a world with too much concentration of power.” “If an extraordinarily powerful AI,” he reflected, “is used by a person or company to impose their worldview on everyone else, the results could be extremely dystopian.”
That is why he commented that “avoiding these two threats requires walking a narrow middle path; for example, one country could gain too much power. Another example is a lab ending up with too much power.” In both cases, everything is very likely to happen in the United States.
Along those lines, Altman indicated that “the world deserves to have confidence that U.S. companies developing increasingly capable AI will act responsibly, especially now that the trajectory of progress has accelerated.” “Every frontier lab,” he proposed, “must comply with this, and there is no reason why any of us should go to work if we cannot do so.”
Read more Marc Aloy: “We cannot normalize young people walking the streets with knives”
What Altman suggests is a federal legal framework in the United States “that establishes consistent safety requirements for frontier AI — more advanced models —“, but he does not consider that we should wait for legislation or government antitrust measures “to start the work of providing this confidence.”
The OpenAI leader’s proposal is to apply “consistent rules to manage frontier risk” so that benefits can be maximized. Altman says he is excited about the idea of having independent auditors to oversee the enforcement of those rules. The executive talks about collaboration among companies: “we believe shared standards for misalignment, monitoring, and safety will lead to better outcomes. We look forward to collaborating with our colleagues across the industry to formulate the best version of these.”
Also read
Although that pace control, for Altman, does not mean “stopping,” even though he admits it should go slower. “Progress has been fast and will continue to be. But it should be slower than it might otherwise be; interventions like safety cases and monitoring have significant costs,” he observed. “The pace will be well worth this cost; no amount of U.S. competitive pressure should justify recklessness, nor allow capabilities to outpace alignment and monitoring.”
Finally, the OpenAI CEO has called for help from the U.S. federal government “for international coordination.” “But first,” he concluded, “we must do what we can ourselves.”
Read more One day, one human week
