Unique Research · Industry · January 7, 2025
In a personal blog post, OpenAI CEO Sam Altman wrote that he believes OpenAI already “knows how to build AGI in the traditional sense,” and is beginning to shift its aim toward “superintelligence.”
“We love our current products, but we are here for the glorious future,” Altman wrote. “Superintelligent tools could massively accelerate scientific discovery and innovation far beyond what we alone can do, and thereby greatly increase abundance and prosperity.”
Altman has previously said superintelligence may be “a few thousand days away,” and that its arrival will be “more intense than people think.”
AGI is a fuzzy term, but OpenAI has its own definition: “highly autonomous systems that outperform humans at most economically valuable work.” OpenAI and close partner-investor Microsoft also have a definition tied to economics: an AI system that can generate at least $100 billion in profit. Under their agreement, when OpenAI hits that bar, Microsoft loses access to its technology.
Which definition did Altman mean? He did not say; the former seems most likely. He wrote that AI agents — systems that can autonomously carry out specific tasks — may “join the workforce in some form” this year and “materially change companies’ output.”
“We still believe that iteratively putting great tools in people’s hands leads to great, broadly distributed outcomes,” he wrote.
That may prove true — but it is equally undeniable that today’s AI still has major limits: it hallucinates, makes mistakes any human can spot, and can be very expensive.
Altman seems confident about overcoming all of that — and on a tight clock. Yet if the past few years of AI taught us anything, it is that timelines can slip.
“We are pretty sure that in a few years, everyone will see what we see, and will recognize how critical it is to act with extreme care while maximizing broad benefit and empowerment,” Altman wrote. “Given what is possible with our work, OpenAI cannot be a normal company. How lucky and humbling it is to get to play a part.”
One hopes that as OpenAI turns attention toward what it sees as superintelligence, it will put enough resources into keeping such systems safe.
OpenAI has repeatedly written that a successful transition to a world with superintelligence is “far from guaranteed,” and that it does not have all the answers. In a July 2023 blog post it wrote: “We have not yet figured out how to steer or control a potentially superintelligent AI and prevent it from going rogue. Humans will not be able to reliably supervise AI systems much smarter than ourselves, so our current alignment techniques will not scale to superintelligence.”
Since that post, however, OpenAI has disbanded some teams focused on AI safety — including safety for superintelligent systems — and several influential safety researchers have left. Some said they left because of OpenAI’s growing commercial ambitions. The company is currently restructuring to make itself more attractive to outside investors.
In a recent interview, asked about critics who say OpenAI does not focus enough on safety, Altman replied: “I would point to our track record.”
— END —