Sam Altman Warns Against AI Control Concentration, Admits He Was Wrong About GPT-4 Disruption

Sam Altman voices fears that control of AI could be centered in too few hands

Sam Altman voices fears that control of AI could be centered in too few hands OpenAI Group PBC Chief Executive Sam Altman is worried that artificial intelligence technology will end up being controlled by just a handful of companies or people in future, resulting in nobody else having any say about how it impacts society. Speaking in an interview with the podcaster David Senra on Sunday, Altman (pictured) warned against a situation in which most people lose their voice as AI changes the world. "The right approach is for people to deeply control the future," he said. "Society and the model will evolve together." By that, he meant that the direction of AI should be steered by social choices, rather than in the direction that some companies want it to go. Altman also said he's worried that AI could one day escape human control, but believes that, paradoxically, the fear of this happening could well enable the first situation to arise. "There are a lot of people who are so nervous about the magnitude of those risks and get so taken by that and feel a need to protect the world," Altman said. "Like, you know, 'we should trade off a lot of liberty for safety.'" Altman's comments were almost certainly a thinly veiled reference to his rival, Anthropic CEO Dario Amodei, who has been one of the most vocal alarmists about the potential risks of AI and has consistently stressed the need for safeguards around the technology. In an essay in June, Amodei warned that the company's then newly developed Mythos model presented "very real risks" to cybersecurity, and could wreak havoc with the world's financial sector, critical infrastructure facilities and national security if it fell into the wrong hands. Because of these dangers, Amodei has become one of the industry's biggest advocates of greater government regulation. But Altman said he disagrees, because if the industry heads in that direction, it's effectively going to be giving up its liberty for safety reasons. It's notable that, last month, Anthropic refused to endorse open-weight models, which can be shared, downloaded, modified and run on independent infrastructure. They're essentially the open-source version of AI. Virtually all of the major Chinese AI companies release their models as open-weight, and by doing so, they have been able to match the capabilities of Anthropic's and OpenAI's best models, and they're far more affordable to run, too. Companies like Nvidia Corp., Microsoft Corp. and Meta Platforms Inc. argue that American model makers should also adopt the open-weight approach, lest they end up losing market share to their Chinese competitors. OpenAI's most powerful frontier models are still entirely closed source, but the company did release a couple of open-weight models last year in the shape of gpt-oss-120b and gpt-oss-20b. So far, Anthropic has not released a single open-weight model. Impact will be huge, but slower to arrive Altman also told Senra that the AI industry has failed to show people how the technology can give people more autonomy and more opportunities. "We are about to see the greatest boom in people starting smaller businesses than we have ever seen," he insisted. "AI is empowering that." However, Altman conceded that this boom will be slower to arrive than he once believed. He admitted that his initial estimate on AI's timeline was wrong, and said society and the economy are going to take more time to adapt to it than he first thought. "I thought when we got to GPT-4, which was back in 2023, that very quickly after that there was going to be much more disruption," he said. For one thing, he said he thought software businesses would become vulnerable to AI much sooner, but now realizes that there's too much inertia in the economy for that to happen overnight. The CEO said he got it wrong because he assumed that organizations would reinvent the way they work around frontier models almost as quickly as they could be improved. AI models gain new capabilities very quickly because their development follows a tight product-release schedule, Altman said. But he now understands businesses can't move that fast, for they have to change software, data access, compliance processes, and think about their budgets, employee responsibilities and customer expectations before they can work out how to take advantage of those new capabilities. Fortunately, Altman believes this is a good thing. He said the lag between development and implementation will make the world's transition "smoother and slower," acting as a stabilizing force that prevents too much economic and societal disruption.

Sam Altman says he expected faster AI disruption after GPT 4, admits he was wrong

Altman also said he is worried AI being controlled by a handful of companies, models or people one day. OpenAI CEO Sam Altman has admitted that he was wrong about how quickly AI would change business and the economy. In an interview with podcaster David Senra, Altman said he expected much more disruption after the launch of GPT 4 in 2023. He believed software companies would face major changes. However, Altman admitted that the pace of change is slower because businesses and society have a natural resistance to sudden changes. Altman said he expected the arrival of GPT 4 to create much more disruption in the software industry. "I thought when we got to GPT-4, which was back in 2023, that very quickly after that there was going to be much more disruption, software businesses up for grabs right away, than it turned out to be," he said. Also read: Apple iPhone 18 Pro may launch with higher price tag: Here is how much it might cost OpenAI CEO believes the slower pace is primarily because of the way the economy works. Businesses often take time to change their systems. "I think I was wrong about a few things, but one in terms of the speed: the economy just has so much inertia," he added. In the interview, Altman also talked about OpenAI's efforts to build Codex, its coding-focused product. He said that the company initially faced a difficult challenge because Anthropic's Claude Code had already gained strong momentum. Altman said, "We were way behind Claude Code, and it seemed like a kind of crazy kamikaze mission to try to beat them with a coding app." Also read: OpenAI ChatGPT can now read and send iMessages on Mac, may raise concerns for Apple: Here is why He said the usual business advice would have been to avoid competing with a product that was already ahead. "The consensus is that this kind of thing never works, and you just move on to the next one. It's a fool's errand to try to win when someone else already has momentum in a particular product category," he stated. Despite the early gap, Altman said Codex has since become widely used by some of the best programmers he knows. Speaking at the podcast, Altman also said he is worried AI being controlled by a handful of companies, models or people one day. "The right approach is to say we want people deeply in control of the future," he said. "Society and the models are going to co-evolve."