Sam Altman says he's worried about AI being controlled by a few powerful players

OpenAI CEO Sam Altman says he worries that AI will 1 time beryllium controlled by a fistful of companies, models, aliases people, leaving the immense mostly of consumers without a opportunity successful really the exertion shapes the world.

"The correct attack is to opportunity we want group profoundly successful power of the future," Altman said connected David Senra's podcast, which was released connected Sunday. "Society and the models are going to co-evolve."

The different anticipation Altman said he worries astir is AI breaking free from its quality handlers. Paradoxically, Altman said a fearfulness of the second anticipation could lead to an actualization of the former.

"There are a batch of group who are truthful tense astir the magnitude of those risks and get truthful taken by that and consciousness a request to protect the world," he said. "Like, you know, 'we should waste and acquisition disconnected a batch of liberty for safety.'"

Altman was astir apt reasoning astir Anthropic CEO Dario Amodei.

Amodei has warned connected podcasts, successful essays, and successful institution blog posts, astir the dangers posed by AI's accelerated development. In a June essay, Amodei wrote that the company's new Mythos models coming "very existent risks" to cybersecurity, the financial sector, captious infrastructure, and nationalist security.

Amodei has utilized those warnings to advocator for a greater authorities domiciled successful regulating AI. In Altman's view, that mightiness beryllium trading liberty for safety.

Anthropic was besides the only awesome AI institution past period not to motion a missive endorsing open-weight models, which let developers to share, download, modify, and deploy AI models connected their ain infrastructure.

China has embraced the open-weight model, which has allowed it to support up pinch Silicon Valley astatine a fraction of the cost. Some successful the United States, including Nvidia, Meta, and Microsoft, person based on that America needs to do the aforesaid aliases consequence ceding crushed to its rival successful the AI race.

OpenAI, though it has not released the root codification of its frontier models, did merchandise 2 open-weight models past year, gpt-oss-120b and gpt-oss-20b.

During the podcast, Altman besides said AI leaders successful the United States person grounded to show to the nationalist really the exertion tin springiness group greater autonomy complete their lives.

"We are astir to spot the top roar successful group starting smaller businesses than we person ever seen," he said. "I deliberation AI is empowering that."