The former head of OpenAI’s AGI department explained his departure from the company as a desire for “independence”

“I want to be independent and less critical. So I didn’t want my views to be rejected, rightly or wrongly, just because I was promoting a corporation’s products,” Business Insider quoted Brundage as saying in an interview with the Hard Fork podcast.

In the coming years, Brundage said, the industry will develop “systems that can, in principle, do everything that a person can do remotely on a computer.” This includes mouse and keyboard control or the ability to imagine an avatar as a “real person in a video chat.”

“Governments should think about what this means in terms of taxes on employees or how it can help in terms of investment in education,” he says.

The timing of AGI creation by large companies is the subject of very persistent debate in the industry. Most experts, like Brundage, believe this is a matter of years. But Dario Amodei, CEO of OpenAI’s main competitor Anthropic, believes the first iterations of the technology could appear as early as 2026.

Brundage, who announced his departure from OpenAI last month after just over six years at the company, should understand OpenAI’s timeline better than anyone.

During his time at the company, he advised its executives and board members on how to prepare for AGI implementation. He was also responsible for some of OpenAI’s biggest security innovations, including bringing in outside experts to find potential problems in the company’s products.

OpenAI has seen a succession of prominent data security researchers and leaders, some of whom have expressed concerns about the balance between AGI development and digital security.

Brundage said his departure was certainly not motivated by doubts about OpenAI’s success in the AGI space (“I’m pretty sure there’s no other lab that’s at the same level of standing”), but he did hint that security issues They were becoming increasingly important. obstacles to testing some innovations.

“I couldn’t work on everything I wanted, which was often related to general industry issues. And it’s not just about what we do within OpenAI, but also what rules should exist in general,” he explained vaguely.