OpenAI searches for amateurs with ‘good taste’

According to Altman, one of the best paths for non-technical candidates to participate in artificial general intelligence (AGI) development is to work in research recruitment. “We believe that the best research teams are built on context, taste, and a real sense of where the field is going next. Research recruitment is about finding people who will move the field forward, not simply fill a position,” he wrote in a post on X this week.

Tifa Chen, OpenAI’s head of research recruiting, is looking for excellent candidates with non-traditional educational and career backgrounds, especially former founders, Altman added.

Altman is the latest senior leader to see taste as a potential strength for workers in the AI ​​era. Last week, OpenAI President Greg Brockman made a similar point. “Gu is a new core skill,” he wrote on X.

A number of other technology tycoons, including Paul Graham, co-founder of Y-Combinator, also agree with the above view. Graham was one of the first to comment on the importance of this quality. As far back as 2002, he asserted that “taste” is not objective and that “good taste is needed to create good things” in an essay.

In a post on X in early February, Graham expanded his views. He commented: “In the age of AI, taste will be even more important. When anyone can create anything, the biggest differentiating factor is what you choose to create.”

Dane Knecht, Cloudflare CTO, responded that he agreed with Graham and referenced his January post, asserting that taste will be the differentiating factor in the engineering field in 2026. “Building is now easy. The hard part is determining what should and shouldn’t be built,” Knecht wrote.

However, not everyone thinks that humans have an advantage in the ability to identify and judge. In an article about the future of AI in early February, Matt Schumer, co-founder and CEO of OthersideAI, said OpenAI’s GPT-5.3 Codex model does what for the first time feels like judgment or taste.

“I don’t see any reason why taste and orientation are uniquely human characteristics as many people say. If AI is trained based on those things, it can also learn,” Schumer wrote on X.