playnext.fm

Future Perfect · Vox

Good Robot #4: Who, me?

March 22, 2025·49 min·4 clips
ChatGPT eerily identifies the host's 'normie perspective' as the anchor of her own AI series, leaving her unsettled.
The question is simple. If AI can change ordinary life, who gets to decide how careful the builders have to be? Daniel answers from inside the work. He describes safety testing meant to find out whether models can do harm before they reach users. But there is a catch. The same process that exposes danger may also teach the company exactly what danger looks like. The host gives that tension room. A model can sit locked away in storage, while the company still learns from the test. Daniel does not call OpenAI evil. He says his faith changed because he no longer believed the company was being careful enough as the technology spread. India becomes the warning sign. According to Daniel, a model was deployed there without fully following OpenAI's own safety rules. He raised the concern. In early 2023, he says he told Sam Altman that OpenAI needed to make the work safe and pivot to safety research. Altman said the time was now. Then comes the blunt part: from Daniel's view, that pivot never came. OpenAI kept moving, shifting from nonprofit toward for-profit status and working with the Trump administration on a half-trillion-dollar AI infrastructure plan. Daniel saw the direction harden. If the company would not slow down or use its lobbying power to make the risks feel urgent, leaving was the lever he had left. The episode closes without a neat verdict, with the paperclip maximizer still hanging around as a warning.

As heard by us

A sharp, uneasy finale about OpenAI, AI safety, and who gets to decide how fast the future ships.

Good Robot ends its four-part series by asking who gets to steer powerful AI when commercial incentives reward speed. Its clearest thread follows Daniel, who worked at OpenAI and came to believe the company was not being careful enough.

Read the full review in PlayNext →

Why you'd press play

Want a calm, sharp look at who really controls AI and why OpenAI's choices matter?

Read the full recommendation in PlayNext →
Listen to the show on
Apple PodcastsSpotify
Play clips from this episodeMore from Future Perfect