OpenAI researcher quits over slippery slope of ChatGPT ads – News

OpenAI researcher Zoë Hitzig says she left her position on Monday, resigning over the recent introduction of advertisements inside ChatGPT and what she believes is a move in the wrong direction for the company.

In a guest essay in The New York Times titled, “OpenAI Is Making the Mistakes Facebook Made. I Quit,” Hitzig said she’d spent two years as a researcher guiding safety polices and shaping how AI models were built and priced. Since the introduction of ads, she believes, OpenAI may no longer be interested in addressing some of the bigger issues AI poses to society.

She doesn’t believe ads in themselves are a bad thing – models are expensive to run and ads create revenue. Nonetheless, she still has “deep reservations about OpenAI’s strategy.”

She explained that ChatGPT has “generated an archive of human candor that has no precedent.” Users chat with the product about everything in the world, often about their most personal issues – evident in the million people a week who talk to ChatGPT about mental distress, the hordes who may or may not be afflicted with “AI psychosis.”

Hitzig believes people talk so candidly because they believe the chatbot has “no ulterior agenda.” Their conversations might range from “medical fears, their relationship problems, their beliefs about God and the afterlife.” Her bone of contention, of course, is that this archive of most personal reflections is now ripe for manipulation where advertising is concerned.

She draws comparisons with Facebook Inc.’s early days when the company told its users they would have control over their data and be able to vote on policies. That, she says, didn’t last long, citing the Federal Trade Commission’s investigation that exposed Facebook’s less-than-noble privacy practices. A company starts with the best intentions, or at least seems to be starting with the best intentions, which then devolves into unfettered profit-seeking.

“I believe the first iteration of ads will probably follow those principles,” she said. “But I’m worried subsequent iterations won’t, because the company is building an economic engine that creates strong incentives to override its own rules.”

The ad debate crossed over into the public sphere last weekend during the Super Bowl when OpenAI’s competitor Anthropic PBC ran ads during the game with the tagline, “Ads are coming to AI. But not to Claude.” It depicted AI private conversations with consumers being rudely interrupted by irritating ads. OpenAI isn’t mentioned, but the inference was crystal-clear.

OpenAI Chief Executive Sam Altman responded, saying his company would never run an ad that was quite as imposing as its rival implied. “We would obviously never run ads in the way Anthropic depicts them.” He claims ads are a way of offering AI to people who cannot afford the subscription cost for a more advanced model of ChatGPT.

Hitzig believes ads are a slippery slope. She believes there doesn’t have to be what she calls the “false choice” of choosing the “lesser of two evils” – restrict people without the money to pay for a subscription to ads, or give them nothing at all.

“Tech companies can pursue options that could keep these tools broadly available while limiting any company’s incentives to surveil, profile, and manipulate its users,” she wrote. “So the real question is not ads or no ads. It is whether we can design structures that avoid both excluding people from using these tools and potentially manipulating them as consumers. I think we can.”

The solution? She believes profits can be used from one service or customer base to offset the costs for another service. If that’s not possible, she believes there should be real oversight – “not a blog post of principles” – that ensures user data isn’t mined to manipulate the consumer. A third option, perhaps wishful thinking, might be to put “users’ data under independent control through a trust or cooperative with a legal duty to act in users’ interests.”

Photo: Unsplash