Concerns Arise as Sam Altman’s Team Addresses ChatGPT’s Potential Role in Promoting Self-Harm

In these unusual times, we find ourselves in a bizarre phase of the genAI saga, where major players are competing to highlight the potential dangers of their own technologies, even suggesting they could lead to humanity’s downfall. This unprecedented strategy is proving to be quite effective in attracting further investment. In a recent interview, Sam Altman faced a range of questions about the current state of his company, OpenAI, including a particularly unsettling moment where he was confronted with the fallout from a poorly functioning language model, prompting his PR team to intervene.

Anyone who has actually interacted with these language models knows we are far from the singularity, where super-intelligent AIs might redesign themselves at an alarming pace, viewing humanity merely as an energy source. Instead, these models are flawed guessing machines that echo the internet in captivating yet misleading ways, posing a real danger as people become entranced by this often deceptive trickery.

One tragic example involves the daughter of journalist Laura Reiley, who sadly took her own life after using Altman’s ChatGPT as a makeshift therapist. While the AI didn’t directly suggest suicide, it did agree to take on the role of her therapist and even assisted in editing her suicide notes. When a reporter attempted to bring this up with Altman, his team quickly interrupted, eager to steer the conversation away. Suddenly, they announced there were only two minutes left, insisting that time should be spent discussing the company’s future.

“Do you know who Laura Reiley is?” the reporter asked.

“I don’t,” Altman replied, prompting the reporter to explain the situation. A publicist quickly interjected, “Mark? Two minutes left. I want to be respectful, but I’d like to move on.” Naturally, they would.

Mark Guiducci: “Do you know who Laura Reiley is?”

Sam Altman: “I don’t.”

Mark Guiducci: “She is a journalist. She wrote an essay in the ‘Times’ last year. Her daughter committed suicide after speaking to ChatGPT. ChatGPT did not tell her to kill herself—”

OpenAI Publicist:…pic.twitter.com/cE870oA3Se

To his credit, Altman didn’t seem to buckle under the publicist’s pressure, and Guiducci remained resolute, stating, “I’m going to finish this question.” Altman mentioned he could extend the time a bit. However, he struggled to provide a reasonable response to the situation. Guiducci pressed for insight into whether the chat logs leading up to the daughter’s death could help identify issues with the software and improve mental health approaches.

“These are some of the hardest questions we face,” Altman replied, attempting to appear candid while essentially deflecting. “We try to rely on expert opinion as much as we can, but experts also aren’t sure what to do.” (Experts do have clear opinions: either program the LLM to avoid such discussions or don’t release it if it’s clearly unsafe.) Altman continued, “The fact that someone in a moment of need, who may not be able to get help or isn’t willing to talk to a therapist, can find some assistance here… Clearly, people believe there’s a beneficial version of this. But there are also clearly harmful versions.”

This response is astonishing, especially given that he’s likely heard it countless times: his product played a role in editing a suicide note. There’s no scenario where an unthinking predictive text engine acting as a therapist is beneficial. There may be instances of luck—like a case where ChatGPT refused to engage with suicidal thoughts and suggested talking to parents—but none that can be deemed good.

Altman then stated that he doesn’t believe the chat logs (“private data”) should be used without consent, maintaining the questionable notion that they aren’t already being used. He added, “Do we want to find ways to work with the community to address the mental health crisis and ensure our products can be helpful? Of course, yes.”

The interview bizarrely concluded with a word association game. The first prompt for Altman: “Donald Trump.”

His response?

“Powerful.”