Most conversations about AI risk are between people who have worked on it for years. On For Humanity #95, host John Sherman does something different: he talks with Dewayne Moore, a Cincinnati dad who, by his own account, learned about it about six weeks earlier. Dewayne found the topic through an audiobook, then listened to nearly every episode of the show.
Sherman's reason for the conversation is that people who are new to the topic have something the experts do not. In his words, the biggest problem in communicating about AI risk is how different the people in the AI safety community are from the people they are trying to reach. What follows is Dewayne's account of how it landed.
What did Dewayne notice first?
Dewayne says what struck him was how openly the people building AI talk about the risks. "They're not running from it," he says of the main players in the industry. That made him wonder whether they might be "sincerely scared."
Their statements are on the record. When Anthropic researcher Jacob Coxon announced in early September that he was leaving the company, he wrote that the people building AI "earnestly believe it could kill us all by the end of the decade," according to TechCrunch. Anthropic's Evan Hubinger replied that he personally puts the chance above 10 percent within the next decade, ABC News reported. Coxon did not give a probability himself.
For Dewayne, the conclusion follows directly. "There's no excuse to keep doing it," he says, and the argument that someone else will build it anyway does not change his view: "So just stop."
How did the people close to him react?
Dewayne says he told his wife the day after finishing the book. "I think I was just trying to make her aware of the risk," he says. She "just wanted no part of it." His parents said they had heard about it in the news. He does not describe any of this as disagreement: "It was more avoidance."
Sherman calls that a rational reaction and offers an image for it. Touch a 200-degree oven and you pull your hand away immediately. "These ideas are like a 5 million degree oven," he says, and the human body has no matching reaction.
"You just make them aware of it, and then they can decide." - Dewayne Moore
Dewayne's view is that telling people is the right thing to do, however they respond: "It's bad to not tell somebody." Sherman adds a line he has heard, that people believe their neighbors more than experts, because a neighbor puts their own credibility behind what they say. He describes the kind of "telephone tree" he would like to see for AI risk, the way schools once passed along snow-day news.
Should the conversation be louder?
Sherman asks whether AI safety voices, whom he describes as brilliant academics presenting "like a college class," should be louder. Dewayne's answer: "Maybe I need to be an alarmist." He compares it to yelling fire in a theater when there is smoke in the hallway, and says that with everyone competing for attention, people have "the attention span of a gnat."
What does he plan to do next?
Dewayne says he wants to keep up with the news, make sure his family and friends know, and call his elected officials for the first time. He ends the conversation hopeful about people and doubtful about institutions. "I'm hopeful we'll figure it out, I really am," he says. "I'm not hopeful that the CEOs are going to wake up and decide to do the right thing."
A note on what was said on air
In the opening, Sherman describes Coxon's post as saying there is a "very real chance" AI kills every human by the end of the decade. The published post says builders "earnestly believe" this, and gives no probability of its own. Dewayne also recalls builders giving ranges like "ten, 30, 50%." We could not match that range to a single published source, so we attribute it to him.
What else did the hosts cover?
- How AI is changing the job market, as Dewayne's family sees it
- Whether data centers and other near-term harms have come up among Dewayne's friends and family, and this month's reports of AI-assisted discovery of flaws in PHP, which PBS NewsHour says is used by about 70 percent of websites
- Dewayne's view that the "we must beat China" argument would not land with his neighbors
- Where Dewayne gets his news, and how hard it is to tell what is real online
- Groups looking for volunteers, as Sherman describes at the end of the episode
The bottom line
This conversation adds no new evidence about AI. It is one person's account of how the message lands when it reaches someone with no background in the field. His reaction, avoidance at home and conviction from the builders' own words, is a single experience, not a survey. It is still a useful reminder that the first-hand statements from people inside the labs do a lot of the persuading.
Guard Rail Now works to make AI risk a kitchen table conversation. If you want to support that work, you can donate to Guard Rail Now.
Have you told someone close to you about AI risk? What did they say? Let us know in the comments on YouTube.
Watch the full conversation: For Humanity #95 on The AI Risk Network
Read more AI safety coverage: The AI Risk Network on Substack