NPR's Sacha Pfeiffer talks with CNBC's Ashley Capoot about OpenAI's ChatGPT for Teens feature and the safety concerns it's designed to address.<\/p>\n
Many teenagers use artificial intelligence not just for homework but for companionship. Many parents worry AI could harm their kids' learning and their mental health. Now OpenAI has a new version of its chatbot that it says addresses some of those concerns and gives parents more control. It's called ChatGPT for Teens. We've called Ashley Capoot to learn more about this. She's a reporter who covers artificial intelligence for cnbc.com. Ashley, good morning.<\/p>\n
ASHLEY CAPOOT: Good morning. Thank you for having me.<\/p>\n
PFEIFFER: How is ChatGPT for Teens different than regular ChatGPT?<\/p>\n
CAPOOT: That's a great question. So ChatGPT for Teens is a dedicated chatbot experience for users under 18 years old. So it builds on a lot of the features OpenAI has been introducing over the last year after the company has faced a number of lawsuits and government inquiries over its safety practices. So basically, what this means in practice is that users between 17 and 13, which is the youngest age that the platform allows, will be automatically placed into the ChatGPT for Teens experience. So adults 18 and older will have access to the chatbot's standard features.<\/p>\n
PFEIFFER: How is the company going to know if people are in that age range?<\/p>\n
CAPOOT: That is another great question, and there are a couple ways this happens. So the first is the most obvious. When you sign up to create a ChatGPT account, you have to input your full name and age. So again, users under 18 will automatically be placed into the ChatGPT for Teens experience.<\/p>\n
PFEIFFER: But people often lie about their age.<\/p>\n
CAPOOT: Right. Exactly. So if a kid tries to circumvent that and claims to be older than they are, OpenAI has a second option that they lean on here. In January, they introduced an age prediction model which helps estimate whether a user is under 18, and they do this by tracking a number of behaviors, including usage patterns, how long an account has existed, times of day when a user is active as well as some other indicators.<\/p>\n
PFEIFFER: Very interesting. So let's say a teenager goes to ChatGPT for Teens and says, write my term paper. What's going to happen? Is it different than what ChatGPT would do? Because ChatGPT might write the term paper.<\/p>\n
CAPOOT: Yes. So there are a number of educational features in this experience, and one of them is called responsible homework reminders. So if a child is trying to take a shortcut quickly to get to an answer, ChatGPT for Teens can show them and redirect them to a mode called study mode, which aims to help students work through problems step by step instead of arriving at an answer immediately.<\/p>\n
PFEIFFER: We also know there are a lot of examples where people get chatbots to tell them things they're not supposed to do or to encourage them to do things that are not healthy. So let's say a teenager goes to ChatGPT for Teens and says, find me a liquor store where I can buy alcohol without showing a driver's license. Is there – are there limitations there?<\/p>\n
CAPOOT: Yes. So OpenAI says the ChatGPT experience is designed with stronger built-in safety protections, and there's a few examples that they give. So they say the experience reduces exposure to content that's considered harmful or developmentally appropriate like the example that you gave there. So some other areas that that covers includes protections around topics like eating disorders or self-harm or violence or other explicit content. It also encourages users to take breaks. So OpenAI said it includes, you know, reminders to step away and turn off the chatbot. And OpenAI also said it includes consistent reminders to tell users that they're talking to AI and not to a person.<\/p>\n
PFEIFFER: Well, teenagers and kids in general are really good at finding workarounds that can defeat technology, so we'll have to see whether they do it in this case as well.<\/p>\n
PFEIFFER: That is CNBC's Ashley Capoot. Thank you.<\/p>\n