Hello! I hope you had a great week—and, unlike me, didn’t have to ask ChatGPT why your neck sounds like a bowl of Rice Krispies every time you move it. More below on how many of us are turning to ChatGPT and other AI tools with our health questions.
Then: All your apps would like a word. Spotify, ChatGPT and Claude now want you to talk to them. Plus, a throwback to one of the greatest gadgets of all time, which celebrates its 47th birthday this month.
And coming up next week, my video interview with OpenAI President Greg Brockman. (Check out our clip of me asking him about the Hugging Face hack here.)
How I got a herniated disc isn’t important. (Actually, if you know, please tell me.) What is important: Like a growing number of people, I turned to ChatGPT for help navigating a week of symptoms, appointments, prescriptions and questions—and it was remarkably equipped to follow along.
Three in ten adults use AI for health advice, according to KFF. OpenAI estimates 300 million users ask ChatGPT health-related questions every week. As I said in a Today Show segment this week, AI medical advice has obvious appeal: It’s available 24/7, costs little to nothing and can generate simple answers from complicated links and medical jargon.
But. A big BUT! These tools can be wrong. They’re hungry for data. And they don’t have a complete picture of your health. That same KFF study found 77% of adults are concerned about the privacy of medical information shared with AI—even as 41% of people who use AI for health say they’ve uploaded personal medical information.
On Thursday, OpenAI began rolling out ChatGPT Health to U.S. users after testing it with a small group of users. The feature lets you connect Apple Health, supported medical records and certain health institutions, and then ask questions grounded in that information. Amazon and Microsoft have released health-related AI tools, too.
It’s a Catch-22 for these companies: People are already asking AI medical questions, so making those conversations more informed and safer is better than pretending they aren’t happening. Yet building a dedicated health product encourages people to hand an AI company even more of their most sensitive information.
I don’t have all the answers here, but I do have a few guiding principles for using these tools thoughtfully:
Ask the basics. I didn’t take ChatGPT’s neck-and-back diagnosis at face value, but I did ask what might be causing the pain, how to relieve it and how to sleep better. As NBC News medical correspondent, Dr. Akshay Syal, said on the show, “Use it to learn about potential things going on in your body. Don’t use it to label yourself with a medical condition.” I wish that were obvious, but recent lawsuits show people are trusting these tools that can sometimes get it very wrong.
Watch your data. It’s tempting to upload that full blood panel workup and ask the bot to make sense of it. A safer bet is to paste in only relevant results and strip out your personal data (name, address, Social Security number, etc.) ChatGPT Health lets you log into some of your institutions via MyChart and other common medical-records dashboards. OpenAI says these are securely connected and their information isn't used to train its foundation models or target ads. But once you start talking to the bot about this info it can be part of its memory. And remember AI companies are not generally bound by HIPPA.
Also, don’t forget to use the incognito, temporary mode in these chatbots. (Here’s my piece on that from a few weeks ago.)
AI is not a real doctor. In my Siri video a few weeks ago, I drew a fake rash on my arm with a marker and asked Siri and ChatGPT for medical advice. Siri flatly refused. ChatGPT, meanwhile, suggested heat rash or an allergic reaction. The point was partly to show Siri’s guardrails—and partly to show what a mixed bag these can be. A real doctor wouldn’t have needed long to diagnose me with “ink.”
Showing my fake rash to a real doctor on the Today Show. Credit: NBC News
All these products warn that AI can make mistakes and isn’t a substitute for a medical professional. OpenAI says it has improved its models’ handling of health questions, including recognizing when professional care may be needed and explaining uncertainty.
Throughout my week of asking ChatGPT about my neck, it repeatedly told me to seek medical care if certain symptoms worsened. They did, so I did. Of course, when I got to the doctor’s office, we talked about what ChatGPT said.
Every day, data brokers profit from your sensitive info—phone number, DOB, SSN. You name it, they sell it to the highest bidder. What happens then? Best case: Companies target you with ads. Worst case: Scammers and identity thieves breach those brokers, leaving your data vulnerable or on the dark web.
This is where Incogni comes in. It scrubs your personal data from the web, confronting the world’s data brokers on your behalf. Incogni also helps remove your sensitive information from all broker types, including those tricky People Search Sites.
Help protect yourself from identity theft, spam calls and health insurers raising your rates. Plus, just for being a New Things reader, you get 55% off Incogni using the code JOANNA55.
Talk to Spotify—and Everything Else
A few weeks ago, I wrote about the end of typing and how many now spend their days talking, talking, talking to computers.
Next week, I’ll go deeper on all of this in my interview with OpenAI President Greg Brockman, where we discussed ChatGPT’s new voice mode and our increasingly chatty relationships with machines. Until then, here are this week’s new voice tools to talk to, test out and possibly argue with:
Claude: On Thursday, Anthropic announced that Claude’s voice mode is getting a better brain. It can now tap the company’s more capable Opus, Sonnet and Haiku models. Now, voice mode picks up whichever model you last used in text chat and defaults to its fastest version.
ChatGPT: OpenAI gave voice mode a similar brain boost in its mobile apps a few weeks ago and made it sound more human in the process. On Thursday, it brought voice mode to the desktop, meaning you can now tell ChatGPT what to do across your computer. Finally, someone else to boss around at work.
Spotify: Earlier this month, Spotify launched a new beta feature that lets you have back-and forth conversations with the app’s AI—entirely by voice. Ask it to change your playlist when the rain starts, tell you how your favorite artist got their big break or debate whether Third Eye Blind is the greatest band of all time. Its response to that last one: “I can’t really crown a single ‘best band of all time.’” Coward.
Name: Kevin Radzewicz
What's your old thing? Sony Walkman WM-F 36
What year is your old thing from? Manufactured from 1987-1989
Why do you love your old thing? This was revolutionary to me: portable music, so I wasn't anchored to the immobile stereo anymore. Also, the cassette tape was an international standard. I remember buying a Bon Jovi tape in Yokohama, Japan, and a Dire Straits tape at a souk in a Middle Eastern country. Finally, its greatest value was the ability to play the half-dozen or so bootleg Grateful Dead concert tapes that I had collected. Those were unique.
I’m tired of ending every newsletter with “See you next week,” so…did you know Nintendo was founded in 1889, before powered airplanes and commercial radio broadcasts? It originally made handmade playing cards. 🎴 This signoff is basically the back of a Snapple cap now.
This newsletter was written and curated by Joanna Stern and Adele Lowitz.
|
