Intellectually Curious
Intellectually Curious is a podcast by Mike Breault featuring AI-powered explorations across science, mathematics, philosophy, and personal growth. Each short-form episode is generated, refined, and published with the help of large language models—turning curiosity into an ongoing audio encyclopedia. Designed for anyone who loves learning, it offers quick dives into everything from combinatorics and cryptography to systems thinking and psychology.
Inspiration for this podcast:
"Muad'Dib learned rapidly because his first training was in how to learn. And the first lesson of all was the basic trust that he could learn. It's shocking to find how many people do not believe they can learn, and how many more believe learning to be difficult. Muad'Dib knew that every experience carries its lesson."
― Frank Herbert, Dune
Note: These podcasts were made with NotebookLM. AI can make mistakes. Please double-check any critical information.
Intellectually Curious
SymptomAI: Conversational AI for Everyday Diagnostic Assessment
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
We explore Google's SymptomAI built on Gemini models, turning AI from a passive chatbot into an active medical interviewer. Using data from 13,000 Fitbit users, the system proactively asks targeted follow-ups and, in blinded tests, delivered diagnostic lists that were more accurate than those from independent clinicians. We also discuss how wearable data correlates with early physiological signals—days before people notice symptoms—hinting at a future where healthcare is proactive, wearable-enabled, and highly accessible.
Note: This podcast was AI-generated, and sometimes AI can make mistakes. Please double-check any critical information.
Sponsored by Embersilk LLC
So it's 2 a.m., right? You've got this uh this mysterious ache. And like three internet searches later, you're entirely convinced you have some rare Victorian disease.
SPEAKER_01Oh, absolutely. The classic late-night symptom spiral. We've all been there.
SPEAKER_00Right. Well, welcome to Intellectually Curious. Today we are doing a deep dive into a massive new Google research study that honestly it might just kill that late-night panic search forever.
SPEAKER_01Yeah, it's a huge leap for we're looking at Symptom AI today.
SPEAKER_00Exactly. A system built on Gemini Flash 2.0. And the goal of this deep dive is to see how AI is, you know, shifting from a passive chatbot into an active, highly accurate medical detective.
SPEAKER_01Aaron Powell And that active part is really the key. The researchers looked at data from over 13,900 Fitbit users.
SPEAKER_00Wow, that's a massive data set.
SPEAKER_01Right. And what makes Symptom AI so different is it doesn't just passively accept your uh your fragmented 2 a.m. rambling.
SPEAKER_00Right.
SPEAKER_01It actually conducts an end-to-end medical interview.
SPEAKER_00So instead of a standard chatbot that just nods and goes, uh, tell me more, it's more like a seasoned investigator.
SPEAKER_01Exactly. It actively cross-examines you.
SPEAKER_00Oh, wow.
SPEAKER_01Yeah, it hunts for those missing clues you didn't even know were important.
SPEAKER_00Yeah.
SPEAKER_01The researchers discovered that when the AI is prompted to proactively ask standard follow-up questions.
SPEAKER_00Which I assume makes a big difference. Trevor Burrus, Jr.
SPEAKER_01A massive difference. It dramatically improves its accuracy in generating a differential diagnosis.
SPEAKER_00A differential diagnosis, meaning uh rapidly narrowing down that final list of possible conditions.
SPEAKER_01Right. It takes the wheel to dig for context rather than just letting you drive the whole conversation. Trevor Burrus, Jr.
SPEAKER_00That's incredible. But uh before we get into how it actually stacks up against human doctors, a quick word about our sponsor.
SPEAKER_01Go for it.
SPEAKER_00This show is sponsored by Embrasilk. If you need help with AI training, automation, integration, or software development, they are the ones to call, uncovering where agents could make the most impact for your business or personal life. Check out Embrasilk.com for your AI needs.
SPEAKER_01A very fitting sponsor for today's deep dive.
SPEAKER_00Right. Okay, so back to Symptom AI. If it's asking better questions, does that actually translate to better answers than a real human doctor?
SPEAKER_01It does. They ran a blinded test where clinical experts evaluated the chat transcripts.
SPEAKER_00Okay, and what did they find?
SPEAKER_01The experts actually preferred Symptom AI's diagnostic lists over those from independent human clinicians. They found the AI to be significantly more accurate.
SPEAKER_00Okay, that makes sense in theory, but I have to push back a little here.
SPEAKER_01Sure. Go ahead.
SPEAKER_00How can a machine outperform a trained doctor? Especially when you, the patient, are giving, you know, incomplete or vague info in the middle of the night.
SPEAKER_01Well, it comes down to how AI processes probability. Human doctors often fall victim to anchoring bias.
SPEAKER_00Like subconsciously latching on to the most obvious symptoms.
SPEAKER_01Exactly, or just your most recent complaint. But Symptom AI leverages what are called distributional priors.
SPEAKER_00Distributional priors? What does that mean in plain English?
SPEAKER_01It basically means it has mathematically absorbed massive statistical patterns from its training data. So it weighs the probability of all your vague symptoms simultaneously.
SPEAKER_00Oh, I see. Without the human bias.
SPEAKER_01Right. It shines brightest in those low confidence, ambiguous scenarios where human context is sparse. It sees statistical threads a tired doctor might just miss.
SPEAKER_00That is wild. So if we take that highly accurate reasoning and we pair it with the physiological data your body is already broadcasting.
SPEAKER_01Like from a smartwatch.
SPEAKER_00Yeah, exactly. What happens then?
SPEAKER_01This is the most staggering part of the research. They correlated Symptom AI's diagnoses with the user's Fitbit metrics.
SPEAKER_00What did the data show?
SPEAKER_01For acute respiratory infections, physiological shifts, like an increased resting heart rate and interrupted sleep, were clearly visible in the data.
SPEAKER_00Aaron Powell Okay, but when? Like as they got sick?
SPEAKER_01Days before the users even reported feeling sick.
SPEAKER_00Wait, days before.
SPEAKER_01Yes, days before the symptoms even consciously registered to the person.
SPEAKER_00Wow. The implications there are just massive for you know expanding global medical access and stopping disease transmission early.
SPEAKER_01Absolutely. It means we're moving toward a highly optimistic era of proactive health care. The AI could notice those subtle changes and initiate a health check-in.
SPEAKER_00Before you even realize you're sick, it completely flips healthcare from a reactive system to a proactive support system.
SPEAKER_01Working in the background of your everyday life to keep you healthy.
SPEAKER_00It totally redefines what going to the doctor even means, which leaves you with this incredible thought to mull over.
SPEAKER_01With that.
SPEAKER_00Imagine a near future where the first sign of a cold isn't a sneeze, but your watch gently checking in to ask how you're feeling.
SPEAKER_01Oh, I love that.
SPEAKER_00Right. We might literally be the last generation to experience getting sick as a sudden unpleasant surprise. Kids could grow up never knowing the feeling of a full blown flu because their personal digital immune system caught it early.
SPEAKER_01A truly fascinating, positive new baseline for human health.
SPEAKER_00Totally. Well, if you enjoyed this deep dive, please subscribe to the show. Hey, leave us a five star review if you can. It really does help get the word out. Thanks for tuning in.