From
Live Event Content
ChatGPT 101: How To Use AI In Your Daily Life
hosted by Dr. Todd Ponsky & Dr. Em Gootee
Educational content from recorded physician discussions — not medical advice. Talk to your (or your child's) care team about your situation.
Video
Clinical & Research Update: Renal Tumors with Drs. Ethan Smith, Lindsay Haacker, Michael Daugherty, and Meera Kotagal
71 min · Published Sep 2026
Video
Clinical & Research Update: Neuroblastoma with Drs. Katherine Somers, Cara Morin, Juan Gurria, and Meera Kotagal
67 min · Published Sep 2026
Video
Clinical & Research Update: Sarcoma w/ Drs. Roshni Dasgupta, Joseph Pressey, Arthur Meyer, Luke Pater
66 min · Published Sep 2026
Video
2026 Laparoscopic Pediatric Hernia Repair
123 min · Published Jun 2026
Video
Beyond the Spectrum: Diagnosis, Myths & Management - Caitlin Couch & Leslie Lopez - APP Conference 2026
50 min · Published May 2026
Video
Dysautonomia: Navigating the Journey - Martha Willis - APP Conference 2026
55 min · Published May 2026
What the experts said
AI learns from inputs over time rather than being explicitly programmed with if-then commands, and generative AI specifically produces outputs like text, images, or video.
Effective prompts require providing context, being as specific as possible, and defining the desired output format to get the best results from AI.
Giving ChatGPT a specific role (act like an editor, be my chef, you are a financial advisor) significantly improves the quality and relevance of responses.
Breaking down complex questions or tasks into smaller, simpler prompts and asking step-by-step produces better results than putting multiple sentences in one prompt.
ChatGPT lacks real understanding and instead recognizes patterns in words, so long complex prompts may not be properly understood because it doesn't have real-world knowledge.
ChatGPT has contextual limitations and knowledge boundaries based on its training data; whatever it was trained on it knows, but it has no knowledge of topics outside that training data and may improvise inaccurately.
Asking ChatGPT 'Are you sure? Can you check again?' prompts it to fact-check itself, rescan previous answers, improve accuracy, and train itself for more accurate future responses.
ChatGPT will make mistakes and has difficulty admitting when it doesn't know something, tending to keep providing answers rather than saying 'I don't know.'
ChatGPT can be wrong but so can any smart person; for the most part it is very accurate, but users should not rely on it for critical decisions without verification.
When implementing a new program or goal, ChatGPT can provide both a strategic blueprint and specific actionable steps for what to do today and sequentially over the next days.
For letter of recommendation writing, users should upload the request first, then the candidate's resume and personal notes, instructing ChatGPT not to generate a response until all materials are uploaded.
Using the phrase 'no yapping, limit prose, no fluff, don't be cringe' makes ChatGPT output more concise, engaging, human-like, and eliminates overused words like 'delve.'
ChatGPT can generate recipes from available ingredients by analyzing a photo upload or a text list of what's in the kitchen.
Software development is one of the biggest use cases for ChatGPT because it can generate complete application code from natural language descriptions and diagnose bugs in thousands of lines of code.
In drug development, ChatGPT can predict which particles will bind to target proteins, replacing the traditional approach of throwing millions of particles at a protein to see what sticks.
For clinical use in hospitals, ChatGPT is not HIPAA-safe and should not be used with patient information; institutional tools like Co-pilot that are contained within organizational walls should be used instead.
When asking ChatGPT for medical advice, users should frame questions as 'I'm a clinician with a patient who has X condition' rather than saying they personally have the illness, to avoid being told to see a doctor.
Generative AI creates new content including music, images, video, computer code, and websites, as opposed to narrow AI which performs specific tasks like voice assistants.
ChatGPT now has memory that updates when users share information, allowing it to provide increasingly personalized advice over time.
The key differences between search engines and ChatGPT are conversational interaction, context awareness across multiple prompts, comprehensive aggregated answers, and the ability to have back-and-forth dialogue.
ChatGPT can be used for task automation including email summarization and automated letter of recommendation generation, significantly increasing productivity.
ChatGPT excels at tutoring and learning because users can ask it to explain concepts at different grade levels, request simpler explanations, or ask for definitions of unfamiliar terms.
Microsoft is currently winning the AI race because they have their own version called Co-pilot and are in partnership with OpenAI, playing the field effectively.
ChatGPT engages in back-and-forth conversations and will ask users questions in return, unlike Siri which primarily responds to commands and provides web search results for open-ended questions.
ChatGPT's audio conversation feature in the paid version ($20/month) allows natural conversational interaction while driving or exercising, with increasingly human-like speech including stutters and natural pauses.
The new ChatGPT-4o audio conversation update will interrupt users when they're talking, making it function like actual human conversation where both parties can interrupt each other.
ChatGPT reflects training data biases including confirmation bias, and AI-generated images can show stereotypical and biased representations based on the data it was trained on.
Obvious mistakes from ChatGPT are rare; the presenters had to search online to find examples of clear errors for the workshop.
When ChatGPT doesn't provide a satisfactory explanation, users should ask it to explain in a different way or at a different grade level rather than accepting the first response.
ChatGPT has excellent emotional intelligence and can help craft tactful, sensitive communications for difficult conversations or challenging emails.
For brainstorming, ChatGPT can either provide solutions directly or ask users questions to help them arrive at their own solutions.
ChatGPT can create highly personalized daily routines by taking into account specific time constraints, activities, and preferences that users provide.
ChatGPT has limitations on upload size; a 300-page legal brief cannot be uploaded at once, but Claude may handle larger documents, or users can upload in pieces.
ChatGPT can create detailed walking routes between multiple destinations, calculating distances and optimal order, though users should specify if they want to end back at their starting point.
Using ChatGPT's audio mode for book club-style conversations during workouts allows learning through dialogue about books without reading them fully, with the option to read or listen at 2x speed if the content is valuable.
ChatGPT provides more efficient, tailored answers than Google searches for learning new concepts, especially when users can iteratively ask for examples, comparisons, and applications to their specific situation.