145: The Role of Generative vs Non-Generative AI in Medical Diagnostics – 7-Part Livestream 1/7

Send us a text

Generative vs. Non-Generative AI in Pathology: Why the Difference Matters

If we don’t start defining what kind of AI we’re talking about, we risk letting buzzwords replace real science. 🧠
This is where we begin—at the foundation.

Welcome to the first episode of our 7-part Guided Journey through AI in Pathology, inspired by two must-read articles from Modern Pathology that you’ll want bookmarked forever (links below 👇).

In this episode, I clarify one of the most misunderstood distinctions in artificial intelligence: generative vs. non-generative AI. Spoiler: ChatGPT is not the same kind of AI that segments nuclei or detects tumors.

These differences aren’t just academic—they affect how we train, validate, and regulate tools for diagnostics, research, and clinical care.

🔍 Episode Highlights & Timestamps

[00:00] Welcome to our AI journey! Kicking off the series with a few tech hiccups (as always 😅) and giving context for why these 7 episodes matter.

[02:00] Meet the minds behind the Modern Pathology AI series—Drs. Hanna, Pantanowitz & Rashidi—and why their work inspired this podcast.

[04:00] Core question: What is generative AI? And how is it different from traditional (non-generative) machine learning?

[07:00] Real-world pathology examples: From generating synthetic H&E slides to classifying tumor subtypes—what’s what?

[10:00] Use cases broken down:

• Generative AI: Text-to-image models, LLMs, synthetic training data
• Non-generative AI: Segmentation, classification, detection, clustering

[14:00] Visual metaphors to simplify complexity—think: cake baking vs. quality control inspection 🧁🔍

[17:00] Why understanding these types matters for regulation, validation, and ethical use

[20:00] Real lab examples: When generative models hallucinate vs. when non-generative tools are just wrong

[23:00] What pathologists need to know before choosing or deploying AI tools

[26:00] Sneak peek: How these AI types intersect with statistics, ethics, and regulatory frameworks (in upcoming episodes)

📚 Resource from this Episode

📄 Featured Publications:

1. The Evolution of AI in Medicine: Generative and Non-Generative AI Tools
   🔗 ScienceDirect – Article 1
2. Emerging Use Cases of Generative and Non-Generative AI in Clinical Practice
   🔗 ScienceDirect – Article 2

🛠️ Tools & Terms Mentioned:

• Generative AI Models: ChatGPT, Gemini, Stable Diffusion, GANs
• Non-Generative Models: CNNs, SVMs, Decision Trees, Regression, Clustering
• Key Concepts: Bias, hallucinations, prompt tuning, model performance metrics
• Software Mentions: QPath, Techcyte, LLMs (GPT-4, Claude, LLaMA)

This episode sets the tone for everything we’ll explore in the next six sessions. Whether you’re building models or simply want to understand what AI is actually doing in your lab, start here.

🎧 Ready to become AI-literate in pathology? Hit play and follow along.
 Let’s build better tools—with clarity, ethics, and science.

Support the show

Become a Digital Pathology Trailblazer get the "Digital Pathology 101" FREE E-book and join us!