Unblinded: A Clinical Trial Podcast Debunking the Myths of AI
Sep 21, 2023
In this discussion, Andy Cooper, CEO of CluePoints and a veteran in clinical data analytics, tackles the misconceptions surrounding AI in clinical trials. He explains the essential differences between supervised and unsupervised machine learning, emphasizing the need for domain-specific models. Andy highlights the potential of AI to predict site issues and streamline operations by reducing manual tasks. He also addresses concerns about data privacy and trust in AI tools, advocating for consistent results over transparency to build confidence in machine learning.
AI Snips
Chapters
Transcript
Episode notes
Choose Supervised Or Unsupervised Wisely
- Use supervised learning when you can label outcomes to teach models faster and more accurately.
- Apply unsupervised methods for anomaly detection but expect noise and longer learning time.
Reduce Edit-Check Noise With ML
- Many manual edit checks are unused and waste resources while ML can surface meaningful anomalies.
- Unsupervised and self-supervised models can reduce noisy checks and highlight real risks.
ChatGPT Failed On Raw Clinical Data
- We threw clinical data at ChatGPT and it produced nothing because it lacked clinical context.
- That led CluePoints to build customized ML that understands clinical data semantics.