Creating Markets, Creating Trustworthiness: Standards for AI
summary
Mindpeak describes an AI that locates and classifies single tumor cells on gigapixel whole-slide images, trained on data totaling about 5–10x COCO in pixels to capture huge inter-lab variability. A joint dinspec for vision AI in health documents concrete development steps: define a narrow data domain, collect cross-lab specimens, validate robustness, and run a clinical validation. Product pipeline targets CE certification in roughly ~6 months and full development plus validation around 12 months; claimed out-of-the-box robustness is the selling point.
AI-written summary · about
chapters (13)
- 0:00AI standards introduction
- 1:26ISO standards overview
- 3:10AI Act and standards
- 4:33MindPeak AI video
- 7:25AI in medical diagnosis
- 10:36Scaling AI challenges
- 13:24Bias in medical diagnostics
- 14:49AI model generalization
- 15:53Training data and models
- 16:56AI standard key steps
- 18:00Collaboration and competition
- 19:22Customer trust reactions
- 20:25Future certification plans
related talks
26mFrom Pixel to Patient: Making Medical AI Work in the Real World
29mCancer diagnostics enabled by machines: finding and classifying single tumor....
36mResponsible AI
24mBreast cancer detection with IBM PowerAI Vision
37mAI in medicine: where do we stand?
47mAI for making decisions in clinical medicine
21mFighting Cancer with AI
06mTransitioning AI from the lab to real-life applications