Loading video...
Video Failed to Load
Microsoft's AI can now detect cancer from a $10 tissue sample. For context, every time tumor cells are tested, doctors create a basic microscope slide to study tissue up close. These slides show cell shapes and structures, but they can't reveal which immune cells are actually fighting the cancer.... show more
43,728 views • 6 months ago •via X (Twitter)
20 Comments

Research paper link: I break down stories like this every day in my free newsletter. Keep up with the latest in AI/Robotics in 5 min a day:

the $10 part is wild. half the reason rural hospitals miss stuff is they cant justify sending slides out for specialized analysis. if this works across tissue types thats huge for places without a dedicated pathologist on staff

This is the shift people miss: AI isn’t only creating new capabilities, it’s upgrading the value of existing infrastructure without changing the hardware.

这技术能普及就太好了,10刀真的不贵

@grok how can this be implemented in hospitals

This is the true potential of AI: moving beyond chatbots to high-precision execution. When we bridge the gap between AI analysis and real-world results, everything changes. Precision in diagnostics today means precision in every industry tomorrow. Great breakdown, Rowan.

I love this. But hospitals can’t even fill in basic paperwork. How are they going to run an open source model?

The $10 part is what nobody's mentioning. → Current cancer diagnosis: $1,000+ test + specialist → With AI: commodity hardware + computer vision = same result This is the Path to Abundance playbook for healthcare. GPT moment for diagnostics is closer than you think.

Rowan with another banger

For this technology to reach its full potential, the hospitals need an easy and intuitive user interface and clear step-by-step instructions. Is this already planned!?

The $10 barrier is huge for accessibility, but I'm curious — how confident are doctors in relying on AI scores vs traditional pathology reviews? Trust is the bottleneck, not the cost.

wildest part is every hospital already had these slides sitting around. the data existed, nobody could extract signal from it without a model trained on 40M cells. classic case where the product isn't new data, it's new inference on old data

...which is exactly where the regulatory conversation gets complicated in ways this framing skips over. Detection accuracy is not the same as regulatory readiness. A system that performs well in a controlled research setting still has to clear premarket review, demonstrate performance across diverse patient populations, and answer the liability question nobody has resolved yet: when the AI flags a false negative and a patient goes untreated, who owns that outcome? Current frameworks weren't built for this. The FDA's SaMD classification can technically accommodate a cancer detection tool, but the harder problem is what happens when that tool learns from new data after approval. Post-market surveillance assumes a relatively static product. Adaptive systems break that assumption, and neither the FDA's AI/ML Action Plan nor the EU MDR has fully worked out how to handle a model that shifts its own decision boundaries over time. The $10 price point matters, but it creates a second problem. Cheaper access at scale means deployment in settings with less clinical oversight, which pushes the tool further toward the autonomous end of the spectrum. That's where risk-based classification frameworks start to strain, because the same tool can be low-risk in a hospital with backup review and high-risk in a remote clinic where the AI output is the final word. Cost democratizes access. Regulatory gaps democratize harm too.

This advancement is truly impressive! Harnessing AI to enhance cancer detection can significantly improve early diagnosis and treatment. I also found a helpful video explaining how to transfer Bitcoin to PayPal:

Full gist

Huge breakthrough. If it holds up at scale, this is real-world AI impact.

Watched this space for 2 years. Real unlock: → $10 vs $500+ traditional pathology → Same tech already being applied to lung and breast cancer detection The bottleneck isn't the AI — it's hospital procurement. 18-24 month integration cycles. Opportunity window is right now.

🙏🏽

Microsoft has finally done something positive for the world

cancer detection for 10 bucks fr
