Welcome, MedTech Professionals.

W5 Edition of The MedTech AI Monitor

​

Hey there,

The big news this week in the world has to do with Chinese AI Startup DeepSeek, with some claiming it’s a game-changer that will upend OpenAI and NVIDIA. Let’s set the record straight.

Lower inference costs don’t automatically translate to better models. Research in distillation shows that reducing cost and improving capabilities are increasingly separate challenges.

DeepSeek’s AI progress is solid but overhyped. Lower costs don’t mean better models- capability and efficiency are separate challenges. OpenAI is scaling aggressively and remains ahead. DeepSeek won’t bankrupt OpenAI or NVIDIA, and better models are coming soon. No magic or revolution, just forward steady progress

Have a great rest of your week!

Greg Matson

1. Caranx Medical Submits AI TAVI-TAVR Guide Software To FDA

Caranx Medical has submitted TAVIPILOT Soft, an AI-powered TAVI/TAVR guidance software, for FDA approval. It enhances real-time precision in heart valve implantation, aiming to expand access beyond specialized centers. Compatible with major imaging systems, it will also integrate with Caranx’s TAVI PILOT robotic system, set for human trials this year. With the TAVR market projected to hit $15B by 2033, competitors like Edwards Lifesciences and Abbott are also advancing innovations. [Link]​

2. Use a Military Approach to Get C-Suite Buy-in for AI

Dr. Michael S. Barr, a healthcare consultant and veteran, advocates using the SBAR (Situation, Background, Assessment, Recommendation) framework: a military-derived communication model to secure C-suite buy-in for AI investments in healthcare. Barr’s approach helps align AI proposals with leadership concerns, emphasizing financial ROI, data-driven insights, and clear, structured communication. By tailoring AI discussions to organizational priorities and presenting SMART goals, healthcare leaders can effectively build confidence and drive AI adoption. [Link]​

3. Why DeepSeek’s New AI Model Thinks It’s ChatGPT

DeepSeek’s newly released AI model, DeepSeek V3, has outperformed rivals on benchmarks but mistakenly identifies itself as ChatGPT, even insisting it’s OpenAI’s GPT-4. This likely stems from training on ChatGPT-generated text, raising concerns about data contamination and AI model integrity. Experts warn that such practices can degrade AI quality, amplify biases, and potentially violate OpenAI’s terms of service. While DeepSeek hasn’t disclosed its training data sources, the incident highlights growing challenges as AI-generated content floods the web, making it harder to filter and ensure originality in new models. [Link]​

📌 AI Job Board

​Senior Strategy Consultant - Life Sciences, Pharma and MedTech (NYC), LSC

​Director, MedTech Supply Chain ERP Global Product Owner Loaner & Consignment, J&J

​Head, Clinical Science (J&J MedTech Vision), J&J

​Associate Director, MedTech Ops, Boston Consulting Group

​

​🗞️ In Other News

• ​OCT Medical Devices 2025: European AI Act compliance ‘not a cause for panic’​
• ​LinkedIn co-founder Reid Hoffman connect to launch AI drug design startup​
• ​Mitigating model risk in AI​

The views and opinions expressed in this email are for informational purposes only. They do not constitute legal advice. Readers are encouraged to consult with their own legal counsel regarding any specific legal questions or issues related to the topics discussed.