ATX - Health AI, AGI, longevity 🛸 CEO Health Universe (Kleiner Perkins) 🚀 Co-Founder @arrivehealth (Acq Bain Capital) 🤖 Hacked insulin pump. FDA approved.
Latent Space
Joined December 2008
- Tweets881
- Following857
- Followers444
- Likes5.3K
Totally agree. Ongoing monitoring and evaluation is the only way. This is something we do for customers, since a static benchmark isn't enough.
We built this already. Different orgs on Health Universe have workspaces where they can send data/artifacts to other orgs on Health Universe, and the agents in their respective workspaces respond accordingly.
Dan Caron⚡ retweeted
If you’re a physician sitting across from a patient and you hear this:
“My LLM suspects I have XYZ diagnosis based on these symptoms, labs, and so on…”
What is your reaction?
Sorry Mark, your take is incorrect. "About 1 in 5 patients said they have caught an error in their radiology report or other physicians’ notes, with 40% perceiving the mistake as “serious” in nature.
That’s according to a new survey of almost 30,000 patients, led by Harvard Medical School and published Tuesday in JAMA Network Open. Such serious errors, the authors noted, included incorrect medical histories or medication regimens and notes taken on the wrong subject." - Using humans is dangerous.
20% is the generally accepted rate of human errors.
I run a Health AI startup and see how fallible humans are everyday. Not their fault, they are overworked.
1. AI can “see” (multimodal models) and reason in real time (looping) 2. AI can “see” and reason if CPR may be necessary. Currently limited without better autonomous robotics. 3. AI can access more (and potentially private data - i.e Meta chats) to learn patient was being abused by parent. AI already scores higher in empathy in medical setting than clinicians. “No doctor knows what will happen when patient steps in..” they are already using pre-visit AI summarization and AI chat with medical records to enable better care. AI can understand outcomes with vision models and access to labs and verbal/textual human feedback from patient.
