YC’s founder-matching service helped medical records AI startup Hona • Hacker News Board

Re: YC’s founder-matching service helped medical records AI startup Hona

#41

Post by all2 » Thu, May 09, 2024, 4:27 PM UTC

bdw5204 wrote: post #40

all2 wrote: post #38

Earlier quoted context omitted.

> Robotics (either to automate household drudgery or to eliminate terrible jobs nobody wants to do) There's a lot of potential for integrating automation into existing heavy industry. Everything from inspection, materials handling, manufacturing, quality assurance. The way forward is 1) knowing an industry well enough to know the pain points, and 2) addressing a pain point in a compelling manner. > Terraforming (to r…

> There's a lot of potential for integrating automation into existing heavy industry. Everything from inspection, materials handling, manufacturing, quality assurance. The way forward is 1) knowing an industry well enough to know the pain points, and 2) addressing a pain point in a compelling manner. One of the top things that comes to mind for me in robotics would be cooking robots. Restaurants have a big "nobody wa…

> Terraforming

Let me see if I can find it, but you might be interested in multilayered permaculture. My mother knows of a guy who has made an in depth study of permaculture and the outcomes are very impressive. If I can find it, I'll drop it here.

Re: YC’s founder-matching service helped medical records AI startup Hona

#42

Post by Zenzero » Thu, May 09, 2024, 11:21 PM UTC

Suppafly wrote: post #37

Zenzero wrote: post #6

Given my experience with LLMs I am skeptical that enough prompt engineering and tuning can adequately summarize patient records. I don't expect an LLM to adequately interpret data in a way that is useful to me. While much is missed on skimming records, an AI summary sounds like a false sense of security.

>Given my experience with LLMs I am skeptical that enough prompt engineering and tuning can adequately summarize patient records. What sort of experience do you have though? I'd say anyone familiar with the use of AI specifically as it related to medical records would probably disagree with you.

I'm both a doctor and a SWE.

Re: YC’s founder-matching service helped medical records AI startup Hona

#43

Post by Zenzero » Thu, May 09, 2024, 11:29 PM UTC

danenania wrote: post #15

Zenzero wrote: post #6

Given my experience with LLMs I am skeptical that enough prompt engineering and tuning can adequately summarize patient records. I don't expect an LLM to adequately interpret data in a way that is useful to me. While much is missed on skimming records, an AI summary sounds like a false sense of security.

Something to keep in mind with every AI startup is that they’re trying to skate to where the puck will be in a year or two. You could still be right, but it’s worth considering what happens to any particular use case if the model gets 2 or 3x smarter.

The difficulty partially lies in that there are numerous different "flavors" of medicine in any given specialty. Coupled with the vast amount of medical knowledge that is known in part or simply hypothesized given current evidence, you are going to run up against doctors disagreeing on how the summary selects what information can be typically left out.

Even for myself, there is individual variation case to case what I would want to see. Sometimes it depends on things that aren't measurable. If my brain is exhausted after a long week, or if I couldn't sleep on Wednesday, I may conduct rounds a bit differently than if I am fresh. An LLM that isn't responsive in the same way can be problematic.