> we have recently launched the world’s first LLM-powered symptom checker
Why make the claim of being the world’s first? ChatGPT was probably first, but even if you don’t consider it because it is more general, it still sounds like others like drgupta.ai launched before you.
Under AuxHealth.io we launched before Dr Gupta (late 2022, very shitty mvp using davinci02 before ChatGPT release), but then spent most of our energy on b2b. This experience did end up helping refine our chatbot significantly.
Relative to chatgpt/gpt4 — it’s a terrible symptom checker because it doesn’t really ask detailed in depth questions and pivots to trying to answer way too early. This is because it’s trained more for question answering and tries to be helpful in suggestions way before it is clinically appropriate. Honestly, this is a problem with Dr. Gupta as well as it asks a bunch of compound questions without pursuing medically relevant info from a not-entirely-forthcoming patient.
I don’t see any new info from this article. I thought the author would buy one to see if it’s real and investigate a little deeper. Everything stated was already publicly available info.
I guess that's not a bug but a feature, it has to do that to be able to handle some situations (e.g. a week where all days have more contributions than the "level" it's currently eating would block it from getting to the other side of the "playfield"). Still, there probably are some situations where it can't reach all the dots?
It looks like it prefers to eat cells in order but it can eat them out of order if going in order is impossible. I assume the timeline under the game is there to show how well it’s doing.
[0] https://codepen.io/jh3y/pen/oNqdmbW