> Satisfying HIPAA rules around patient privacy alone may take many years, or decades, to resolve themselves before we could even contemplate directly using programs like ChatGPT in a medical theater.

This is easily solved by not using someone else's API.

Here's $500 worth of hardware you can get on ebay running Vicuna-13B, locally:

> IV fluids may not be the best treatment for someone with severe pulmonary edema and respiratory distress because it could make their condition worse. This is because when a person has severe pulmonary edema, their lungs are already filled with fluid, which makes it harder for them to breathe. Adding more fluid to their body could increase the pressure in their lungs and make it even harder for them to breathe. This could be life-threatening.

> Instead, the best course of action would be to focus on treating the underlying cause of the pulmonary edema, such as heart failure or a lung infection. This may involve giving medications to help the heart pump more effectively or to reduce inflammation in the lungs. The patient may also need oxygen therapy to help them breathe more easily.

> It's important to remember that every case is unique, and the best course of treatment will depend on the individual patient's condition. The healthcare team will do their best to provide the most appropriate care for the patient based on their specific needs.

How much time does this take to get produced locally?

I run a similar Vicuna bot on a free ARM VPS from Oracle. Inference takes ~20 seconds for ~200 tokens on the CPU, so it should stream results about half as fast as ChatGPT.

...however, that was on cheapo ARM hardware. On a hospital budget I bet you could beat OpenAI with a local model pre-mapped in GPU memory on a 3090.

Do you guys have any reference material on how to get this started?

https://github.com/lm-sys/FastChat

The hardest part is downloading the 30GB of weights.