4 ms·
Maybe we need to redefine the problem. Some of the suggestions below were about automating the paperwork. Why not automate the treatment? I mean consider how
by dividendium 18y ago
Maybe we need to redefine the problem. Some of the suggestions below were about automating the paperwork. Why not automate the treatment? I mean consider how unscalable it is to have a doctor, an actual person, see each patient. At most they can get through one patient every 15 minutes.
A doctor normally looks at a patient, asks them a few questions to rule out certain conditions and then prescribes a pill or requests a test for more information.
Sounds like the perfect kind of job that software can do if it knew the parameters.
The hard part is training the software, so you'd need doctors to perform treatments in a way that a computer can understand it (codified in some way) and then you need patients to report back on whether or not that treatment worked.
There's your business opportunity and a real opportunity to help people, both the doctors and the patients.
- nradov 18y agoSounds good. This will be practical just as soon as you invent an AI smart enough to make it through medical school and pass the boards.
- yummyfajitas 18y agoAll the AI needs to do to be medically useful is obey the samurai principle: meet/exceed the effectiveness of a doctor on some subset of patients, and `raise GoSeeARealDoctorException` on the rest. I'm sure regulators/politicians/medical establishment won't like it. But the technical problem isn't all that bad (and has actually been solved by machine learning for a few use cases).