Emily Moin @emilymoin.com · Jun 15 I don't say this lightly but this is antiscientific.
10 likes 5 replies
Replies
In this specific context, I suspect this is here largely for regulatory “don’t deanonymize the data” reasons with health data.
Not meeting the moment at all! That’s the exact type of work we should be thinking about. Prob will be very useful (especially in healthcare). Using a random internet R package for processing data is allowed…but someone doing cool work with a model they built and host locally isn’t??
Are people using LLMs on health data directly? I'm hoping that instead they're at most using AI to write scripts to process the health data?
like whaaat?
I am a huge and I think quite vocal proponent for careful use of 3rd party AI when manipulating data – I essentially do not expose my data to remote models, ever. But restricting the use of a locally-hosted offline model? How is this anything other than ideological animosity?