A lot more durable than software engineering in this day and age...
294 karma · joined June 13, 2021
Who is going to enforce this?
I also checked out some other options along the way, and ultimately realized that pretty much all of the options come with enterprise-oriented features that are just added complexity for the self-hosting use case.
Ultimately, I've gotten at least somewhat familiar with all the complexities of Authentik, so I'd have a hard time switching off. Would definitely love to see a solution geared towards selfhosting that's more barebones, though.
If you could do a perfect job of training a GPT on the entirety of published academic literature, the total of what that GPT could spit out would be limited by the knowledge contained in academic literature. At the same time, you'd have created a tool that is cheap and does a good job of synthesizing knowledge/answering questions across all disciplines. The model will never replace the scientists who are working at the very bounds of their fields, but it doesn't have to in order to be extremely useful, even useful enough to replace a majority of knowledge workers.
Just because GPTs can't be smarter than the smartest humans doesn't mean they can't be smarter than most humans.
More broadly, all big 3 cloud providers (Azure, AWS, and Google) have offerings for FHIR data storage and API access, as well as common NLP based healthcare data analysis workflows. Many of these seem relatively new, or as if they have had a lot of recent attention focused on them. I'm definitely interested in how/why these companies (as well as some other VC funded ones, like Medplum), are entering this space with products that are not directly sellable, but are rather things that other tools would have to build upon. It seems like AWS works directly with end-customers to use their APIs to build products, but I'm not sure what Azure and Google are doing.
This one's probably too complex for my use case, but I thought the concept looked very neat and wanted to share.
Based on my experience with model organisms (flies & rats, primarily), it is actually pretty amazing how analogous the techniques and goals used in this sort of research are to those we use in systems neuroscience. At a very basic level, the primary task of correlating neuron activation to a given behavior is exactly the same. However, ML researchers benefit from data being trivial to generate and entire brains being analyzable in one shot as a result, whereas in animal research elucidating the role of neurons in a single circuit costs millions of dollars and many researcher-years.
The similarities between the two are so clear that I noticed that in its Microscope tool [1], OpenAI even refers to the models they are studying as "model organisms", an anthropomorphization which I find very apt. Another article I saw a while back on HN which I thought was very cool was [2], which describes the task of identifying the role of a neuron responsible for a particular token of output. This one is especially analogous because it operates on such a small scale, much closer to what systems neuroscientists studying model organisms do.
[1] https://openai.com/research/microscope [2] https://clementneo.com/posts/2023/02/11/we-found-an-neuron
If you want to get a sense of what a correct, properly labeled example might look like, you should visit the CMS's guidance [1]. Health insurers should be providing documents that follow the format described there, and most importantly, reporting values that are accurate.
I think it's very fair to have differing attitudes/moral thresholds for an impoverished mother shoplifting a week's worth of baby formula versus "a couple in Alabama [which] pled guilty to shifting $300,000-worth of stolen baby formula on eBay".
My reading of the article suggests that the trend discussed is a result of the latter, which is more recent and problematic, and not the former. Comments here discussing the morality of crime or a desire for policy change are missing this distinction.
The difference being that anyone can sift through images and identify good ones, whereas few people can create good images. The amount of time it takes to complete the task may be the same, but the number of people who can do it greatly increases.
Just as I wouldn't have liked to live through the Protestant Reformation and the wars of religion, I'd much rather have spent my adult life in the peace and prosperity of the late 1900s than on the trajectory people agree the world is headed for now.
Seems kind of like insurance - you pay for it when you don't need it so that you have it when you do.
The better analogy would be Musk indiscriminately enabling full self-driving in every Model 3, claiming its Level 5 as is, and encouraging drivers to sit in the back with a blindfold on.