the8thbit

You can fine tune the models, and you can modify binaries. However, there is no human readable "source" to open in either case. The act of "fine tuning" is essentially brute forcing the system to gradually alter the weights such that loss is reduced against a new training set. This limits what you can actually do with the model vs an actual open source system where you can understand how the system is working and modify specific functionality.

Additionally, models can be (and are) fine tuned via APIs, so if that is the threshold required for a system to be "open source", then that would also make the GPT4 family and other such API only models which allow finetuning open source.

This isn't what Meta releases with their models, though I would like to see more public training data. However, I still don't think that would qualify as "open source". Something isn't open source just because its reproducible out of composable parts. If one, very critical and system defining part is a binary (or similar) without publicly available source code, then I don't think it can be said to be "open source". That would be like saying that Windows 11 is open source because Windows Calculator is open source, and its a component of Windows.

"Eventually though, open source Linux gained popularity – initially because it allowed developers to modify its code however they wanted ..."

I find the language around "open source AI" to be confusing. With "open source" there's usually "source" to open, right? As in, there is human legible code that can be read and modified by the user? If so, then how can current ML models be open source? They're very large matrices that are, for the most part, inscrutable to the user. They seem akin to binaries, which, yes, can be modified by the user, but are extremely obscured to the user, and require enormous effort to understand and effectively modify.

"Open source" code is not just code that isn't executed remotely over an API, and it seems like maybe its being conflated with that here?

What does "soundness" mean here? I love typescript's type system, except for the quirks that are inherited from javascript, which I sometimes find infuriating, and I'm wondering if you're talking about the same thing. For example, the fact that typescript- through javascript- only exposes a single `number` type is so annoying. Sometimes that's fine, but other times it would be nice to be able to say "No, this isn't just any damn number, this is an integer! This is an iterable ID for god's sake!" so you at the very least get a static error when you try to pass in `1.001`, or something that could be `1.001`. And that's just a very basic example of how a collection of number types with more granularity would be an improvement. Especially if they were a little more robust than that and were composable. Imagine being able to type `integer | negativeFloat` or other such wacky composed types. Ideally you could compose your types programmatically, and define shit like "the type is a `float` who's floor value has a % 3 of 0".

Logo?