They’re releasing it open weighs mostly, not open source necessarily.
Not all that different from the freeware model, where you get a binary that you can run, but you don’t really have the building blocks to make it yourself.
Yeah, for me, I am very strict about that. If it is not open source, I will not use it.
I’m aware of the difference between open weight models and open source models, and I will not use open weight models because they are not open source.
Before I learned that I should care, I used closed source software and so now have some closed source software that I still use because of the fact that I got used to it and can’t really easily get rid of it.
AI is the very first technology where I can draw the line immediately and say I will never use a closed source system.
I have switched to open source software and operating systems as much as possible, but because of the fact that I used operating systems and software before I cared about open source, I still have some dead weight to drag around. And with AI, I’m hoping to avoid that issue.
For example, every application on my phone is open source, except for one, and I find that app useful enough that I cannot get rid of it.
source includes the training data. If i gave someone the full source code of a wasm interpreter and a wasm blob that contains all of the logic of the actual application, with no way of building that blob yourself, and call that open source, I’d be laughed out of the room.
That is true, and so I suspect the only truly FLOSS models we will ever see will be highly specialised to a particular task; which is, in the long run, fine by me. A coding llm that comes with a huge corpus of open source training code, maybe even just in a specific language, a speech to text model with a corpus of transcribed creative commons audio, probably a single language or pair of languages if its for translation, etc. That way the datasets, while still huge, could actually be curated by a single community.
They’re releasing it open weighs mostly, not open source necessarily.
Not all that different from the freeware model, where you get a binary that you can run, but you don’t really have the building blocks to make it yourself.
Yeah, for me, I am very strict about that. If it is not open source, I will not use it.
I’m aware of the difference between open weight models and open source models, and I will not use open weight models because they are not open source.
Before I learned that I should care, I used closed source software and so now have some closed source software that I still use because of the fact that I got used to it and can’t really easily get rid of it.
AI is the very first technology where I can draw the line immediately and say I will never use a closed source system.
I have switched to open source software and operating systems as much as possible, but because of the fact that I used operating systems and software before I cared about open source, I still have some dead weight to drag around. And with AI, I’m hoping to avoid that issue.
For example, every application on my phone is open source, except for one, and I find that app useful enough that I cannot get rid of it.
source includes the training data. If i gave someone the full source code of a wasm interpreter and a wasm blob that contains all of the logic of the actual application, with no way of building that blob yourself, and call that open source, I’d be laughed out of the room.
“open source” models usually do exactly this
That is true, and so I suspect the only truly FLOSS models we will ever see will be highly specialised to a particular task; which is, in the long run, fine by me. A coding llm that comes with a huge corpus of open source training code, maybe even just in a specific language, a speech to text model with a corpus of transcribed creative commons audio, probably a single language or pair of languages if its for translation, etc. That way the datasets, while still huge, could actually be curated by a single community.
I use olmo from AI2 which also includes the training data.