It is not about being a simp or whatever that means. It is about LLM’s being trained on licensed data and the output being a transformation (literally a collection of functions) of that data. Meaning that it is virtually a certainty that the output without a license notice is a violation on the license of the data it was trained.
Not to mention all the others shit show that AI’s companies do , are trying to accomplish and stand for.
There are many, many reasons to not use AI (the ones you mention as “other shit shows” for example…).
But when the linux, open source, copy-left guys, who only ever use licenses to keep code free and not locked away in some proprietary project and who would claim that copying is not stealing but freeing code any other day of the week, suddenly complain about AI using “stolen data” to make it accessible to anyone, it will never be not funny to me.
[…] who would claim that copying is not stealing […]
I’m not sure what do you mean by this, one is either complying with the license or not.
[…] suddenly complain about AI using “stolen data” to make it accessible to anyone, it will never be not funny to me.
The data is not accessible to anyone, you need to have a subscription to one of these services to even access the output of these LLM’s, not the data itself. The data is closed behind AI companies’ doors.
If someone would take all data out there (licensed or not) and really make it accessible to anyone (like what Aaron Schwartz tried to do) I would not be that mad about it. This is certainly not the case though. AI companies are trying to rent access to this transformed data for profit.
AI companies business implementation of LLM’s goes against FLOSS in practice and in spirit.
Either it gets ruled as fair use or the licence terms of the og code apply at which point GPL licences apply and and all the ai companies gonna become open source
It is not about being a simp or whatever that means. It is about LLM’s being trained on licensed data and the output being a transformation (literally a collection of functions) of that data. Meaning that it is virtually a certainty that the output without a license notice is a violation on the license of the data it was trained.
Not to mention all the others shit show that AI’s companies do , are trying to accomplish and stand for.
There are many, many reasons to not use AI (the ones you mention as “other shit shows” for example…).
But when the linux, open source, copy-left guys, who only ever use licenses to keep code free and not locked away in some proprietary project and who would claim that copying is not stealing but freeing code any other day of the week, suddenly complain about AI using “stolen data” to make it accessible to anyone, it will never be not funny to me.
Copying is not stealing is coined by the piracy community, not FOSS, the FOSS guys are really into their licenses.
I’m not sure what do you mean by this, one is either complying with the license or not.
The data is not accessible to anyone, you need to have a subscription to one of these services to even access the output of these LLM’s, not the data itself. The data is closed behind AI companies’ doors.
If someone would take all data out there (licensed or not) and really make it accessible to anyone (like what Aaron Schwartz tried to do) I would not be that mad about it. This is certainly not the case though. AI companies are trying to rent access to this transformed data for profit.
AI companies business implementation of LLM’s goes against FLOSS in practice and in spirit.
Either it gets ruled as fair use or the licence terms of the og code apply at which point GPL licences apply and and all the ai companies gonna become open source