That is incorrect though, it follows the fallacy that it’s just like a big database where all that (much larger) data is being copied and compressed into. It’s called machine learning and denying the reality of how it works is just not useful.
Imagine you study as an engineer in whatever field, but now laws have been passed that you only licensed the knowledge from university and publishers. If you work you have to demonstrate who you learned it from and then pay royalty fees. Obviously that would be insane for humans, but I do forsee that they will try to do this for machine learning. Because of the argument you made.
So any open source / weight model you find and could run locally (like e.g. deepseek) will now be illegal because you can’t prove where “dey tuerg dur dartae” from.
Thus all potential future gains from AI will be monopolized, while the costs socialized.
That is incorrect though, it follows the fallacy that it’s just like a big database where all that (much larger) data is being copied and compressed into. It’s called machine learning and denying the reality of how it works is just not useful.
Imagine you study as an engineer in whatever field, but now laws have been passed that you only licensed the knowledge from university and publishers. If you work you have to demonstrate who you learned it from and then pay royalty fees. Obviously that would be insane for humans, but I do forsee that they will try to do this for machine learning. Because of the argument you made.
So any open source / weight model you find and could run locally (like e.g. deepseek) will now be illegal because you can’t prove where “dey tuerg dur dartae” from.
Thus all potential future gains from AI will be monopolized, while the costs socialized.