Sounds like they are preparing to “pull an Apple” with more than just pricing there.
Part of the benefit of Apple’s M series is the unified memory model. They’re able to convert that into increased GPU performance because you no longer have to transfer data in and out of VRAM.
But Apple can only pull that off because they control the CPU, GPU, and the OS (specifically the graphics SDK). Writing graphics code in a unified model is quite a bit different from the conventional x86 model.
Intel would need their own equivalent to Metal if they wanted to do a similar move.
I don’t know enough about Vulkan to say if it’s compatible with this kind of approach, but if not then is Intel really up to starting from scratch?
If they got Unreal and Unity on board, I guess that would give them a good chunk of the market right off the bat for new titles, but what about legacy ones?
If I could predict what happens to the tech market when XYZ policy is enacted, I wouldn’t be posting on Lemmy during my tea breaks. Whatever policies end up sticking around, success is gonna require a lot of us having ideas, trying them out, and recombining them.
But I’ll claim this about my personal metric of “success”: If the future of open source looks like copying the extractive data-mining model of big tech and hoping we can shove the entire history of human thought into a blender faster than them, I think we’ve failed.
I don’t see why those are the only two options.
We could update GPL, CC, etc. licensing so that it specifies whether the author intends to allow their work to be used for LLM training. And you could still put a non-commercial or share-alike constraint on it.
Hooray, open source is saved while greedy grubby hands are thwarted.
As it was with standardized testing, so shall it be with personal behavior: the goal is not to inform the student why, but to enforce compliance.