I think if you have a market where you don't license distribution for your software mostly because "hey, you can sell that for more, maybe", then changing the market so that everybody has to buy distribution should actually force the middleware price down, if anything, because they're no longer able to segment their market on it.
It's not like the market for middleware changes by this. I honestly don't see it having much of an effect on price. They're gonna take their middleware and go where exactly?
That's correct. This will make it more expensive to make games. However, the games will be better for it, and the games made the most expensive are those with exploitive business models such as lootboxes.
I am very uncomfortable with the idea of "this person or system cannot create, they can only steal". It seems very dehumanizing, and though LLMs aren't humans I could see the argument very easily turned on people. There is nothing specific to LLMs in it.
The people filling the wires with AI content they didn't write aren't fundamentally incapable of creation; they have chosen not to do it, and so are incapable only until they make a better choice.
It's not dehumanising to see the output of someone who isn't creating as uncreative.
There's this famous saying about the design of a bear-safe waste container for American national parks: "there is a considerable overlap between the intelligence of the smartest bear and the dumbest tourist." Similarly, I suspect there is a considerable overlap between the creativity of the least creative human and the most creative LLM. Maybe it's that I've seen the same arguments raised for years about comic artists who traced off other people's work...
This will not remain limited to AIs. Learning requires imitation. Our artistic system is already pretty heavily copyrighted, but it could still be a lot more so, and I don't think this would be good for our culture.
While I certainly like parentheses highlighting and rainbow parentheses, I've programmed Clojure without syntax highlighting and while it’s not as nice as it would be with, it’s fine.
I’ve also written C++ and Java in Notepad long ago. Not ideal, but hardly a problem.
Huh? The images are tokenized in the same way language is and it’s just fed into one single model. Not multiple smaller expert models.
Image gets rasterized into smaller pieces (eg 4x4 pixels) and each of those is assigned a token, similarly how text is broken up into tokens. And the whole thing is fed into a single model.
> Imagine face recognition to work like a text chat, where the PC gets the frame from the camera and writes in the chat: "Who's that? Here's the RGB888 image in hex: ...".
Gamers simply don't have the impression that they're getting value from the "support," rather than getting shafted come end of life.
reply