I've probably been building PCs for longer than you've been alive. Consumer usage has not a single thing to do with how data centers (and specifically data centers for AI) are deploying hardware.
You're talking about something completely different and much smaller scale than anything I'm referring to. nVidia is propping up this market by selling its highest end cards. They aren't selling low power cards to data centers.
This has nothing to do with the consumer replacing video cards in their PCs.
Son, I built my first computer when Tandy was new on the block. I've used everything from punch cards to reel-to-reel to work on systems. I've worked in data centers for over 30 years (close to 40 at this point), working on everything from Cray to PDPs to DEC and everything in-between.
I've built and upgraded more data center systems than I can even count, so many they could stack end-to-end to the moon and back. I've seen consolidation happen where the workloads of 100s of machines that took a whole floor were taken down to 1 that took up less than 1% of the floor space and used only 0.001% of the power.
No, they will obviously be replaced. But your original statement - about power consumption - is completely irrelevant, because those new cards are going to draw MORE power as they are more powerful. Your 'perfect' argument is about as convincing as someone's 'perfect' call with Zelensky. Replacing the cards with more powerful cards is more problematic, as it adds more strain to the system, not less. You're trying to make the argument that replacing a card for a newer one solves something, but it doesn't. They aren't saying 'we have 6 cards in this array, we can replace it with 4'. No, the new array will have 6 as well. Can it maybe perform operations better? Sure, but that doesn't mean it isn't drawing more power overall. They are more efficient on a per-calculation basis, but that doesn't mean they aren't drawing more TDP per use hour.
Your assumption that every new generation of card that replaces those 6 are always going to draw more power is flawed. Because that assumes that no improvements to power consumption ever is made. That power requirements only ever goes up never down, which again if you actually know anything about the computer industry, especially data center systems, is 100% incorrect. So either you're talking out your posterior here or you're not being honest and just arguing just to argue because you don't like this decision Disney has made.
Or you put the term limit because OpenAI wanted a limit so they could get Disney to put more money in.
So you really think that Disney with all its lawyers aren't smart enough to put term limits on their own on the use of their IP into their contracts? And that it was OpenAI that requested the term limit just to fleece Disney out of more money and that without that request Disney would have just allowed the use of their IP in perpetuity? Come on, get real, now I know you're talking out your posterior.
So they should just give up and give in, instead of protecting their IP like they've done for literally 100 years?
Also, please point out this lawsuit that they've won against an AI company. They sent a cease and desist to Google, and they sued Midjourney, but that lawsuit has yet to go anywhere, and Midjourney can still pull Disney IP if you know how to backdoor the commands. As I said previously, its basically impossible to pull stuff out of a data set. I've never seen an example of it actually happening. Even when Muskrat continuously neuters Grok, the content is still there, he just puts in restrictions on accessing it, and you can, again, get it to repeat the content if you work at it correctly.
Who said they are giving up? If they were giving up they would drop every lawsuit against every AI company they have pending now. Again if you look at this as just about the AI slop, you're missing the forest through the trees. This isn't just about the dumb AI videos on your Tiktok feed, its about a whole lot more, some of which was discussed already.
If you look at part of the statement of the announcement this was added -
"As part of the agreement, OpenAI has committed to continuing its industry leadership in implementing responsible measures to further address trust and safety, including age-appropriate policies and other reasonable controls across the service. In addition, OpenAI and Disney have affirmed a shared commitment to maintaining robust controls to prevent the generation of illegal or harmful content, to respect the rights of content owners in relation to the outputs of models, and to respect the rights of individuals to appropriately control the use of their voice and likeness."
Do you really think Disney would still be able to do that from the outside, maybe, but would be more effective and have a voice while being on the inside.
Also just because you can backdoor your way into something right now doesn't mean that backdoor exists forever and won't be plugged up.
Also its not impossible to pull things out of data sets, its just not been very practical as of now outside of a test environment. And just because Musk is too lazy to do it in Grok doesn't mean it can't be done. I can tell you from personal experience you can purge data sets of specific content as needed. Many of the early models had to be purged because they weren't producing the correct outcomes. So it is possible that even if the models have Disney IP today, doesn't mean they can't be purged of that later down the line and any related content created from it scrubbed.
