Generative AI is so energy intensive ($$$), that Google is requiring users subscribe to Gemini.
Google is entirely dependent on advertising sales. Ad revenue subsidizes literally everything else, from Android development to whichever 8-12 products and services they launch and subsequently cancel each year.
Now, Google wants to remove web results and just use generative AI instead of search as it’s default user interface.
While I agree in principle, one thing I’d like to clarify is that TRAINING is super energy intensive, once the network is trained, it’s more or less static. Actually using the network isn’t dramatically more energy than any other indexed database lookup.
It’s static, yes, but the static price is orders of magnitude higher. It still involves loading the whole model into VRAM and performing matrix multiplication on trillions of numbers
To be fair, I wouldn’t include “loading the whole model into VRAM” as part of the cost, given they can just keep it in there between different requests, and it might be down to hundreds of billions or dozens of billions instead of trillions… but even after all improvements it should still be orders of magnitude more expensive than normal search, which just makes their decision even crazier
Awesome. Truly spectacular.
Generative AI is so energy intensive ($$$), that Google is requiring users subscribe to Gemini.
Google is entirely dependent on advertising sales. Ad revenue subsidizes literally everything else, from Android development to whichever 8-12 products and services they launch and subsequently cancel each year.
Now, Google wants to remove web results and just use generative AI instead of search as it’s default user interface.
So, like I said: Awesome.
While I agree in principle, one thing I’d like to clarify is that TRAINING is super energy intensive, once the network is trained, it’s more or less static. Actually using the network isn’t dramatically more energy than any other indexed database lookup.
It’s static, yes, but the static price is orders of magnitude higher. It still involves loading the whole model into VRAM and performing matrix multiplication on trillions of numbers
To be fair, I wouldn’t include “loading the whole model into VRAM” as part of the cost, given they can just keep it in there between different requests, and it might be down to hundreds of billions or dozens of billions instead of trillions… but even after all improvements it should still be orders of magnitude more expensive than normal search, which just makes their decision even crazier
Training will never stop, tho.
New models will keep coming out, datasets and parameters are going to change.