It seems you are confusing training and inferring in AI. Training indeed takes insane amount of energy, while inferring (including generating) is pretty cheap. One can do it using their computer.

It seems you understand about AI, but not about figures of speech

The energy issue stands even if this video required as little as .1 kWh

Producing such video manually (without AI) would take more energy. Driving a car to buy and deliver decorations, shooting, editing, CGI-ing, rendering, etc. Another order of magnitude of wasted energy.

But in this case the energy is feeding a productive chain that can benefit its surroundings in many ways, even beyond economics. When you put those kWh in a AI, you just fueled a machine with little to zero boost to the productive environment in a global perspective. That's the waste.

In the message above, I give an example of driving a car to buy decorations (that then will be thrown away at some point). What's "productive"/"benefit"-ial about that? Waste of much more energy, and also much more pollution. In either case, the productive outcome is anti-Trump agitation.

Not to mention the effectiveness of this "anti-Trump agitation"...

AI machines denies the roles of the ppl needed to make this possible. It means you won't need the director, the editors, the sales clerk at the local shop, etc. Proccess may not be as sustainable as i wish, but simply excluding them can cause greater harm, with environmental side effects as a bonus

It is possible to explore many different and more sustainable ways to make content. This path can lead to innovation in various stages of the process, and also to the develpment of new artistic concepts, for example.

This is a 3 minute video, not a crappy picture. It was not done using a PC, it took a shitload of supercomputing.

um, nah I can do that much on my gaming computer in a day. The compiling and editing is the hard part. That takes time. mark got called out on it too, which is probably why he posted it.

No you can't lol

Okay yeah, you are right, but my son can LOL He has skillz.

Yeah that boy is gunna bring the next generation

He has a degree in film, too. He could do some major damage. I have sat there and learned soooo much from him about film, film-making, editing, just the whole thing. It is intense what goes into it all.

I never said it was a picture, my point still stands. Again, you confuse training and inferring. It'll likely take like 10 minutes of a single H100 GPU (nvidia) to generate 1 minute of video.

neither needed to happen.

Despite the agenda you're trying to push, the time-frame does not dictate the power consumption

1. I'm not talking about time-frame. I'm only talking about how much energy required (joules). 2. I don't have an agenda, I just don't like misleading information/statements; and I push back against them. I myself never generated a video using "AI".

I think he was talking about creating the scenario into 'being'... manifesting.