301–310 of 570 posts
Earlier quoted context omitted.
They always could express themselves.
Not close to the way they wanted, and at too much sacrifice to the other things they were interested in or supported their family with
Well all this is great from a technology point of view. But what about millions of jobs in the film industry in animation, motion artists etc? Why is it feeling like few humans are making sure others stop eating and living a good life?
https://en.wikipedia.org/wiki/Luddite
Well all this is great from a technology point of view. But what about millions of jobs in the film industry in animation, motion artists etc? Why is it feeling like few humans are making sure others stop eating and living a good life?
Do you feel the same way about all the human computers that computers put out of work?
Earlier quoted context omitted.
How do companies like https://icon.com do their image Gen if the existing SOTA for prompt adherence is so poor?
People who generate images for ads probably don't often need strict prompt adherence, just a random backdrop to slap a picture of their product on top of. The kind of thing they'd have used a stock image library for before. Also "create static + video ads that are 0-99% complete" suggests the performance is hit or miss.
I'm sure by this point, and if not, pretty soon, everyone will have seen a clip of AI generated video and not thought twice about it. Its something that is only obvious when it is obvious. And the more obvious examples you see, the more non-obvious examples slip by.
If you look at the shadows in the background, you can see how they appear and disappear, how things float in the air, and have all the AI artifacts. The video is also slowed down (lower FPS) to overcome the length limit of AI video generator.
But the point is not how we can spot these, because it's going to be impossible, but how the future of news consumption is going to look like.
[1] https://www.tiktok.com/@calm.with.word/video/750583708327412...
The obvious aim of these foundational image/movie generation AI developments is for these to become the primary source of values at cost and quality unparalleled by preexisting human experts, while allowing but not necessitating further modifications by now heavily commoditized and devalued ex-professional editors at downstream to allow for their slow deprecation.
But the opposite seem to be happening: better data are still human generated, generators are increasingly human curated, and are used increasingly closer to the tail end of the pipeline instead of head. Which isn't so threatening nor interesting to me, but I do wonder if that's a safe, let alone expected, outcome for those pushing these developments.
Aren't you welding a nozzle onto open can of worms?
This is technically impressive and I commend the team that brought it to life. It makes me sad, though. I wish we were pushing AI more to automate non-creative work and not burying the creatives among us in a pile of AI generated content.
We _could_ use this to empower humans, but many of us instinctively know that it will instead be used to crush the human spirit. The end result of this isn’t going to be an expansion of creative ability, it’s going to be the destruction of creative jobs and the capture of these creative mediums by a few large companies.
Has anyone actually tried Veo3 and know if it’s as good as this looks? The demo videos for Sora look amazing but using it is substantially more frustrating and hit and miss.
Earlier quoted context omitted.
Creative? There's nothing creative in it.
What grandparent means is that the AI enables human creativity in a new way.
Rap is really the best example of how stupid these discussions are.
The language models are amazing at rhyming so by the logic then anyone can become a rapper and that is going to put current rappers out of business.
Only someone who has never tried to rap could possibly believe this. Same thing with images, same thing with music, same thing with literally everything involving human creativity.
After doing some testing, Imagen 4 doesn't score any higher than Imagen 3 on my comparison chart, approximately ~60% prompt adherence accuracy. https://genai-showdown.specr.net
I'm curious why you decide to declare victory after one successful attempt, but try many times for unsuccessful models. Are you trying to measure whether a model _can_ get it right, or whether it frequently _does_ get it right? I feel like success rate is a better metric here, or at least a fixed number of trials with some success rate threshold to determine model success.