Dreamix: Video Diffusion Models Are General Video Editors

dreamix-video-editing.github.io

36 points·Michelangelo11··4 comments

A Google Research paper and a tiny answer to the alarmist cries of the beginning of the end of Google Search or nonsense such as Google being 'a year or two away from total disruption'. [0] without considering their response to OpenAI.

The age of fake is already here thanks to ChatGPT, DALL-E 2, Stable Diffusion and Dance Diffusion and Eleven Labs Voice AI, but what you are looking at with this is another level of deception with AI video generation from images.

So full dystopia it is then. The internet's own proverb still stays true: Don't believe any you see or hear on the internet.

[0] https://twitter.com/paultoo/status/1598434161332981760

AI feels very different from previous waves of disruptive technologies (e.g., Web 2.0, mobile, cloud, social). More than ever, negative consequences are in plain sight. AI proponents do not even deny that this is going to eliminate a lot of jobs. And it is going to produce an endless barrage of generic, soulless, manipulative content. Wouldn’t surprise me to see people retreating from online platforms as a result of that.

Where is the positive vision in all of this?

Perhaps this is the answer to the question “how do we combat the epidemic of loneliness?” —- once the internet truly cannot be trusted, we’ll be forced to once again connect offline (where, at least for now, deepfakes are not possible).

It's a genie in a bottle situation. Short of an apocalypse that bombs mankind back to the stone age, we can't "unlearn" these new skills. Sure, you can try to legislate them away, but you'd be fooling yourself at that point. There was "then" and there's now: we have to learn how to adapt and hopefully end up with a net gain positive change by the end.

Reply on news.ycombinator.com