Google's new AI video tool floods internet with real-looking clips

When initial Sora demos came out well over a year ago they looked impressive at first but still in the uncanny valley territory (current paid Sora offering is nowhere near those demos). Veo doesn't seem to be any better, it just has more features - sound and video.

I agree with your point though. To me it looks uncanny but if somebody was insisting that some generated video was real, I might not have counter-arguments to prove otherwise.

Based on what I've seen of its output, there seem to still be issues with consistency between shots. Items, people, and spaces change a fair bit between shots, and I've not seen any videos out of it yet that utilize long running single shots.

On a scale of "how trust ending is this?" I'd personally rate its current iteration as low-moderate risk. It will be a matter of when rather than if when we do reach the point where any arbitrary footage can be questioned, but I don't think we're quite there.

I am a bit concerned about what this is going to do, or maybe already has done, to aid in scams. People were already falling for AI Elon, never mind video of this quality.

Various writers warned about this years ago before the AI hype. But we keep pushing anyway.

> we came to the conclusion that this pretty much ends trust.

I see a business opportunity in selling trust and a way to watermark. It will have to be a better business and technical model than trust authorities as those have proven unreliable. You have to become your own certification authority and start creating a trust web.

It works perfectly with JS disabled by uBlock Origin (at least in Firefox).