AI Deepfakes: Why You Can't Trust Your Eyes

The technology behind deepfakes is advancing rapidly. Yet, there is no bulletproof way to identify them. Now, Meta is trying something new - but will their idea work?
▸ subscribe to our channel: goo.gl/UupwgM
#ai #deepfakes #fakenews #sora #factchecking #meta #instagram #facebook #threads #artificialintelligence

[@sl4983]

Yes, there's sooo much AI fakes on YT, and others. Including commercials.

[@siddharthb2633]

Somebody will find a way to get rid of these watermarks.

[@samuelsiaw6265]

AI is here to stay and so we better get ready to live with it considering its merits. It’s sad it has its own downsides. Fortunately we are not oblivious to this so let’s keep our eyes opened at all times. Thanks to DW for bringing this up, such a priceless article. ❤

[@sjoervanderploeg4340]

This is so stupid, especially for open source models... you got a script that works like this in layman terms:
input prompt
feed data to model
retrieve image from model
composite a watermark over the retrieved image
save image
Now this is going to be SUPER SIMPLE to remove that watermark, every brain cell I have shrinks a little when I hear regulators talk about this stuff as if they took zero effort or are completely oblivious.