The amount of AI-generated deepfakes that we have seen so far is honestly far lower than what I or most people would have predicted we would see at current capability levels.

The amount of AI-generated deepfakes that we have seen so far is honestly far lower than what I or most people would have predicted we would see at current capability levels.

https://twitter.com/GarrisonLovely/status/1729977755826315473

Do deepfakes actually have that much power? How many people are deceived by it?

My understanding is that the main harm caused by deepfakes so far is actually people getting away with dismissing genuine recordings of themselves saying or doing bad things by falsely claiming they're deepfakes.

To me the refuting of a video forks:

1. "You can't trust this video" where provenance is needed. Thankfully that problem is being worked on.
2. "My followers will believe me" is a social problem that won't be solved easily. Given the smallest particle of doubt allows followers to maintain their believe/disbelief.