@Gargron kinda roundabout way of thinking of that, no?

@Gargron fun fact: this technology appeared in Vernor Vinge's "Zones of Thought" sci-fi novel series in the early 1990s.

There, it was hijacked in such a way that a video call was "faked" by only information about what the person was saying, and the receiving AI reconstructed an image automatically based on previously seen frames.

@Gargron it's a pretty good idea in my opinion. Maybe with many Keyframes or just 80% of the image reconstructed by machine learning it can be non-distinguishable even with environment changing fast etc.
Just goes to show what could be possible with slow connections, mesh networks etc.

@Gargron this is absolutely horrifying

@Gargron maybe that is the whole point 🤷🏻‍♂️

@gargron
I feel that even ignoring the dangerous implications of this. This would make video meetings even less of a substitute for actually talking to people. A lot of subtle nuance would surely be lost with this right? We already lose so much with body language mostly disappearing out of frame, I think this tech would increase that issue.

@Gargron This is actually really cool. And referencing @kingannoy, it would probably require a good balance between taking enough key points and bandwidth usage. Because I'd assume that given enough key points, we'd be able to pick up on subtle body language.

@Gargron It's like a real-time self deepfake.

@Gargron "Videoconferincing is going through a rennaissance."

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAH