Hanson addresses the use of his persona by AI, saying that there are about a half dozen youtube channels using his face and voice to make it appear he says things he is not actually saying.

quote:

Originally posted by PASig:
Facebook is now like 85% AI slop

It’s awful

It really is. I really don’t trust so many videos I see anymore to be real. It is hard to tell unless you notice the subtle things in an AI video. There really should be a law to require a disclaimer.

.

quote:

Originally posted by Black92LX:
I think all AI generated videos need to have a big AI watermarked in the middle of the video.
I wish there was a way to just filter out AI content.

I'm sure something automated will come along to filter out the fraud crap. I'd be willing to bet that solution will involve the use of AI to detect said crap. A little touch of irony.

I have researched many popular AI sites by asking for information that already know very well and in many iterations. When the answers are incomplete and/or incorrect = FAIL!!!

I have settled on “Perplexity” as my go-to search engine. Here is the Perplexity advice on AI Video detection…

“ If you are trying to decide whether a video is fake, I would use this order:
1. Find the earliest source.
Search for the first appearance, not the viral copy.
2. Check for Content Credentials.
Use a C2PA/Content Credentials viewer if the file or platform supports it.
3. Look for corroboration.
Multiple independent videos, witnesses, news reports, geolocation, and timestamps matter.
4. Inspect the technical details.
Look at lips, teeth, hands, shadows, reflections, background motion, and audio synchronization.
5. Extract frames and reverse-image search them.
Sometimes a “new” video is an old clip with a new caption.
6. Be suspicious of low resolution.
Low-quality, highly compressed, cropped, or screen-recorded video is harder to authenticate and easier to fake.
7. Treat detector scores as advisory only.
A detector result can justify more investigation, but it should not be treated as proof.”

Sadly, not the definitive answer I wanted, rather a decent answer to what is a difficult issue!

I add that Perplexity footnotes the sources it uses for reference in many cases, allowing for verification of its responses - I check those frequently, worrying about derivatives rather than original sources such as from scientific journals. If you decide to utilize Perplexity, I strongly advise you to do the same: “Trust but verify.”

A couple guys I've been following on youtube who seem pretty sharp, have been appearing in AI videos with their exact likeness, but it's not them after you watch it for a minute or two. It really pisses me off. If the net becomes MAINLY AI generated crap as a substitute for the real thing, I'm done with it.

If in the future if you can't really tell any longer, that will be the day the internet died. Maybe the best thing to happen short of shopping for specific special stuff.

It was obvious that this was going to be a problem when all of a sudden you had a computer asking you to prove that you’re human

The best solution to AI is NOT to use it.

Go back to thinking and do your own work.