Recently Hank Green has gotten into some hot water for using a very odd, very out of place AI-ism in one of his videos. Hank has claimed he only uses AI for "research" -- setting aside the idea of using a probabilistic hallucination machine for anything that could be described as research. This AI-slip strongly suggest the AI is being used for script writing as well. To say the least, he'd heavily pulling directly from the AI output into the script for that AI-ism to make it through.
However, I would like to suggest that it goes much, much farther.
If you watch the clip, you'll notice a jump cut essentially between the end of one paragraph and the start of the next. In itself, this is not all that suspicious. Anyone who's edited videos know that recording is really hard. You generally record the script in chunks with jump cuts between those chunks so that if you screw up you don't ruin your whole clip. Usually, you do each chunk several times. The thing is, this would mean that he started the paragraph with this blatant AI-ism multiple times and never noticed it was strange during recording. He didn't choose to do any in the moment script corrections, and he didn't make note or correct it during editing. I call bullshit.
Far more likely, in my opinion is that Hank Green has access to AI Influencer software with a model of himself and his voice. I propose that he generates a script then feeds it into the pipeline to create a video of himself speaking the script with some basic editing style triggers like paragraph jump cuts. I'm not saying he does this on every video, rather I suggest that he's making occasional low quality / low effort AI videos to effectively boost his ad revenue from his most loyal viewers while making the low effort nature of the videos obvious enough to not be of wide interest and avoid more viral attention.
For those who are not aware, AI influencers are extremely sophisticated these days and do not require the intensive computation they once did. It is likely that you cannot distinguish many of the AI Influencers from real people, and it would not be surprising at all if Hank Green had access to marginally more advanced models to fool the people who think they're good at picking out an AI. What AI Influencer start up wouldn't want to collaborate with an internet celebrity? Lots of data for a model and more potential revenue than starting from scratch.
Has anyone seen any hints of AI-isms in his other lower quality videos?
2 Comments
PunnyName@lemmy.world · 4 pts · 20d
Do you have any proof to your assumptions, or should we start avoiding that entirely?
BillyGruff82@lemmy.dbzer0.com · -2 pts · 20d
Yeah, the video clip of him saying 'I appreciate the push back' is a piece of evidence that supports the theory because it doesn't make sense in any other context. It's fine if you don't believe me from one piece of evidence, I genuinely get that, but also I'm not obsessed with Hank Green. I'm not willing to dig through every video he's made in the past three years to build up a case.
I'm sharing the theory because I think it's something that is now well withing the realm of reasonable possibilities, and I thought it'd be something people would want to keep an eye out for. Clearly, a large number of people don't think this reasonable. I disagree. I think most people have probably only noticed the poor quality AI Influencers and haven't bothered looking up the higher quality models. There's not going to be much I can do to convince you to look those up.
I get the need for evidence, but I think that's really hard to acquire when deception is point, and it's more work than I'm personally willing to put into the problem. Evidence for any specific case is going to be hard to find for people who aren't AI experts. I suspect in a few years someone more interested in the problem than me will build up a larger collection of evidence documenting where the models slip up, and only after the first big scandal will people take the concern seriously. I'd hoped this would provoke more inquiry, but clearly I'm just coming off as an insane conspiracy theorist -- and I don't know what to do about that other than to just let it go.
halfapage@lemmy.world · 1 pts · 20d