Beep boop - this is a robot. A new show has been posted to TWiT…
What are your thoughts about today’s show? We’d love to hear from you!
Beep boop - this is a robot. A new show has been posted to TWiT…
What are your thoughts about today’s show? We’d love to hear from you!
@Leo, I posted this a couple of weeks back, but I think this sums up the problem with AI slop perfectly:
The problem isn’t that we have AI generated content, it is the lack of QA on the results, or people just generating stuff at random that distorts things for no reason. If it was used for something clever, fine. But the slop is a real problem and it is drowning out real content.
In the video, as Mike says, the use of AI on these Titanic videos is hideous, distorts a historical event (I mean, over the last 50 years, I haven’t seen one piece of evidence that there were great white sharks swimming through the corridors) and is so stupid. The creator used it to make money from views of the video, but it displaces serious documentaries and other “real” content.
Another one that is very common at the moment is “porch pirates getting their just desserts”, which seems good and fun at first glance, then you realise, it has nothing to do with porch pirates getting caught, it is AI generated slop of people grabbing boxes and them exploding in a fountain of coloured powder or paint. There are some very good genuine videos on this subject, for example from Mark Rober and you get a certain satisfaction from the pirates getting caught and reported to the police, but the AI videos are just empty, the are so obviously AI, many have a Sora watermark! And there is no real satisfaction in watching an AI puppet getting covered in coloured powder or paint.
Another is AI filtered clips from films in Shorts on YouTube, they are filtered to hell and back, probably so they don’t get a take-down from the copyright holders. Or AI generated and read stories with generic AI images, about unfaithful husbands or wives getting their comeuppance, or historical “documentaries”. I saw one about the fuel cans used in WW2 and how the German design was so superior, but the AI narrator and the slop "period” videos were a real put-off, for what could have been a really interesting story.
In the video above, Mike does show some examples at the end of still images brought to life, one has a young boy turn into an old woman and a light turns into a clock, but otherwise they are actually interesting and add to the context of the image. But these things need the prompt author to double check the videos before publishing, to ensure they are correct, but people mostly don’t seem to bother, they just throw out the “slop” as soon as it comes back and hope that they can make a few bucks, before it gets downvoted to hell or pulled.
Again, it isn’t that it is AI generated, it is that it is AI generated, but it has no merit or it is so bad that it is just unwatchable. It is the prompt writers who are at fault and the lack of quality control, before they publish the results. This can be extrapolated to other areas, such as the millions of AI “tunes” on Spotify.