A trend on TikTok in which “time travelers” interrupt old videos is making waves, and it's making many users uneasy.
These videos feature a well-known video, usually a Vine video, but the video gets cut off, takes odd turns, unexpected people appear in the video, and usually ends up ruining the ending.
The videos often depict a “time traveller” going back in time to avoid the original video, stop the sharing of jokes, and prevent the video from going viral. The time traveller is often a shadowy, sinister figure, and the videos are accompanied by ominous music.
A lot of the videos use beloved old Vines, which just makes them more creepy. One of the first videos might be an old King Bach video with a guy popping out of nowhere.
The videos show a new and creepy way artificial intelligence is being used to create new videos, and while most commenters seemed amused by them, some also pointed out that the videos show the dangers of the technology and how it could change the videos people watch in the future.
Others say it's a way to revive the once-popular Vine video service, which was shut down in 2017 and was in many ways a precursor to TikTok.
Creating videos is relatively easy, but can be quite time-consuming, and many developers use a system called Luma AI, which they call “Dream Machine,” that can create videos from a single still image and a prompt.
Users can capture a classic video and tell it to create a video in which, for example, a threatening man walks into the frame, which can then be spliced together with the original video, complete with threatening music, etc., resulting in the kind of effect seen in many viral clips.
Luma uses a Transformer model trained on video and, like other generative AI, uses patterns in that training data to create new patterns. The company describes this as “a first step towards building a universal imagination engine,” and the videos it generates often resemble dreams or nightmares.
The status of AI-generated videos has sparked controversy on social networks, including TikTok, where many have chosen to allow AI-generated videos to be posted as long as they are clearly tagged as inauthentic, but enforcement of these rules is not always consistent.
