Unlike Make-a-Scene by Meta, which produced fun, fully realized images with more user control, and DALL-E 2 with its realistic images, TikTok’s image generator produces more abstract art with colorful blogs that may (or may not) be recognizable to the regular eye.
While artists won’t be flocking to the new generator because of this, the images created could work as fun backgrounds for the more than one-billion-a-day TikTok users.
Text-to-image generators are flooding the Internet as AI research has produced deep-learning models that can create art that is realistic and beautiful, unlike the AI art generated in the past.
Like other new media models, it’s not yet clear exactly what dangers the new technology could create, but we do know a few possibilities…
In TikTok’s case, The Verge tested violent suggestions spoken to the generation, like the assassination of world leaders. The outcome of those tests were blurbs of color with barely recognizable figures. However, the tested suggestions included nudity with the same results if the text specified how the body should appear. So while artists won’t be flocking to TikTok’s model as of now, since they are working with higher-end text-to-image generators, does that cause problems? We don’t know yet.
What we do know is AI is changing the world and if you can think of it, researchers are already working to make it a reality.