This is why it is so important to be critical and double check everything you generate using image generators and text-based AI.
The smiling thing was probably partly the “photojournalism” thing since those pictures tend to be a little more dramatic and somber. BUT otherwise this is a perfect example of this issue.
And now that Google Image results are all clogged up with AI-generated results it’s getting worse. The good news is this causes problems for the AI models as well, since new versions can’t be easily trained on bulk data anymore - too much of that data would include AI-generated stuff which quickly takes this problem to a level that actually makes the model worthless.
That means there’s actual reason for people that want to use this stuff to take time and deliberately develop data sets with non-stolen and diverse sources, but unfortunately that’s so much more work that it remains to be seen if that’s actually the route they’ll take.
