"Visual Elevator Music"

"Generative AI was trained on centuries of art and writing produced by humans. ... When generative AI was left to its own devices [however], its outputs landed on a set of generic images – what researchers called ‘visual elevator music’ ... pleasant and polished, yet devoid of any real meaning. ...

"The findings ... show that the default behaviour of these systems is to compress meaning toward what is most familiar, recognisable and easy to regenerate ... [resulting in a form of] cultural stagnation. ...

"It’s the slow flattening of creativity into polished sameness.

"AI is like a robot that learns by looking at lots of pictures, stories and songs. But it mostly remembers the ones it sees the most. So when it makes new things, it keeps making very similar stuff again and again.

"It’s why so much AI imagery looks the same.

"The algorithm just doesn’t know how to be weird and creative like humans do."

~ Ahmed Elgammal