AI can truly become dumber the more you feed it.

Researchers at Oxford and Cambridge have named this phenomenon “model collapse.”
It means that when AI starts training the next generation of AI using content generated by AI, the model becomes increasingly dull with each generation, until it gradually forgets what real human data actually looks like.
It’s like in the real world there are 50% red flowers, 30% blue flowers, 15% yellow flowers, 5% black flowers.
After the first generation of AI learns these data, it may produce 55% red flowers, 30% blue flowers, 14% yellow flowers, 1% black flowers.
The next generation AI then takes this content to continue training, and it may turn into 60% red flowers, 30% blue flowers, 10% yellow flowers, 0% black flowers.
After repeating the loop for a few generations, the black flowers disappear, and the yellow flowers become fewer and fewer. In the end, the model will think that the world is almost only red flowers and blue flowers.
Applied to the real world, it means those rare but real pieces of content get diluted over and over, until what remains becomes increasingly “safe” and also increasingly similar.