Reminds me of the beginning from the novel “The Swarm” by Frank Schätzing…

While being an environment issue, the plastic wrappings have a practical purpose: protect food from roaches. In many japanese cities you cannot have food open without attracting gokiburi within a few hours. This is also why the japanese keep everything as clean as possible. Even in the shadiest places there is someone with a vaccuum and a stickytape floor roller(!) to prevent the smallest crumb from staying on the floor too long. Eating on the move in the streets is frowned upon, because fallen down crumbs attract roaches. Public trashcans are rare, because - you guessed it - roaches. You are expected to carry any trash back home and put it in a sealed bag in your trashbin. The typical size of japanese houses and flats does not offer much space for storing large food containers, so you buy your food in small portions.

Of course a more environment-friendly wrapping would be better, but it has to be able to withstand a roach nibbling on it, which is not the case for various organic-based polymers.

But…isn’t unsupervised backfeeding the same as simply overtraining the same dataset? We already know overtraining causes broken models.

Besides, the next AI models will be fed with the interactions from humans with AI, not just it’s own content. ChatGPT already works like this, it learns with every interaction, every chat.

And the generative image models will be fed with AI-assisted images where humans will have fixed flaws like anatomy (the famous hands) or other glitches.

So as interesting as this is, as long as humans interact with AI the hybrid output used for training will contain enough new “input” to keep the models on track. There are already refined image generators trained with their own but human-assisted output that are better than their predecessor.