> More interesting: in the future with gpt5 and image generation tech, how will we know if anything is real ever again online? Actual question Another question

> More interesting: in the future with gpt5 and image generation tech, how will we know if anything is real ever again online? Actual question Another question... Say we discover a new planet in 2030. Countless low quality, click/ads driven websites are going to report about it using gpt-4 text and dall-e 4 pictures. Now how do we train gpt-5 and dall-e 5 to know about that new planet, yet somehow sort between what's "real" about that planet and what's poorly generated by inferior AIs (by that I mean inferior to the new version of the AI we're trying to train)? Do we now need an AI to first determine what's real and what's not to know how to correctly train the new AI without using polluted data? Mind blown.