The nature of preference-ranking in LLM post training means that "slop" is something that the average person actually prefers. Maybe there's an argument that it's not intentional but just emerges as a form of reward-hacking (e.g. people like a little bit of slop or sycophancy but the model dials it to 11) but I imagine if the end result wasn't rewarded ultimately that would get penalized.

The fact that a clearly sloptacular image won first place seems to corroborate that indeed people do like it. Until people's tastes change and they learn to recognize and immunize themselves against slop, the models will continue to generate it.

▲

internet2000 15 hours ago | parent | next [-]

A parallel I like to draw is Thomas Kinkade and HDR photography. This style has always been popular, even pre-AI.

▲

happytoexplain 14 hours ago | parent [-]

Analogies are lazy rhetorical devices. Different things are, unsurprisingly, different. Saying "X is like Y" does nothing but invite the conversation to take a turn into identifying all the differences between X and Y.

Just defend the original subject on its own merits.

	▲	Barbing 8 hours ago \| parent \| next [-]
		This comment is like an opinion someone had and then shared digitally with the world. - I didn't mind the Thomas Kinkade reference, but you do have me thinking about to what extent analogies are lazy. Do you abstain from using them completely, or keep them rare?
	▲	kdidiwkdkw 14 hours ago \| parent \| prev [-]
		[flagged]

▲

Barbing 9 hours ago | parent | prev [-]

Great points.

Wouldn't it be interesting if we could take this image back to ourselves in 2021 and see how we thought about it? I'm not a visual artist, so I would not have been overly critical of it. Today, AI art has a ton of baggage & as a result, the sloptacular is bothersome to me. But before I could recognize it, before certain styles became played out… probably much less offensive.

Will the generators keep up and become dynamic enough the immunization will never happen at scale?