[ checks own name ]

Phewph, came real close to being a Daren, there.

It is like the prisoner dilemma. If you think you'll do the right thing enough people will cheat that you are then obligated to cheat also to keep up.

There's a difference.

When Daren goes 80 in a 30, they actually might be punished as that abuse is recognized and penalized.

With LLMs, there's no such thing as consequences for the bad uses, only rewards.

Internet video feeds are chock full of slop now because that is rewarded. Video platforms are even making "AI remix" buttons to accelerate taking an actual video with thought and effort behind it into uninspired slop. People are making knockoffs left and right but the courts are largely ruling that AI is 'transformative' so those knockoffs that a human would get sued over are getting passes. Managers are micromanaging worker use of AI in hopes that maybe they can prove they can fire most of them.

LLMs enable the worst users more than they enable good users. In software development, the responsible operation of an LLM might speed up a developer 20-50%, depending on the context. An irresponsible one that just assumes the output is good will post a whole lot of crap. Same for all fronts, prose, video production, music. People who care about the medium can get some speedups, but people who are just lazy, uninspired, but see an opportunity can drown out the quality content.