The world was a better place when AI was limited in scope. Like, The most AI was doing pre 2020s was beating the shit out of chess grand masters, and letting niche game nerds do Tool-Assisted Speedruns as an entire subcategory of speedrunning. There might be more than that, but those are the most public examples I'm aware of. It showed us that focused datasets could do some crazy things. Then the wrong people got involved and said "What if I just feed everything it and market it as a replacement for labor entirely?" and now, we're in this fucking mess where it produces slop, and in spite of everyone else who is a professional pointing out how shit its output is, anyone looking to make a buck off of it thinks its wildly successful because as always, they don't have to deal with the fallout, care, or maintenance of whatever the bullshit generator spit out.