I would imagine it is like transcribing, an industry I was in for a little bit when I was younger. I saw the same transition there and imagine it will be elsewhere. First it's a bunch of people saying "AI can't take our jobs, our jobs are thinking jobs." Then it's "Sure, you could use AI, but there's no real advantage to it because it makes so many mistakes."

But pretty soon after that it's "Why am I paying a transcriptionist $3/minute when I can just have the machine auto-transcribe it and then my admin assistant can just scan it for mistakes."

Even if there still IS a quality difference between great writers and AI product, "good enough" is good enough for most customers, especially if you have to pay professional rates to get better.

▲

rfgplk 5 days ago | parent [-]

Exactly, time amortized LLMs are already unbeatable at this point.

	▲	daveshistory 3 days ago \| parent [-]
		I go back and forth on it a lot myself; and it's not just in the office context. Grandkid's sports club had an AI-made song about the group at Christmas. It was "good enough" for that. Did they steal the job of a local band? Well, in the sense that the club would have had to commission a song. But in reality, the club wouldn't have paid money for that. They won't pay money to commission Anthropic (or in that case I assume something like Suno) to make the song in the future either. They just won't pay the money at all. A lot of "valuable" human work will be replaced, but it won't be profitable for the companies. I bet more stuff is being transcribed now than ever before -- but not much money is being made on it.