With a lot of the changes we're going through, I think the main difficulty is the rapidity of change being greater than our speed of adaptation.

We could get used to any of these situations, but by the time we do, it's a completely new situation. Perhaps we're just going through something like the industrial revolution and we'll get to settle into some Information Age norms on the other side.

You can outsource what you want and how much you want to AI. A blog post can be anything from entirely written by AI (prompt: "Come up with an interest topic and write a blog post on it.") to written by you and the spelling being corrected by AI (which is great if you're dyslexic).

I like to write what I like to think of as wholly out of my brain, but use AI for research and fact checking, which before AI I'd have used a non-LLM-based (but still somewhat smart, and increasingly so) search engine for.

One of the things that makes me uncomfortable is that it's often hard to tell to what extent AI was used. I've recently butted heads with a good friend when I realized his messages for me were LLM output. I felt offended at first, because my first thought was that the level of effort I was putting into the communication wasn't reciprocated; that he was going the easy route and not even reading my messages; that he was being lazy. He explained that it was all his own thoughts, and the LLM was just an aid to express them more clearly. I appreciated his explaining himself, because it made me think, but also told him I didn't need that, that he was clear enough without it. But then there is nothing I can do; he'll continue to use it, and maybe one day I'll start using it for that purpose too, just like in the 1990s I might have sworn never to use a mobile phone.

Another thing that makes me uncomfortable is that the content goes through the Big Tech. Unless you self-host, which my friend doesn't; he's not even tech-savvy enough to know that DeepSeek's advantage is in it being smaller and therefore self-hostable, and has zero concern about passing his communication with me through the CCP.

I'll share some thoughts as someone who both writes and codes and uses AI:

I've found AI more useful for coding than for writing. AI doesn't produce good writing. You can easily tell ai-driven writing, and it doesn't have a soul or a personality. It might be good for boilerplate, but nothing else.

Nothing I write on SN is ai-generated. That's because my purpose for writing is to gather and express my own thoughts. Only I can do that, AI can't do it for me.

Code, however, doesn't need a soul. It just needs to get the job done. So i've used AI to write snippets of code and I feel fine about it.

I don't think AI will create a reduction in demand for economists, any moreso than any other job.

I do think it could reduce the number of junior positions, or another way to say that is that junior positions will require senior level work.

But that's gonna be true across most knowledge-producing occupations, I reckon.

All of the above, because large portions of our work can easily be done by AI. I'm not even sure which of those would see the greatest reduction. My guess is private>public>academia and public would be first if they responded to market signals at all.