The biggest UX work in the short term is going to be around giving people the right amount of transparency so they still feel in control. The actual LLM work is relatively easy and impressive.
You just ask them. I literally just prompted ChatGPT (3.5) to think about its answer again and explain what's wrong with it, and it did. Kept going 2 or 3 times, until it was sure it wasn't doing anything wrong, and magically the script it produced at that stage just worked.
Wanting to do some things by hand might be increasingly a thing only people over a certain age do.
What are some techniques that these products use apart from bounding context and prompt engineering? ReAct sounds like it might help: https://react-lm.github.io/
Suppose you configure your inbox such that you unpredictably don't see a small percentage of important emails. Suppose you miss an email from your boss telling you a very important client needs you to fix a critical bug today, and as a result you start working on that bug slightly later when your boss contacts you another way. Your boss might tell you to uninstall your LLM and just read all your email
By analogy it's socially more acceptable to have an outage because AWS is down than because insert_nich_provider is down even if insert_nich_provider has fewer outages. If insert_nich_provider has more outages it's even worse.
"Sorry I didn't see your important email; my AI assistant deleted it"
Do I really want to wave my arms to control an interface? No, they'd get tired immediately. A mouse on a desk where I can rest my arms isn't sexy because it's not new, but it's pretty much ideal.
They would, but only because you're not used to using them much.
Using our arms more would probably be really good for us.
If we had to use our legs and feet to move a mouse then computer programmers might be some of the fittest people on the planet, instead of some of the fattest.
I say this while lying on my bed and tapping with my fingers on my tablet... we're doomed...
Convert your command into the correct syntax.
Preview what the execution of the command looks like.
Review the changes after they occur.
Undo.
I have a difficult time imagining myself trusting something as unpredictable as an LLM to perform any kind of "cleaning up" task on important data like my inbox. Perhaps future developments in utilizing language models in conjunction with less stochastic systems will make this an easier sell, but for now, I'm not seeing it.