It can follow instructions well. Just give it a spec in a formal, well defined language. runnable language and some reinforcement learning.

It can also handle vague stuff.

That's certainly more powerful than a regular computer language.

you're not wrong. there is a part to this that does feel like a devolution of engineering when we're just asking the computer really really nicely to do the thing we want.

on the otherhand, LLMs are a really easy way to get results that are previously fairly difficult. While i was tooking dinner last night I built a tool that turned movie puns like "the podchowski casters" into an actual director, using llms. it wasn't that hard.

It's a tiny fraction of the amount of work it takes to get people to build software to specifications. There is an entire professional discipline dedicated to this, and we're here just giving a name to the bucket of tools you use to get LLMs to do some of the same things.

idk, its continued attempts at writing down in detail the processes and culture your setup wants to have

its not really a loss to try the writing and at worst you have a better idea of what it is that you want

True. People arguing over prompts is akin to people arguing over which way to pull the lever on a slot machine.

And lies to boot. No, I won't take the bait of "Models can't lie, they don't have intent" either. One can lie by omission, and just watching thinking traces is more than enough to demonstrate the machine is more than happy to deceive end users if it's training set, RLHF, or other harness quirks tell it to. Only difference is that now you don't have code that's as easy to track down to implement the deception as a red flag.

Made me sick the first time I ran a model on my machine. I'll take honest malware over a "well meaning liar" of an LLM any day.