Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?
This is your opportunity to ask questions. No question too simple or too silly.
Culture war topics are accepted, and proposals for a better intro post are appreciated.
Jump in the discussion.
No email address required.
Notes -
How do you all interact with LLMs?
I’ve seen a few articles recently noting the rise of AI as a buddy / therapist or whatever. It’s usually beside the point of the article but an implicit notion is that lot of folks regularly ‘chat’ with AI as if it were a person.
Which I find baffling. Outside of the very early novelty, I find this way of interacting extremely boring and tedious, and generally find the fact that AI wants to get conversational with me a general frustrater.
If I’m not using AI as a utility ‘write X, troubleshoot Y, give me steps for Z’, and I’m using it recreationally / casually, it’s more akin to web surfing or browsing Wikipedia than chatting on a forum or whatever. I will use it as an open format encyclopedia and explicitly not as a conversationalist sounding board. And i genuinely find negative value in the fact that the former is constantly interrupted with the attempt to be the latter.
So my question is again, how far outside of the grain am I?
I got it to write some emergency (and therefore generic) substitute plans, and it produced a downloadable doc, which probably saved me three hours, so I appreciate that. Apparently Teachers Pay Teachers is now a bit scammy, and also I'm unwilling to spend my own money on that kind of thing.
I tried getting some advice on a personal project a month or so ago, and GPT 4 kept saying things like "that's awesomely profound and deep!" with each step, which was annoying, but I hear the new model might be better, and also it does better when just told to knock it off, which I didn't try.
GPT 4 hasn't been very useful for conversation, since by default it produces essay length answers (and sycophancy), but I haven't tried any other models.
It's been reasonably useful for summarizing light research and making concept art.
If you go to the personalization settings in the ChatGPT app, you can set custom instructions for how the LLM should behave with you.
Tell it to be less verbose, and to avoid sycophancy. The latter step may or may not work, but GPT-4o is mostly dead now (they were going to kill it entirely, but so many people have become addicted that Altman relented. Big mistake.)
Back when this became an option, I went for:
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link