[chalsprhebaodu]

I’ve commented it before, and surely it’s something I’m doing wrong, but I cannot believe system prompts or GPTs or any amount of instructing actually works for people to get ChatGPT to respond in a certain fashion with any consistency.

I have spent hours and hours and hours and hours trying to get ChatGPT to be a little less apologetic, long-winded, to stop reiterating, and to not interpret questions about its responses as challenges (i.e when I say “what does this line do?” ChatGPT responds “you’re right, there’s another way to do it…”).

Nothing and I mean NOTHING will get ChatGPT with GPT-4 to behave consistently. And it gets worse every day. It’s like a twisted version of a genie misinterpreting a wish. I don’t know if I’ve poisoned my ChatGPT or if I’m being A/B tested to death but every time I use ChatGPT I very seriously consider unsubscribing. The only reasons I don’t are 1) I had an insanely impressive experience with GPT-3, and 2) Google is similarly rapidly decreasing in usefulness.

[marviel]

1. Use the API 2. Use Function Calling, with detailed parameters, well named output variables describing the format of the output you want.

You'll get much, much, much better results.

[mpalmer]

My evergreen system prompt prefix:

"You are a maximally terse assistant with minimal affect."

Works well for the most part.

[ojosilva]

Another issue I have, especially when demanding terseness, is that it tends to bail out of writing long code snippets with ellipsis comments like "// And more of the same here" which sometimes defeats the purpose. Except when the code is illustrative to a concept, I want it to be thorough and code the damn thing to the last semicolon.

My solution, which works sometimes, is to instruct it to "not write comments in the code." The drawback is that ChatGPT normally does a good job adding comments, but not something I can't live without.

This "code-trimming" effect does not show up for me in API requests.

[flir]

I second "terse". Damn useful word.

"No moral lectures. No need to mention your knowledge cutoff. No need to disclose you're an AI. Be detailed and complete, but terse."

Gonna try rewriting in the second person, based on your prompt.

I often feel like I'm trying to undo the damage done by OpenAI though. The API doesn't seem to need this crap.

[hackerlight]

OpenAI really should fix this. I've started using Bard and brevity comes out of the box. When I used ChatGPT I always had this background feeling of irritation at the ridiculously verbose responses.

[Manouchehri]

Using JSON mode with the GPT 3.5/4 API works well for us. So much so that we have to intentionally fake errors to test that our retries/fallbacks actually work in our code.

[BOOSTERHIDROGEN]

Have you compared this to chatgpt plus?

[Manouchehri]

ChatGPT Plus does not expose JSON mode in their web UI. You have to use the API via OpenAI (or Azure OpenAI).

[crooked-v]

I would assume a lot of that has to do with whatever obsequieous nonsense they've got in the RLHF 'safety' training, and you're not getting rid of that without pushing it into a totally different context via DAN-like 'jailbreaks'.

[Solvency]

It wasn't always like this. GPT in early 2023, hell late 2022, was incredible. I could have it fully stimulating a Unix terminal on acid for hours, it'd never break character. It's so insanely nerfed now.

[muzani]

It's insanely good every time they have a public release, then deteriorates significantly. There's plenty of evidence around this too - just compare the exact same prompt then and now. Not sure if this is a matter of cost or just playing whack a mole with unintended behaviorial bugs.

[xyproto]

I have a similar experience.

Only asking for things I expect it to be able to find online helps a lot, though.

The moment I try to be innovative or mix two ideas into something new or novel, it falls to pieces in the most frustrating way.

[ben30]

I have this as my custom prefix for when talking to gpt:

Cut unnecessary words, choose those that remain from the bedrock vocabulary everyone knows and keep syntax simple. Opt for being brusque but effective instead of sympathetic. Value brevity. Bullet points, headings and formatting for emphasis are good.

[ojosilva]

Unfortunately, in my experience, as the chat session advances it seems to forget these instructions and become its old apologetic self again.

[losteric]

Do you thumbs down the bad responses?

[chalsprhebaodu]

Religiously.

[supriyo-biswas]

I wonder if people only click on the thumbs down button, thus serving to only provide a negative signal mechanism with no ability to differentiate a positive from a negative one.

[TrickardRixx]

Anecdata: I’ve several times clicked “regenerate” and then thumbed-up the new response when nudged.

[JohnBooty]

It's certainly better than not doing it, but I wonder how much that helps?

I mean, there's no control sample. It's a single custom-generated response read by a single person. I'd like to know how they derive useful insights from those votes.

[muzani]

The hack is using GPT-3, and I don't mean 3.5. It still performs to a production level, at least for creative work. It's been sped up and is significantly cheaper.

[kirkarg]

I'm not sure about web based service but with the API this is easily achievable by tinkering with the system message.

[adamgordonbell]

Share some chats. It will be instructive for others and maybe somebody has a solution.

[kromem]

A fun bug is that ChatGPT will always use an emoji when apologizing. So if you ask it not to use emojis in a chat and it does (which it often will do in promising not to), and point it out, it results in a loop of apologies and self critique that devolves into modeling an existential crisis.

[JohnBooty]

That's interesting. I've seen a lot of apologies from ChatGPT-4 and I don't think I've ever seen an emoji.

I've never asked it not to, either.

[Solvency]

This isn't even remotely true. I've never once seen an emoji from it in over a year of daily use.

[lulznews]

Yea why did they build in the wishy washy wokeness … sigh. Very difficult to get succinct answers from it.