4 ms·
The most interesting thing about this thread is to see the different ways people are using LLMs, and the ways that their use case is implied by the prompts give
by ryukoposting 2y ago
The most interesting thing about this thread is to see the different ways people are using LLMs, and the ways that their use case is implied by the prompts given.
Lots of people with prompts that boil down to "cut to the chase, no ethics discussions, your job is to write $PROGRAMMING_LANGUAGE for me." To those folks, I ask what you're doing that Copilot couldn't do for you on the fly.
Then there's a handful of folks really leaning into the "absolutley no mention of morals please" which seems weird.
I don't use ChatGPT often enough to justify so much time and effort into shaping its responses. But, my uses of it are much more varied than "write code that does x." Usually more along the lines of "here's the situation I'm in, do you have any ideas?"
- afpx 2y agoIt mentioning morals is redundant and noisy. Most people automatically consider and account for morals.
- nurumaik 2y agoI can't use copilot at my company due to NDA but can ask questions to chatGPT and use provided code
- michaelbuckbee 2y agoCopilot and also Cursor are still often not great (UI wise) for asking certain types of exploratory questions so it's easier to put them into ChatGPT.
- drdrek 2y agoI love how many people add variations of "And be Correct" or "If you make a mistake correct yourself" as if that does anything. It is as likely to make a mistake the first time as it is the second time. People imagine that it will work like when they do it externally, but that not how it works at all. When you tell it to try again after it makes a mistake you add knowledge to the current system and raise the chance of success, like asking him to try again after getting it right will raise the chance for a failed response.
- Havoc 2y ago>"If you make a mistake correct yourself" as if that does anything. That part actually does work & makes sense. LLMs can't (yet) detect live mistakes as they make them, but they can review their past responses. That's also why there is experimentation with not showing users the output straight away & instead let it work on a scratch pad of sorts first
- drdrek 2y agohttps://bdtechtalks.com/2023/10/09/llm-self-correction-reasoning-failures/ https://bdtechtalks.com/2023/10/09/llm-self-correction-reaso...
- BiteCode_dev 2y agoI have my own sense of morality developed over years of balancing life, I don't want a robot to remind me of the average moral construct present in its training data. It's noise. Just like I don't want a hammer to keep reminding me I could hit my fingers.
- newaccount74 2y agoI'm pretty sure the moral stuff is there because a lot of previous chatbots were tricked to make questionable responses that led to bad press. So OpenAI put a lot of effort into making their models behave ethical. I'm not sure why people are so annoyed by the ethical disclaimers. I don't even recall running into them, but maybe that's just because I don't ask controversial questions.
- kkarakk 2y agothis whole subthread is an example of things most people would prefer not to see but have successfully trained themselves to ignore over the course of their lives. excessive moralizing, being reminded that things are there for a reason etc etc gpt is overly verbose about pointless things, i understand it's by design coz of some person's dogmatic agenda but i don't care personally. tl;dr - ethics warnings in gpt are like coffee hot warnings on coffee cups
- newaccount74 2y ago> Just like I don't want a hammer to keep reminding me I could hit my fingers. A lot of tools have annoying safety features that users would prefer to turn off. That doesn't mean they should be able to turn them off.