Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Memory and new controls for ChatGPT

openai.com

Post reply on HN
Print view View original

41–50 of 270 posts

Re: Memory and new controls for ChatGPT

#41

Post by kraftman »

polygamous_bat wrote: post #35
kraftman wrote: post #32

Earlier quoted context omitted.

Personally i would like a kind of 2D Map of 'contexts' in which i can choose in space where to ask new questions. Each context would contain sub contexts. For example maybe I'm looking for career advice and I start out a chat with details of my job history, then im looking for a job and i paste in my cv, then im applying for a specific job and i paste in the job description. It would be nice to easily navigate to the…

I think it’s less of a 2D structure and more of a tree structure that you are describing. I’ve also felt the need of having “threads” with ChatGPT that I wish I could follow.

Yeah thats probably a better way of putting it. Like a lot of times I find myself wanting to branch off of the same answer with different questions, and I worry that if I ask them all sequentially chatgpt will lose 'focus'.

Re: Memory and new controls for ChatGPT

#42

Post by isaacisaac »

__loam wrote: post #40
acoyfellow wrote: post #33

Earlier quoted context omitted.

I have some success by telling it to not speak to me unless it's in code comments. If it must explain anything, do it it in a code comment.

I love when people express frustration with this shitty stochastic system and others respond with things like "no no, you need to whisper the prompt into its ear and do so lovingly or it won't give you the output it wants"

People skills are transferrable to prompt engineering

Re: Memory and new controls for ChatGPT

#43

Post by blueboo »

lxgr wrote: post #25

This seems like a really useful (and obvious) feature, but I wonder if this could lead to a kind of "AI filter bubble": What if one of its memories is "this user doesn't like to be argued with; just confirm whatever they suggest"?

This is an observed behaviour in large models, which tend towards “sycophancy” as they scale. https://www.anthropic.com/news/towards-understanding-sycopha...

Re: Memory and new controls for ChatGPT

#44

Post by ldjkfkdsjnv »

BonoboIO wrote: post #39

It got so difficult to force ChatGPT to give me the full code in the answer, when I have some code related problems. Always this patchwork of „insert your previous code here“ This is not a problem of the model, but I suspect it is in the system prompt that got some major issues.

They save money by producing less tokens

Re: Memory and new controls for ChatGPT

#45

Post by minimaxir »

bluish29 wrote: post #28

It is already ignoring your prompt and custom instructions. For example, If I explicity ask it to provide a code instead of an overview it will respond by apologizing and then provide the same overview answer with minimal if no code. Will memory provide a solution to that or will be a different thing to ignore?

Did you try promising it a $500 tip for behaving correctly? (not a shitpost: I'm working on a more academic analysis of this phenomenon)

Re: Memory and new controls for ChatGPT

#46

Post by dr_kiszonka »

nafizh wrote: post #38

My use of ChatGPT has just organically gone down 90%. It's unable to do any sort of task of non-trivial complexity e.g. complex coding tasks, writing complex prose that conforms precisely to what's been asked etc. Also I hate the fact that it has to answer everything in bullet points, even when it's not needed, clearly rlhf-ed. At this point, my question types have become what you would ask a tool like perplexity.

You could try Open Playground (nat.dev). It lacks many features but lets you pick a specific model and control its parameters.

Re: Memory and new controls for ChatGPT

#47

Post by markab21 »

I've found myself more and more using local models rather than ChatGPT; it was pretty trivial to set up Ollama+Ollama-WebUI, which is shockingly good.

I'm so tired of arguing with ChatGPT (or what was Bard) to even get simple things done. SOLAR-10B or Mistral works just fine for my use cases, and I've wired up a direct connection to Fireworks/OpenRouter/Together for the occasion I need anything more than what will run on my local hardware. (mixtral MOE, 70B code/chat models)

Re: Memory and new controls for ChatGPT

#48

Post by keketi »

BonoboIO wrote: post #39

It got so difficult to force ChatGPT to give me the full code in the answer, when I have some code related problems. Always this patchwork of „insert your previous code here“ This is not a problem of the model, but I suspect it is in the system prompt that got some major issues.

Every output token costs GPU time and thereby money. They could have tuned the model to be less verbose in this way.

Re: Memory and new controls for ChatGPT

#49

Post by karaterobot »

What's the difference between this and the custom instructions text field they already have? I guess memories are stored with more granularity (which may not make a difference) and it's something the tool can write itself over time if you let it (and I assume it does it even if you don't). Is there anything else about it? The custom instructions have not, so far, affected my experience of using ChatGPT very much.
Post reply on HN