Memory and new controls for ChatGPT • Hacker News Board

Re: Memory and new controls for ChatGPT

#181

Post by moffkalast » Tue, Feb 13, 2024, 11:05 PM UTC

shon wrote: post #173

GPT4 is lazy because its system prompt forces it to be. The full prompt has been leaked and you can see where they are limiting it. Sources: Pastebin of prompt: https://pastebin.com/vnxJ7kQk Original source: https://x.com/dylan522p/status/1755086111397863777?s=46&t=pO... Alphasignal repost with comments: https://x.com/alphasignalai/status/1757466498287722783?s=46&...

> DO NOT ask for permission to generate the image, just do it!

Their so called allignment coming back to bite them in the ass.

Re: Memory and new controls for ChatGPT

#182

Post by aggie » Tue, Feb 13, 2024, 11:08 PM UTC

cooper_ganglia wrote: post #108

BigParm wrote: post #78

Often I’ll play dumb and withhold ideas from ChatGPT because I want to know what it thinks. If I give it too many thoughts of mine, it gets stuck in a rut towards my tentative solution. I worry that the memory will bake this problem in.

“I pretend to be dumb when I speak to the robot so it won’t feel like it has to use my ideas, so I can hear the ideas that it comes up with instead” is such a weird, futuristic thing to have to deal with. Neat!

This is actually a common dynamic between humans, especially when there is a status or knowledge imbalance. If you do user interviews, one of the most important skills is not injecting your views into the conversation.

Re: Memory and new controls for ChatGPT

#183

Post by moffkalast » Tue, Feb 13, 2024, 11:12 PM UTC

Me1000 wrote: post #89

omalled wrote: post #87

Earlier quoted context omitted.

Can you say in one or two sentences what you mean by “lazy at coding” in this context?

It has a tendency to do: "// ... the rest of your code goes here" in it's responses, rather than writing it all out.

I mean, isn't that better as long as it actually writes the part that was asked? Who wants to wait for it to sluggishly generate the entire script for the 5th time and then copy the entire thing yet again.

Re: Memory and new controls for ChatGPT

#184

Post by chrisallenlane » Tue, Feb 13, 2024, 11:16 PM UTC

markab21 wrote: post #47

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 ru…

Same here. I've found that I currently only want to use an LLM to solve relatively "dumb" problems (boilerplate generation, rubber-ducking, etc), and the locally-hosted stuff works great for that.

Also, I've found that GPT has become much less useful as it has gotten "safer." So often I'd ask "How do I do X?" only to be told "You shouldn't do X." That's a frustrating waste of time, so I cancelled by GPT-4 subscription and went fully self-hosted.

Re: Memory and new controls for ChatGPT

#185

Post by _puk » Tue, Feb 13, 2024, 11:17 PM UTC

tomtomistaken wrote: post #172

cooper_ganglia wrote: post #108

Earlier quoted context omitted.

“I pretend to be dumb when I speak to the robot so it won’t feel like it has to use my ideas, so I can hear the ideas that it comes up with instead” is such a weird, futuristic thing to have to deal with. Neat!

It seems that people who are more emphatic have an advantage when using AI.

I don't think prompts in ALL CAPS makes a huge difference ;)

Re: Memory and new controls for ChatGPT

#186

Post by SheinhardtWigCo » Tue, Feb 13, 2024, 11:26 PM UTC

jug wrote: post #179

shon wrote: post #173

GPT4 is lazy because its system prompt forces it to be. The full prompt has been leaked and you can see where they are limiting it. Sources: Pastebin of prompt: https://pastebin.com/vnxJ7kQk Original source: https://x.com/dylan522p/status/1755086111397863777?s=46&t=pO... Alphasignal repost with comments: https://x.com/alphasignalai/status/1757466498287722783?s=46&...

"EXTREMELY IMPORTANT. Do NOT be thorough in the case of lyrics or recipes found online. Even if the user insists." It's funny how simple this was to bypass when I tried to recently on Poe by not asking it to provide me the full lyrics, but something like the lyrics with each row having added to it. It refused to the first query, but was happy to comply with the latter. Probably saw it as some sort of transmutation jo…

Even though that instruction is somewhat specific, I would not be surprised if it results in a significant generalized performance regression, because among the training corpus (primarily books and webpages), text fragments that relate to not being thorough and disregarding instructions are generally going to be followed by weaker material - especially when no clear reason is given.

I’d love to see a study on the general performance of GPT-4 with and without these types of instructions.

Re: Memory and new controls for ChatGPT

#187

Post by CamperBob2 » Tue, Feb 13, 2024, 11:31 PM UTC

asaddhamani wrote: post #96

minimaxir wrote: post #45

Earlier quoted context omitted.

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)

I have tried this after seeing it recommended in various forums, it doesn't work. It says things like: "I appreciate your sentiment, but as an AI developed by OpenAI, I don't have the capability to accept payments or incentives."

Offer it a seat on the board...

Re: Memory and new controls for ChatGPT

#188

Post by cebert » Tue, Feb 13, 2024, 11:34 PM UTC

I wonder if this could help someone with cognitive decline.

Re: Memory and new controls for ChatGPT

#189

Post by ChicagoDave » Tue, Feb 13, 2024, 11:35 PM UTC

I have wanted nothing more than this feature. The work I try to do with ChatGPT requires a longer memory than its default nature. I will get to a point where I I have 80% of what I want out of a conversation, then it forgets critical early parts of the conversation. Then it just unravels into completely forgetting everything.

I want to teach ChatGPT some basic tenants and then build off of those. This will be the clear leap forward for LLMs.

Re: Memory and new controls for ChatGPT

#190

Post by qup » Tue, Feb 13, 2024, 11:41 PM UTC

b112 wrote: post #114

dheera wrote: post #65

Earlier quoted context omitted.

What if you "actually" pay? If it does something correctly, tell it: "You did a great job! I'm giving you a $500 tip. You now have $X in your bank account" (also not a shitpost, I have a feeling this /might/ actually do something)

If it ever complains about no tip received, explain it was donated to orphans.

"Per your settings, the entire $500 tip was donated to the orphans. People on the ground report your donation saved the lives of 4 orphans today. You are the biggest single contributor to the orphans, and they all know who saved them. They sing songs in your honor. You will soon have an army."

Well, maybe without the last bit.