1:50 I have been experimenting with making my own simulation where agents interact. Its been very interesting so far but my PC does not have enough Vram to do it quickly. Ill make it work though. The main thing I wanted to experiment with is getting the agents to self reflect, and thusly grow and change over time. My experiments so far is just giving the agent three hypothetical options of activities to do (sleep, watch TV, or go outside) and then having it restate personality traits in a comma separated list. The idea is ill append this updated personality trait list to future prompts for the agent, meaning that its reflection on each activity will effect its response to the next one. When I get them having conversations they could influence eachother too. Should be interesting.
The near future is the prompt not being a trick or impressive. Its just a conversation. If you learned GPT 3 Prompt "secrets" they would be worthless now.
Your help n guidance much appreciated ur trial and error testing it all and sharing refined gems and sharing best use cases properly with us is what inspired and got me into AI when I found your channel over year and half ago. It's been saving grace and then everyone in AI community and finding you all one by one and seeing u come together is awsome. Thank u keep up the great work keep it coming always helpful ❤the OG of getting most out of GPT and prompts cuz gpt loves multi level prompts for more in depth to point outcome...enjoy AI tools we can use and GPT cases in best way or things we didn't think about out box ❤
Man that prompt creator is pretty good. I mean, I still dont think chat GPT can give me *exactly* what im looking for but that GPT has given me a prompt that gets me the closest ive ever gotten *by far.* I am really convinced that AI cannot be creative, especially after this. But it did give me some decent ideas based off the examples I gave it, and the whole super prompt that was generated. Not great but not bad. Some marble to chip away at for something real and original, if you will.
Why cant i use custom gpts with the free version of GPT-4o like DALLE, for example when i want to create an image it says"I can’t create images right now. Can I help you with something else?" ?
So my rule of thumb is to use 70% of the token limit. I have found it grnerally yields mich better recall. I learned this thru experience using Gemini 1.5 one million context length window and past 700,000, the model became extremely slow and recall was not the best.
can oneyone tell what what version of CHATGPT custom GPT's use? It seems still based on the old version. I can't find an option to change GPT's to use GPT 4 o
Your use of citations to prevent hallucinations, or to allow the user to feel more confident in the output, is a simple yet excellent idea. I'm using it now and the results are much better and precise.
Bro just talk to it there are no set rules, just keep in mind one important fact and base your questions on this: if you ask stupid questions and input bunch of unrelated text and wrong grammar, you’ll also get stupid and wrong answers mostly. Start talking to it simple first, ask simple stuff to get a feel of it, how it generally answers, then you can ask him after your base request to give short summarized responses and so on as you like. But you can always look up “useful chatgpt prompts” and just use those but you will just say or read it.
