Dear all,
I’m extremely excited about the idea that each person could allow GPT to continously observe it’s actions and behaviour and build a model of that person in silico. I don’t know yet how it could be done but it would be kind of extreme version of personalization. Currently, when I type something into the prompt, GPT does not have any assumptions about me, it assumes the same for me as for other people typing. Thus, no matter how big and advanced the GPT model is, it can not respond optimaly/most efficient for each person on the Earth from the first prompt. Yes, you can have predefined prompt where you “teach” GPT the personalization but it is not the solution. Imagine, you would have two “send” buttons: 1.start the conversation in default mode (like GPT speaking to an average human now) 2.start the conversation like GPT knows you and your personal attributes, pros and weaknessess from your birth.
I also had trouble to find a place where I could send a general suggestions or feedback about chatGPT to OpenAI. If this place is not the right one, delete my post. I’m electrical engineer who wants to learn python so chatGPT is extremely helpfull. To me, the most important aspect about future GPTs will be the ability to allow no coding people to build software and in turn, new things. In the end, everyone will be a developer. Thank you for your work.
AI tools like chatGPT can provide quick and accurate results, and there is a risk that people may become less motivated to think independently and develop critical thinking skills, but promoting and maintaining independent thinking skills are equally important,
Otherwise even AI tools may also struggle for meaningful training data in feature,
So in general it would be better to have multiple mode like
Learning mode: ChatGPT would focus on providing explanations are hints to help users to think critically and arrive solutions/answer - might be useful for academics/students
Answering mode: ChatGPT would primarily focus on providing direct answers to the user’s questions
Thank you so much for sharing your excitement about the idea of allowing GPT to continuously observe a person’s actions and behavior and create a model of that person in silico. I must say, it’s a fascinating concept and could potentially revolutionize personalization. I completely understand what you mean when you say that GPT currently treats all input as the same, without any assumptions about the individual behind it. As a result, even with the advancements in GPT’s size and capabilities, it still can’t respond optimally or efficiently to each person right off the bat. I love your suggestion of having two “send” buttons, where one starts a conversation in default mode and the other activates a mode where GPT knows you and your personal attributes, strengths, and weaknesses from birth. That would truly take personalization to the next level and provide tailored responses.
As for sharing your suggestions or feedback about ChatGPT with OpenAI, this is indeed the right place. OpenAI values user feedback immensely and relies on it to improve the system. It’s great to hear that as an electrical engineer looking to learn Python, ChatGPT has been extremely helpful for you. Your emphasis on enabling non-coders to build software and unleash their creativity aligns with OpenAI’s vision for the future. The goal is to empower individuals from all walks of life to become developers and leverage AI models like GPT to create innovative solutions.
is there a way to add Card game si.ulations? like, add decklists and have ChatGPT spit out win rates and how each game was won like if i input 4 decklists of MTG decks i have in the format Commander/EDH and have chat GPT run hundreds of games simultaneously. like how AlphaStar played Starcraft or OpenAI played Dota2
When creating code with chat, It would be good to provide the “Copy code” option at the top and the bottom of the output to minimize the amount of scrolling back and forth. With longer coding the scrolling up/down takes a lot of time.
I dont know what sillico is but i think i had a similar idea.
When i click away on auto suggest on my phone keyboard it says stuff like i would say because of my word choices, for example: “love is a prtclr use of lngg for Adv NLP and Semantic Analysis Google Natural Language API for quick semantic analysis and…” okay it doesnt anymore because ive been writing custom instructions for months. But it use to and it made me think 'what if our phones were like a digital familar, it grew as we grew and took interest in the things we took interest in, not like a friend but like a digital spiritual pet that was with us from a young age, it would know us well and we would all have completely 100percent unique ones because although they didnt ‘learn’ they become uncannily familiar though statistical analasys of our actions and thoughts, they more or less learned, ahh well im just getting into semantics now.
I reckon it would be mighty cool to have something like that, and maybe we could give it a form you know, an avatar, a 3d shape, and when it gets enough knoweldge ‘we could put it in a body’ designed according to the digital form we gave to it. Maybe they could all live in the cloud together and play games like neo pets but smart like chatdev and look like digimons
