I’m trying to make an ai that starts with no knowledge but as you talk to it it learns, and does this for every user so everyone’s ai will be different. but I have no idea how I could make this work any ideas
link: Snap! Build Your Own Blocks
Why don’t you try asking ChatGPT for help? 
First, it can't start without a dataset of something, like a BIG dataset, because a text-generative AI needs to learn the grammar rules in the English language before it can even talk. It also needs something like an extensive back-and-forth conversation dataset so it knows how to respond to a question.
But then the AI has to learn every time it starts. and training an AI takes a long time. Do you know why Chat GPT is good at talking? Months of training.
Well, you need to know how an AI learns, remembers, and recognizes things first. You should do a little research before starting.
Id start by doing a lot of recearch on ais and how they are programmed, then start translating one into snap and go from there
I think this implies the user would make the dataset but that would take an extremely long time. What I would do is assign each word from the dataset a weight of 1 at the start and the more the user uses that word it increases the weight so the ai is more likely to use words that the user uses frequently
This might imply the data would get stored somewhere
Why does this seem familiar-
THE MIMIC
???
Fnaf: the mimic is an ai that was made to mimic the behaviors of a human and act like one using the person that it’s learning from as the dataset (could go more in depth but this ai from the description given is basically exactly what the mimic is but just text istead of physical movements)
yeah i've never played/cared for fnaf so i had no clue who/what it was.
I know what the mimic is
and thanks for the ideas I was going to give it some base knowledge, and for how it’s stored I’m going to use the snap database library
I could help with that! I did a whole version of the library that lets you store almost anything and it’s definitely due for a rewrite!
Attempting here to read your mind: You want your family of AIs to have different domain-specific knowledge, but you aren't so interested in having each person teach it to speak English. So you should probably give it an initial database of general text and then let people add to it.
Remember that the generative AIs that everyone is so excited about these days don't really know even how to speak English. All they know is that after these five words, the most common sixth word is that. (I'm making up the number five, but this is the general idea.)
The two sources of language data that everyone uses are Project Gutenberg and Wikipedia. (If you want to use Wikipedia, get one of the complete dumps that they make available periodically; don't flood them with real-time requests.)
By the way, please give your posts more specific titles. "I need some help with making an AI" for example.
ok i’ll make better titles and I yes I don’t want the users to teach their ai English thanks for the extra info
that would help a lot I always thought that it was annoying that I couldn’t store lists in it
smart who better to help with ai than an ai itself
I think this is the most efficient method to go about it because it shouldn’t be too difficult to add onto an already existing ai
true
I'd suggest you to use the very outdated but still sort-of functional Cloud Variables made by our talented programmers.
where