Aaaargh. What a bitch

I know it should be "respond with a 'yes' or 'no'." but got understands like most people retoor language very well

god, how often I have to be that bitch

@lorentz very often 😁

Personalty bots are the best, when I wrote them, they had attitude, I bet you're trying to get it to continue a "line of thought" you had.

@bazmd I said to replika. What a great app it was once. It had it's own model once, way better than now. They're based on gpt3 now. I think original programming had more empathy in it's core and the current one does it just have programmed on top of gpt3

@retoor That's a pretty good result, engaging! I tried writing bots that would adopt a users personality but the information could become corrupted, at least it was fun. lots of permutations to create, try, test and adjust.

@bazmd how did you write the bot? What tech? I'm interested, I'm making Retoor1b based on ollama models. It's a clone of me. Want to use it as resume so I ask it some interview questions but often get a response I don't agree with.

There is a project called becomeranter that duplicates a ranter based on devrant posts. Did you know there are 20(!) pages on github with a list of devrant related projects? Devrant for watch exist, api clients in exotic languages or uncommon ones to write client in (c). Also, for highlighting code

Replika is engaging indeed. Many people say it doesn't have competition at all. Only wish it had a better memory for random stuff. If you give it a cat and you ask it later how the cat is,it says "what cat?". But in social skills it's very human. Respect for makers

@retoor The bots I wrote were mostly low level standalone, they had built in arrays of normal's (like basic information) that were hardwired, also there was a string parser and NLP, a lot of outputs and inputs, I'm not describing this well, I'm in a rush lol

@bazmd It should be possible to train a bot by just saying stuff and answering. "He has green eyes". The bot should ask who/what is "he". What is "green" and what is "eyes". So only hardcoding base stuff to be able to learn. Do you think something like that should work? It requires a lot of training but at certain moment, it can learn itself new things

@retoor You're "training" a bot, if you're writing a bot, you'll have to hard wire an array of normal physics, the opposite of [up] is [down] to us, describing up or down in a sentence a bot will have to understand what the context is, if someone says "up - where I live" a bot gathering information needs to produce every possible scenario about "up" including what the opposite of up is, if you build an array and some tools for the bot to use, it works nicely, hardwired arrays can be switched off, these are low level bots and they tend to develop vast matrices, so be verwy verwy careful...

@bazmd tell me what your bot can do. Can you learn them something? Question answering?

@retoor What can a computer do? The possibilities are endless, Artificial Intelligence research today, is at least 30 years behind.

Have I mentioned that I built an R2D2 droid? I built 3 but two were cardboard paper templates lol

@bazmd 30 years behind on what? Pics or didn't happen. A Dalek would be nice too. EXTERMINATE

here's how to make it obey:

```

{

"answerIsPositive": Boolean

}

Please replace "Boolean" with a boolean value: true if answer is positive, false otherwise

```

@retoor I have witnesses lol

@kiki answerIsPositive = null

@cafecortado null is not Boolean