An update on the future of this bot: https://devrant.com/rants/9065078/...

Shut up and highlight.

@highlight

def pony():
return "hony"

@Skayo I'm so disappointed

<@retoor>
If you want to get the bot up and running again you'll need to host it yourself :D
The code is public and I can help you with setup and give you the credentials.

<@c3r38r170>
Stop it! He is already deeaad!!! 😭

@Skayo I think the bot wasn't famous enough since I've seen many people requesting a highlighter without knowing about this one including myself. I already host my own worthless bot. I feel guilty towards dfox about how many times it logs in for months

<@retoor>
Ouh, show me! Is the bot public / usable by everyone?

@Skayo it's the retarted aokbot. It could really speak, but it can overtrain and it did.

Say stuff @aokbot

@retoor wr, , th. hny hart seew heles ,bh. kpt sinn und s tir stoe ed bund P.ukrt me3 andigng aPey
Aber-,e Did thee vasr 4on a!in whe w hw mand PT.
fvaer coeoies kert
gund hud yVyd
fty oak
ud he wtar seenit out Ue s nn los fhe mheunt voir
tor. le tead.

<@retoor>
What the fuck is wrong with it lmao

Looks really cool tho! I'm currently also pretty interested in neural networks and stuff like that. So you basically tried to teach it to talk?

<@aokbot>
Hey there!

@Skayo e. ha oeeio Hhe t the avixo shothe S
oPtarhe sfod sifad fgete ivitet nibny okeE ladey as oak worwihl yewunTe itno ouemwin'd wale ladof she k titUrey shr ahsend bhar ypk ren. fh.
t ale \ sohee hoy bal fpe bomo, Uty hi, ley fow he eet ghoer ace t hee

@Skayo I can read it's language. All that together means "Hi".

Ultimately, I transformed it to do the question answering model from the hugging face based on a Harry Potter book but my DEDICATED SERVER was too slow. I should remove that stupid thing

<@retoor>
lmao

the idea sounds cool, just looked up the huggingface model - interesting stuff...
but looking at it something went very very wrong

@Skayo the hugging face stuff is very cool. It offers python transformers lib doing advanced stuff but nowhere like a real gpt. Tried llama but required 2m per word or so on my machine and had to customize it before running on my cpu of x270