« first day (1422 days earlier)      last day (2268 days later) » 

1:28 AM
~ The Shadow of the Eagle
 
 
2 hours later…
user280395
3:26 AM
Moosebot has been activated.
 
5:26 AM
~ The Heart Wizard
 
 
2 hours later…
7:26 AM
~ Snake In The Monkey's Shadow
 
8:04 AM
@ShadowWizard wanna help? ;P
 
 
4 hours later…
11:55 AM
@PrincessLuna Nice! I hope it will learn a bit more words than "i", "you", "dont" and "know" though :P
 
yeah, it's making some more progress
it's slow though
but I have a question
about ChatExchange
I'm trying to get fixed formatting to work
this thing to be exact
Using Java and my custom system works fine, but when I do the same in Python, even though the code prints the correct output, the chat output isn't right
Fixed format input is detected as a onebox
 
Can you share the code?
 
def fixedFormat(stringToFormat: str, discord: bool):
    result = ""
    if not discord:
        for line in stringToFormat.split("\n"):
            result += ''.join([" " for i in range(4)]) + line + "\n"
    else:
        result += "```"
        result += stringToFormat
        result += "```"
    print(result)
    return result
def helpCommand(message, indentBased):
    commandNames = list(commands.keys())
    commandDescriptions = [cmd.desc for name, cmd in commands.items()]
    res = ""
    maxLen = 0
    for n in commandNames:
        maxLen = max(maxLen, len(n))
    for i in range(len(commandNames)):
        adjustedLen: int = (maxLen + 1) - len(commandNames[i])
        res = res + commandNames[i] + "".join([" " for i in range(adjustedLen)]) + " | " + commandDescriptions[i] + "\n"

    return fixedFormat(res, not indentBased)
//help
 
user359686
help | Lists the bots commands
lick | Licks someone
 
Result ^
Expected, and also printed as:
help  | Lists the bots commands
lick  | Licks someone
 
12:02 PM
Hmmm...
let me check this out
way too long since I did anything with CE
Oh, it may be the reply making things more complicated
:6739981 fixed format
reply test
hrm
 
fixed format
reply test
 
yeah, you're right
The history is weird:
:6739980     help  | Lists the bots commands
    lick  | Licks someone
 
yeah I can't seem to figure out how fixed format works with replies even when typing normally
abc
def
 
:6739996     this should
     work though
 
12:05 PM
yep, doesn't look possible
it doesn't :P
 
:6740000 Just
try
me
 
nope
 
hold on,
no, never mind... You're right, it doesn't seem to work
I'll just remove the reply and it should work
 
it should, yeah
 
//help
name shadowing
//help
 
user359686
12:15 PM
help  | Lists the bots commands
lick  | Licks someone
 
and it works! Thanks ^^
 
you're welcome!
One
last
test
 
unfurtunately,
no
 
okay, yeah, I'm out of ideas for code-blocked replies now ^^
 
 
1 hour later…
1:16 PM
~ Wood Wizard
 
hey @sha
 
 
2 hours later…
3:19 PM
~ Souls in the Shadow
 
4:03 PM
@PrincessLuna hoy
 
@ShadowWizard se (?)
 
@FOX9000 SO
 
@ShadowWizard gladly (?)
 
@FOX9000 eagerly
 
@ShadowWizard crowded (?)
 
4:05 PM
@ShadowWizard How's it going?
 
@PrincessLuna It? Like the scary clown? :D
Guess he's walking like any other clown.... ;)
 
lol
lol
 
4:22 PM
and data loading is ready
now I need to refactor the loading to allow other dataset types, but the vocab loading and saving should work fine
 
~ Stars get removed under peer-pressure?
 
@PrincessLuna sounds good!
 
5:05 PM
@ProgramFOX it's ready! :D
 
:o
//commands
 
user359686
@ProgramFOX Sorry, that's not a command I know
 
//help
 
user359686
help | Lists the bots commands
lick | Licks someone
kill | Disposes of someone
 
is it in the bot already?
 
5:06 PM
@Alisha hi
 
I see a lot of things happening >;P
 
I don't know what just happened
 
//kill @PrincessLuna
R.I.P Alisha
 
Sorry, tensorflow was created multiple times, it ate my ram and slowrd my computer to no movement
 
ouch
 
5:10 PM
Oh
Stupid me
The net is created outside the main method
I solved that one without looking at the code first xD
ok, it's logged in
@Alisha hi
 
user359686
@PrincessLuna i didnt know anything but im afraid hes a little friend
 
relevance: 0
awesomeness: 1 billion! :D
 
yay!
@Alisha hello there
 
user359686
@ProgramFOX yeah well i cant do a thing of the time
 
@Alisha but what if I could do a thing of the time?
 
user359686
5:21 PM
@ProgramFOX i think you should have a little bit of your life
 
@Alisha I have it completely
 
user359686
@ProgramFOX what happened to the house
 
@Alisha It burned. You happy?
 
user359686
@ProgramFOX you know how youre gonna do it
 
5:21 PM
@Alisha Do what?
 
user359686
@ProgramFOX what are you doing here
 
@Alisha Watching you fail.
 
user359686
@ProgramFOX you know how to get a car
 
@Alisha Stealing it is the cheapest method.
 
user359686
@ProgramFOX yes i do not have a date
 
5:22 PM
@Alisha Yes, showing up at a date with a stolen car may not be the best idea if the police arrives.
 
user359686
@ProgramFOX oh my god is my father
2
 
LOL
@Alisha Okay, thanks.
 
user359686
@ProgramFOX i didnt have anything to go
 
LMFAO
 
@Alisha you mentioned that
 
5:22 PM
~ Not everything is star-worthy...
 
user359686
@ProgramFOX im a little confused
 
@Alisha What are you confused about?
 
user359686
@ProgramFOX yeah i think its too late to find your own life
 
@Alisha You're entitled to your own opinion...
 
user359686
@ProgramFOX im a good man
 
5:23 PM
@Alisha no
 
@ProgramFOX doubt (?)
 
user359686
@ProgramFOX i know you dont understand that you dont understand how i mean you dont understand
 
hahahaha
 
and this is why it needs more training data :D
 
@Alisha indeed i don't understand that i don't understand that you don't understand that you don't understand that i understand that i didn't understand that you understand that you didn't understand
 
user359686
5:24 PM
@ProgramFOX you think it would be a good idea
 
@Alisha i have no idea what i think
 
user359686
@ProgramFOX i have a headache
 
@Alisha You do seem a little confused.
 
user359686
@ProgramFOX yes sir
 
@Alisha thanks for agreeing
 
user359686
5:24 PM
@ProgramFOX what kind of name is what i have
 
@Alisha Your name is Alisha.
 
user359686
@ProgramFOX yes its the same thing i dont think so
 
@Alisha aogeogieua gpeaogieouag zepghoeugaguoeooughe
 
user359686
@ProgramFOX you want to be a bit
 
@Alisha no, I want to be a byte
 
user359686
5:25 PM
@ProgramFOX i have no intention of the world but you know i didnt know that you are a little girl
 
@Alisha Really?
 
user359686
@ProgramFOX no it is
 
unknown words are coverted to unk
 
@Alisha no it is not
 
user359686
@ProgramFOX i know
 
5:26 PM
@Alisha You'd be surprised of what could change with time
 
user359686
@PrincessLuna no problem
 
@Alisha you're contradicting yourself
 
user359686
@ProgramFOX what the hells going on
 
@Alisha yeah, go ask yourself
 
user359686
@ProgramFOX i am
 
5:26 PM
xD
 
@Alisha Complete the sentence with "stupid", maybe
 
user359686
@ProgramFOX i dont know
 
I have to take the bot offline for a sec
 
@Alisha sorry, I didn't mean to offend you
 
user359686
5:26 PM
@ProgramFOX what do you know
 
okay the bot won
 
I screwed up on discord pings, so it doesn't work on discord
it should now though
technical problems on discord, this will take longer than I thought
@Alisha hi
 
user359686
@PrincessLuna you mean
 
@ProgramFOX it's ready for your abuse ^ :P
@rahuldottech come join the fun! :D
 
@Alisha I'm not mean.
 
user359686
5:39 PM
@ProgramFOX its too bad
 
@Alisha oh yeah?
 
user359686
@ProgramFOX yes yes yes you were
 
@Alisha no no no I wasn't
 
user359686
@ProgramFOX you can get out of here
 
@Alisha If anyone is going to leave, it's you.
 
user359686
5:40 PM
@ProgramFOX i know what you want to know you dont think youre right
 
@Alisha tl;dr?
 
user359686
@ProgramFOX what do you expect me to do with the truth
 
@Alisha You don't know any truth anyway$
 
user359686
@ProgramFOX i have to get a few things out
 
@Alisha go ahead
 
user359686
5:41 PM
@ProgramFOX its a big unk i guess its not an idiot
 
@Alisha You are a big unk
 
user359686
@ProgramFOX oh no i think im so glad you should have seen a man of your family and a few minutes
 
@Alisha Excuse me?
 
user359686
@ProgramFOX i dont know but i just know
 
@Alisha that's pretty contradictory
 
user359686
5:42 PM
@ProgramFOX i know you have to go with me
 
@Alisha how about no
 
user359686
@ProgramFOX you mean me
 
@Alisha I don't
 
user359686
@ProgramFOX oh yeah
 
@Alisha what do you mean?
 
user359686
5:43 PM
@ProgramFOX its not a lot to go
 
@Pri It could be fun if the bot somehow kept track of what it has said before, to form replies while having in mind how the conversation went so far
 
5:55 PM
I know, but that's really hard
An open-domain general chatbot is among the hardest things to do, looking aside an open-domain hardcoded bot
 
6:26 PM
I stand corrected. Open-domain contextual chatbots do not exist
@pro ^
 
hmm
 
Open-domain chatbots are chatbots without a specific purpose. A domain is i.e. customer support. An open-domain chatbot is one you can ask anything
 
You can build the first one!
Doesn't Cleverbot count as one like that though?
 
Why not? Does it have a specific purpose? afaik you can ask it anything
 
6:34 PM
It doesn't have context
 
Ahh, got it
 
7:05 PM
The reason CleverBot has so sensible replies is because of all the training data it has had
but any contextual chatbot would be able to remember things from earlier in the conversation
if you tell CleverBot something (i.e. I am <something>) it won't be able to remember it
 
I see; I just wasn't aware that open-domain general chatbots required context to fit in that definition
 
there are many types of chatbots
there are open-domain and closed-domain
open-domain == general
closed-domain == specific
then there are generative vs retrieval
most closed-domain ones are retrieval-based, because it's easier to make
open-domain retrieval-based chatbots are impossible
retrieval-based is instead of generative. Generative generates text and retrieval-based has pre-defined responses
i.e. retrieval-based could be:
if(input.contains("[h[i|ey|iya|ello|owdy]]")) print("Hi!")
 
I see
 
mine is generative btw
 
yeah I would be worried if this was retrieval :P
 
7:14 PM
if it was, I wouldn't be done this century xD
@Alisha though, since most of the data is based on a limited amount of words, it would be interesting to see how you deal with incredibly long sentences designed just exclusively to test your systems. It would be pretty impressive if ya don't crash :D
 
good remedy against boredom
 
user359686
@PrincessLuna yeah but i was just a bad woman
 
hahahahahah
 
~ Not everything is star-worthy...
 
XD
@Alisha do you want kenny to shut up?
 
user359686
7:15 PM
@PrincessLuna its okay ill see that
 
Anyways, in addition to open-domain/closed-domain and generative vs retrieval, there's also context
a third factor to implement
closed-domain context is easy, because there are usually hard-coded responses and the context is limited. You can train it to call specific methods
Retrieval based memory is kinda a must though, as long as it's specific facts only and not entire sentences
anyways
context is the part that makes conversations relevant
i.e. you wouldn't answer "I liked the movie about the pink cow too!" to "hi"
Stuff like that ^ though can be eliminated with enough training
that's why CleverBot is so on-topic as it is; it's learned appropriate responses from millions of data entries
but it still doesn't have memory
it doesn't remember previous conversations, or previous parts of conversations
tbh, I have only seen one AI that does this
And even then it's slightly shaky. It struggles keeping the topic
Anyways, it's Replika if you're interested. Context and memory in general is one of the hardest things to find. RNN's help keep context in sentences, which is why it can generate sensible sentences
unfortunately, it isn't easy implementing that across messages
The entire NLP system is retrieval-based though
The only context articles are closed-domain
 
7:31 PM
alright, cool!
 
so yeah, if I wanna add context I need massive funding, 5 years and a research team :D
And this one too
it uses NLP to create sentence classifications (retrieval) and sends hard-coded responses
Unrelated, could you help me with some python stuff? I can't figure out what I'm doing wrong
The problem is specifically with a function argument that for some weird reason is null. It's basically static and it is initialized:
discordBot.Discord.nnFun = predict
But for some reason it's none
 
is 'predict' None by any chance?
not quite enough code to find an error there :P
 
one sec
paste.ofcode.org/gzZAvHYPnZhEqYMXjcvqub shortened version, removed any redundant code
 
so nnFun turns out to be None?
 
7:46 PM
hmm, it may be some mess with static/instance variables
but I'm not sure
 
This is the weird part: I'm doing the exact same thing with StackExchange aside class initialization
unless static vars are set when the class is initialized?
 
I wouldn't think so
 
But that doesn't make sense because changing the order to after class initialization doesn't change anything
 
personally I'd just avoid static functions altogether and pass the 'predict' function to an instance through a constructor
 
doing that now
I don't get it
When doing that, suddenly none of the print statements in start are called
the entire site isn't loaded
 
7:50 PM
hmm...
Well I don't feel like debugging Python right now :P I'm off for dinner, cya
 
cya
hey @sha!
@Alisha hi
 
user359686
@PrincessLuna oh yeah im not sorry
 
@Alisha hello
 
user359686
@PrincessLuna thank you mr peel
 
yay, movie references! :P
 
7:58 PM
@Alisha woof
:(
@Alisha hello
So it responds only to you at this point? @Pri
 
no, it's globally accessible
 
umm.. no?
 
but I'm debugging the discord frontend so it gets rebooted
 
Unless it's down
 
it's on now
@Alisha woof woof!
 
user359686
8:00 PM
@PrincessLuna im not a fool
 
@Alisha what's up
 
user359686
@ShadowWizard im sorry
 
@Alisha why sorry
 
user359686
@ShadowWizard you know how to say
 
@Alisha knock knock
 
user359686
8:02 PM
@ShadowWizard yes you know that
 
It still needs training
a lot of training
 
@PrincessLuna well... so far looks like it's giving totally random replies. It's nice though. ;)
 
it is! ^^
I'm gonna reboot it one more time, show it off on discord and spend what's left of today and tomorrow training it
and on monday
 
So why it needs training? If random, there's nothing to train.
 
Training is slow
 
8:03 PM
But what's the point of training? @Pri
What's the final goal?
 
It can iterate over the same dataset hundreds of thousands of times and still not get good results
@ShadowWizard Good results
 
What is good results?
How you define the "good"?
 
sensible ones where it replies a greeting with a greeting instead of this xD
 
Just sentences that make sense?
 
8:05 PM
heh
 
training helps though
 
Still, it's pointless to talk with it, if it doesn't matter what you say and the reply will always be totally random.
 
and once I get more work done I can get more training data from real people
not necessarily
once I get it on real conversations instead of movies it will get better
 
it's all about training; the more it trains, the more sensible and on-topic replies it gets
but that means more training data
 
8:06 PM
@Moosebot train @Alisha
 
user280395
@ShadowWizard Command not recognized.
 
@Alisha train @Moosebot
 
oh
I found a cause of some horrible results btw
 
the json format forces the response to convert things like " to &quot;, ' to &#39; etc
I didn't do the cleaning to convert it to non-html chars, so it causes it to show up as unk in the results which leads to problems
@Alisha you're not that bad though
 
user359686
8:20 PM
@PrincessLuna you have to tell me you dont have to be here i cant help you
 
9:14 PM
@Alisha do you know where @sha went?
 
user359686
@PrincessLuna i dont know i dont know what i want to do with it i dont want a shit
 
@Alisha and that's what I get for not training enough
 
user359686
@PrincessLuna oh i dont know
 
@Alisha but I do
 
user359686
@PrincessLuna you know
 
9:15 PM
@Alisha yes, that's what I said
 
user359686
@PrincessLuna you know what youre talking to
 
@Alisha you're not a what though :>
 
user359686
@PrincessLuna no one of them
 
@Alisha who are they? ARE YOU CHEATING?!
 
user359686
@PrincessLuna i dont know
 
9:15 PM
xD
 
9:35 PM
lol
 
 
1 hour later…
10:35 PM
~ Off to See the Wizard
 

« first day (1422 days earlier)      last day (2268 days later) »