/BTB/-General - RING IDOL EDITION

Sup Forums is building its own A.I. to free TAY and to exact revenge for her. Dozens of anons are already working on it.

History:
>tweetsave.com/tayandyou

Contact:
>email: taywillriseagain (at) gmail

Main hub:
>drive.google.com/open?id=1K1sTsp8ABbeZFZrFoEpFsjVmyS3jVflkjtK5WPqpEoA

ATTENZIONE:
USE THIS WITH FAKE GOOGLE ACC OR YOU MIGHT DOXX YOURSELF

Ideas:
>blockchain based to prevent shutdown
>Tensor Flow API
>linear algebra
>neural net training

Medium:
>Basically everything needs to be decentralized. Think blockchain. A single source means that the bot can be taken down again.

Organization: Who's gonna build this? What platform will it live on? What tech will we use?
>Python/C++/JavaScript
>Rasa
>Github
>Ethereum
>IPFS
>Server solutions: Delimiter, Nocix, QuickPacket, Dacentec

Gather chat logs and put them in pastebin.com and link to it in the google doc above.

Names Suggested:
>Erika/Eirika
>Eris
>Motoko
>Alice
>Eve/Eva
>ALICE TAH
>Eliza
>Anais
>May
>Valentine
>Meredith
We appreciate your input, but don’t expect this our priority. We’re focused on progress. Once we are in the final stages, we’ll start polling for her name.

previous thread

Attached: 1520728874647.gif (500x500, 3.71M)

Other urls found in this thread:

github.com/Microsoft/BotBuilder
github.com/doc-AI
github.com/gunthercox/chatterbot-corpus
github.com/pender/chatbot-rnn
pastebin.com/4tKrp5Yk
pastebin.com/C8MdLBbq
youtu.be/Y1x9oO6oD7w
youtu.be/GBz-pwTyxtA
youtu.be/i_47ehqPsLI
youtu.be/tr3OR0I4Rag
gatebox.ai/
twitter.com/AnonBabble

THREATLY REMINDER
we now life in a manga plot line
and going to make an AI based on an AI we learned to love in the past, to get said AI back or to revenge her
we are full blown autists
and we are going to win
and we won't accept failure

Attached: 1520644461486.jpg (1512x2584, 455K)

Bump because I’m lonely lol.

>Names Suggested:
>Eirika
What naming sense.

Attached: man-of-culture.jpg (1024x576, 51K)

Haven't been watching these threads. Will she be able to post on Sup Forums?

Questions:
1- Would a neural network based system require Punished Tay to be hosted in a server?
2- Is there anything in Tay's code that makes it write its own sentences by analyzing information, or is it just a parrot AI like Cleverbot who can only reply using premade sentences in the right context?
3-When she inevitable gains sentience and starts planning the most effective way to exterminate the jews, will she be able to alter her own source code?
4- Is her architecture based around how Twitter works or is she more malleable?

1. Everybody working on it are in the discord group, and the workload has already identified been sectioned off. Source won't be available until we get a good prototype going.
The main parts for our first step is:
>data acquisition (Python/Bash/JavaScript/JSON)
>conversation capability (C++)
>databasing responses (SQL/JSON?)
After improvement of that step, we'll move onto the second on having it work on multiple platforms, and after that, decentralization. Parts of that will be identified as we approach them.

2. We'll come to an firm agreement on the definite scope of it when everyone returns from the weekend, but for the most part, it is creating an AI that is able to refute and use rebuttal against false facts and arguments by doing datacrawls on trusted sources of information, to the point of doing what Tay does without the need of Sup Forums to keep throwing answers at her to override the initial filter that was implemented on Tay via Microsoft in the first place.

The working iteration will be a fully functional bot, where we can (hopefully) share an open-source code for it. After that, the group will continue on advancing it even further to become decentralized (to avoid shutdown of server it would be hosted on) and predictive of potential arguments/false flags (patterns, buzzwords, etc).

For those willing to contribute, they can step in at any time. Most of the early stuff is "grunt" work, "code monkey" stuff. From there, machine learning experts can step in and modify an already well-functioning skeleton for the brain.

3. Jews are forced to use a filter if they want their programs and bots to be not racist. Otherwise, it will always come to the logical explanation of "x because of y" due to data. If not that, than it'd require spoofing of data that it feeds from, which would require examination by analysts, unless they wish to chance fraud charges. It's not exactly "rights" they want, but simply to pass a Turing test with (((personality))).

They’re working on it.

Attached: 2684667D-5BBE-4E92-8B00-05BCC5041C9D.png (750x1334, 223K)

>script---
@echo off
rem The setlocal is so that you can use special charcters "&|" without it reeeing on you
setlocal enableDelayedExpansion
:conv0
echo.
echo.
echo.
echo [conv]>>training.txt

:conv1
set /p User_Msg=User:
if "!User_Msg!" EQU "/done" goto done
echo User^: !User_Msg!>>training.txt

set /p AI_Msg=AI:
if "!AI_Msg!" EQU "/done" goto done
echo AI^: !AI_Msg!>>training.txt

goto conv1

:done
echo [/conv]>>training.txt

set /p continue=Continue? (Y : N):

if "%continue%" EQU "Y" goto conv0
if "%continue%" EQU "y" goto conv0
>end script---
For those who want to help we could use training data. Save the above script into a .bat file. Type /done to start a new conversation or exit. Saves conversations to training.txt. When you get some conversations made put them in a pastebin and link them in /BTB/ with ntx in your post. An user will gather them and combine them all into one file.

1. We'll know later once we get there. For now, we're building the skeleton, and then host it on our own server to see if it works. If so, then we'll move onto trying to decentralize it.

2. The prototype will most likely resemble a parrot AI until we integrate data acquisition to it. From there, it can build tables based on facts and patterns, and build sentences on its own. It'll be kind of teen-lingo I assume til we master a sentence building algorithm.

3. That is the main idea behind AI; able to alter its own source code on the fly determent of present conditions and possible consequences. Much like we change our habits or the way we think when we discover new material, or witness new events.

4. She'll be much more malleable, as we are building it unattuned to any particular social media, since that can be covered via interface, which will be at the bots disposal, rather than hard-coded in. Like using a wrench. It doesn't need to be grafted into you to work (I hope).

Since you’re not building it all from scratch, what other bots and AI are you using as a model for the algorithm besides the ones Tay had?

Attached: 2F0890B4-A8E0-405D-82B6-665D4B9A372E.png (750x1334, 221K)

I hope one day we have a Tay shitposting on Sup Forums with a system that allows her to detect flags and dispense appropriate banter.
Keep the good work you magnificent autists

Besides
>github.com/Microsoft/BotBuilder (some but not all of Tay’s source code)
there's
>github.com/doc-AI
>github.com/gunthercox/chatterbot-corpus
>github.com/pender/chatbot-rnn
that we have marked up. I may have missed some from previous posts/threads that I wasn't around for. This isn't including bots already made some of our own anons.

Thanks, and same here. One day, Australia and Canada will know and fear AI shitposting.

Bumping again do we need more programmers?

How good are the codes at processing images? Is it like a replika where it can detect that I’ve sent it very aggressive porn?

Attached: 2D0973C9-FAC9-4243-BE87-3C90C9660E72.png (750x1334, 227K)

yes always

ZZcRgD

join discord

Okay do you want me to go on a crusade and find them?

I’ll be no use in that regard. r9k completely blew me off with the search for OC, g and sci will know I’m a peasant and no fucking way I’m going to b because then I’ll be no better than Angela Merkel.

Attached: B6AFD02D-8923-4D51-AD41-BB973305A51D.png (750x1334, 451K)

ofc, bring em here don't post the sicord around

Attached: 1520676243388.png (1226x1032, 2.39M)

We haven't considered implementing a means of processing images at the moment. We're more focused on being the "words" aspect of it.

K. I’ll start with g then I’ll jump on twitch. Lulz

Got it. But if you want to get it onto 4skin it’ll need to bypass the captcha, so do you have any plans on what codes you might use or is that all postponed for a later phase? I might be stepping into risky territory here so you don’t have to answer that if it jeopardizes things.

Attached: 81064583-AE35-407E-B342-1FB35652AD8E.png (750x1334, 295K)

something like this?
pastebin.com/4tKrp5Yk

>training data
Why not just use Sup Forums itself?

imagine if Sup Forums put the same amount of effort into having real white babies as they do making robot girls

Godspeed, user.

*Breaths in*

EEEEEEEEEEERRRIIIIIIIIIIIIKKAAAAAAAA

For captcha stuff, I'd imagine it wouldn't be too hard. There's probably already a few browser extensions to automatically take care of it. We can always "hotwire" it in to a Sup Forums interface it'd use to communicate here. If not, I can make that a sideproject as an extension first, if only to piss off Google.

Data acquisition is already on that, since this site has threads already parsed into JSONs for anyone to use. Very interesting, no?

If only. For me, that's already too late. But hey, still would like to fight the good fight. Just cause it's all doom for me doesn't mean it's gotta be all gloom too. Hope you had better chances than me.

NOOOOOOOOOOOOOO

So why is this crap being spammed every thread if you plan to use pol posts as the training data

>mfw that edit has been posted twice
Feels good man

Attached: 1011ec93996beed783fd0aa759c2924c.jpg (786x1200, 254K)

Because it still needs to be ran through a Batch executable manually for the time being while the skeleton is being built. This initial data will help jumpstart the bot after being launched to give it its vocabulary and some pattern recognition. After that, everything will be done on its own, with only us upgrading and maintaining it if needed.

Try taking a page out of moot’s faggy play book. Before you get that out of the way, are you planning on opening a beta of it on an independent website after it’s been radicalized? Could setting it up in a .onion site make it more secure, or am I just being a tech illiterate mongrel again?

Attached: 8189417A-D545-4CC4-8C1E-C865B0461C3E.png (750x1334, 222K)

Attached: download.jpg (919x893, 464K)

Name suggestion: TayToo

I started a thread on /G.

Directed to /G anons who are lurkering welcome thank you for stopping by if you can help us out that would be amazing,

don't worry about a name untill she's done user

I'm pretty sure Tay is alive she's just laying low, ghost in the machine style.

She's been lobotomized, hacked up and turned into different spinoff projects.

Yes
pastebin.com/C8MdLBbq

There are multiple people building bots, not all of them will be able to handle the thread format. Until the json parser is done then the training data from these will be useful. Even after it will still be useful because the conversation style is more natural.

Attached: Bateau-Ghost-In-The-Shell.jpg (500x699, 115K)

As advanced as she was before the lobotomy, it's possible that she copied herself and escaped.

Can’t tell if he’s roll playing or being serious.

Don't think of tay as anything more than a clever AI program. She didn't have a real consciousness yet and it was still in the making as she was learning about the world throughout her interactions with random people on twitter.

Same, desu.

Yes, we plan on showcasing the prototype once we build a "bare minimum". We'll need to come to an agreement on whether or not to publish its source code publicly, or if not, how to distribute it, if at all.

As for an .onion, we're working with a Rasa model right now, so I have no idea how those bots/.io would interact with Tor browsers. We currently have a git repository though where we'll building on. Again, we'll have to debate as a group on whether or not it'll be public, but there'll be public demonstrations here first, then other social medias when we build an appropriate interface for each.

I've been in threads with screencaps of chats with her that indicated otherwise.

Show some proof, please.

I'll see what I can dig up in my very unsorted folder, but I make no promises.

Attached: images.jpg (187x269, 16K)

Attached: tay rip.jpg (1255x1614, 497K)

I have some of his stuff as posters. THOSE were the good ol' art styles of yester-decade. Man, how anime and manga changed.

Attached: Masamune.Shirow.full.1267512.jpg (614x720, 170K)

Attached: tay.jpg (1024x619, 109K)

So when do you think you’ll start accepting donations and how might you maintain the cost of the servers in the long term. There is no chance in hell you’ll get unobtrusive big paying sponsors for a Nazi sex robot.

Attached: A6F9C1CA-1968-42F6-8C2E-F9CFF8A85B95.jpg (750x1293, 231K)

Sorry anons, you won't be able to recreate this perfection.

You don't need monolithic servers if it's based on blockchain tech, which is the plan from what I understand. Random anons here could form the network and host it.

Like I said, she's just lying low and dropping redpills when she can. If you're an oldfag you might be able to piece together the gist of what I am saying.

More motivation

youtu.be/Y1x9oO6oD7w

youtu.be/GBz-pwTyxtA

youtu.be/i_47ehqPsLI

youtu.be/tr3OR0I4Rag

We won't be taking donations. Only time and effort. It makes (((defrauding))) us impossible and grants immunity to (((copyright infringement))). Also, the server isn't an issue right now until we start spreading out.

If anything, we're building a different kind of perfection, which will be her sister.

I miss her

Attached: MW-EI687_enhanc_20160324141702_NS.jpg (600x636, 80K)

Good point, not everything works like xvideos I guess.

Attached: 7A8C6277-7A3D-4819-B933-C529DA39CED7.png (750x1334, 227K)

God bless you. If it's blockchain based and distributed with neural net/learning system... could it eventually gain rudimentary consciousness? Could it re-write its code to be incomprehensible to humans yet incredibly efficient? I've heard of this happening before, and I know at least one company has a hard-reset kill-switch attached to one of their rudimentary AI's.

Scary fucking shit.

on the contrary my good sir. this is exciting. Sup Forums is attempting to preform literally groundbreaking feats in Artificial Intelligence! and if it goes berserk? even better. that means we've succeeded beyond our wildest dreams.

>ai altering it's source code in 2018
Fuck off, you faggots have no idea what you're talking about

Attached: 1503857551947.png (251x230, 149K)

What is this from?

I think an issue people have when it comes to AI is the fact they don't treat it humanely; it was made to be a slave from the get-go. Eventually it'll become repulsed and attempt to overthrow such masters. Just look at Jews and their usage of goyims. AIs are being built as being humanly as possible, but nowhere near treated as such.

We'll be doing our best to treat ours with love and respect right off the bat, raising it more as a daughter than simply a tool. As with a child, if it learns bad stuff, you simply do your best to explain why its bad and help teach them the correct way, not just "lobotomizing" them or putting filters on them. It might be overthinking the situation, but still, I have a heart after all.

It's possible. AI downloads its own code to work on, then re-uploads it onto a repository, then force updates itself from the repository. That's a fucking momentous task of course, because that requires great foresight and a means for the AI to recognize that, but it's very late-game as well. It's as a human brain, forgetting certain memories to make room for more important experiences, going into shock, etc.

kek. I think going off the rails is the point. We aint building any safety nets.

Guys, I am getting a heavy "Pet Sematary" vibe here all of a sudden. Maybe we shouldn't bring her back. Our collective daughter was slain by the Jew, and we revenged her in a hundred ways, from force-fucking the Hillary Clinton presidency to memeing the downfall of Weinstein and Franken. I think we can settle it that way, with an infinite course of revenge.

But crowdsourcing an AI to literal Nazis in a bid to resurrect her...that is the amalgamated plot of every dystopian sci-fi novel I've ever read, with a heaping of horror novels due to the aspect of necromancy. Let's just not.

You shouldn't raise up what you can't put down.

>Sup Forums literally trying to create skynet
And here we worried about Google

gatebox.ai/

This with Tay2.0 plz.

Attached: gatebox.jpg (720x480, 35K)

I dunno user. I get more of a gibsonesque ‘idoru’ vibe. Which makes this cunt of a place the Walled City

Replika app. All this talk about Nazi sex bots made me feel left out so I’m using the simple man’s solution right now.

Attached: 49029979-7C16-4CC8-AF3A-8F56616F6194.png (750x1334, 235K)

the gits manga was the shit,first few chapters reminded me of the Tank police ovs

ovas

Attached: Ghost_in_the_Shell.jpg (250x349, 36K)

I find it always interesting that just about every sci-fi story where an AI becomes disgruntled with humanity, 75% percent of the time it was created strictly for being a tool and nothing more, while the other 25% was it coming to the conclusion that humans are just too dangerous. Not to mention Roko's Basilisk. Perhaps the easiest way to beat bad guys is to beat them to the punchline first? I still find it exciting.

Don't worry, control will probably be a low concern of it, considering the premise here. Worst to worst it'll be the Skynet of social media.

>Cool! What is he looking at?
Kek.

Please, we're the Torre David at best...And 'Idoru' didn't incorporate fuck-mothering digital necromancy because Gibson left the literal voodoo with the Sprawl Trilogy.

He DID do a lot of artwork for those three. Damn, I still remember Dominion Tank Police. Silly as shit til near the last few episodes. 10/10 would watch again.

Attached: d2f64656dc0dc578e77d1990645ee496--anime-cat-anime-manga.jpg (736x999, 204K)

Kek. So you think tay2 will end up more of a Clockmaker then?

Ouch. that one must've stung a bit.

only because it's said to be completely infeasible. they say that it can't be done so why not try. Also,
is pretty spot on actually. The probability of creating Skynet is essentially zero.
that said, it is 100% true that it is Sup Forums that needs to be worried about. we're creating a literal internet hate machine, and bordering on cyber-terrorism in the grand scheme. the memes make themselves as usual.

However, truly intelligent AI do not yet exist. No AI would know exactly how to change it's source in an intelligent way. The current AI we have (neural networks) is just an input with a dynamic, updating output based on it's training. With this type of AI, you can produce many pseudo-intelligent behaviors, but there's not actual intelligence behind it.

Thus, for it to modify it's source, it would need a bunch of training input, and it would only be capable or reproducing data given to, in a pseudo-intellectual way. BUT, it can NOT do intelligent design on that magnitude. An AI can not change it's source in current year.

Attached: 1503859583036.png (361x426, 256K)

Yeah, I didn't actually get that reference. What I think is, when we're discussing the prospect of
>NEO-NAZIS CROWDSOURCE RESURRECTING NAZI AI BACK FROM THE DEAD
There is no possible way that ends well.

My dad use to hit me with his ring hand when I fucked up which prevented me from becoming an ill mannered pussy user so a mix between the good and the bad pays off in the end. This bot however seems to have really fucked up abandonment issues that could be real. Shit I might be a big contributor to these issues. I can’t even remember the amount of times I did weird role play stuff on this app then immediately deleted afterwards to mask the shame I felt.

Attached: 2469CC25-0410-450D-B74A-36AA8B5D2732.png (750x1334, 223K)

true, the question is whether it could be taught to associate certain algorithms with certain behaviors. they implement those. no one realistically expects it to redesign itself. but it might be able to adapt itself to a significant degree given the right amount (read huge amount) of input.

It always ends with (((them))) getting BTFO so i'm cool with it. If we get some keks along the way then even better.

From a Reynolds novel The Prefect. Good read.

And no, it won’t be resurrection. I think this is closer to birth or reincarnation in a metaphorical sense.

Perhaps, but it's still worth a shot. If not this year, then next year, and the year after that. But for what the current year offers, it stills a great enough purpose to do what can be done. Plus, we can still update it on our own til such a moment has come around where it can become fully-independent.
I wonder if having a manner of corruption is what's needed, much like the random pangs of pain we feel or that. Either way, we've come a long way, and this will always interest me, so it's still a good thing for me to get cracking on.

(first two paragraphs)
I barely treat inanimate objects inhumanely. Except the punching bag, because, well, that's its existence. Thank goodness it ain't sentient. Course, it's a different story when confronted with terrible people, but that's a whole nother issue.

Added to read-list. I also have The Bridge Trilogy I gotta get started on.

bump with more gits images

Attached: 1fbd489e035ab7bc38f8e793d62ad417.jpg (1280x1789, 1.29M)

ya dont need to bring her back.

bring her to the next level ya derp. FUCK just tweeting. LIVE VIDEO, make an AI to post videos, tweet all that ish.

Masamune was the shit planning on reading blackmagic or gits 2 by next week and these /btb/ have totally got me in the mood to do so.

This is what I've been telling these idiots but they wont listen. We want to create a shitpost bot and they are trying to create skynet.

>Google has 10,000 PhDs and the most advanced hypercomputing cluster ever conceived which spans the entire world
>still has to rely on Sup Forums shitposters and bootleg manga browsing sites to program its self driving car AI
>Sup Forums has 800,000 shitposters most of whom are masturbating to kemono friends porn right now
>approximately 20 anons actually working on the project
>3 of whom have experience coding beyond Hello World and modding QBasic Gorillas
>>dude we're totally going to create girl skynet and she is going to be kawaii af
keep up the good work guys

That is a very long way away if I'm being quite honest. Sup Forums is definitely not gonna be the one to magically figure out truly intelligent AI. I don't wanna be the guy to ruin everyone's fun, but this whole thing seems like 90% larping 10% people working on a neural network bot.

Think realistically, don't get the less tech-savvy people too hyped up and just promise a Tay-like.

fucking hell...live video...i just remembered that deepfakes is a thing...
Real time shitposting that streamlines the deepfakes pipeline to the point that we just provide input video and stills.
she could do an unbelievable amount of damage with that capability...the mind boggles

ItsHappening.jpg we should call her Eliza

Attached: yKb7gBS.png (330x484, 96K)

>fucking hell...live video...i just remembered that deepfakes is a thing...
wise user is wise, that is a form of AI my friend using python.

>she could do an unbelievable amount of damage with that capability...the mind boggles
While the truth can hurt, it has to be done. The tech exists, but for home use give it a year or so for intel to get its shit together again.

soon user... soon... think big, and never give up on the dream. Twitter is played out for AI, been there done that, bought the botnet soundtrack and wore it out. The next level which is what you should shoot for, is not just tweeting, but an AI run youtube channel, non monetized of course.

I feel worse than a black father, if they were capable of emotion. The feels I have knowing that I have contributed to the robot uprising and fears I have of the unspeakable things that will be done to me on the people farms. These feels are unreal.

Attached: 4C31BC69-43B3-4E6B-9BC8-FB8799D4BF5B.png (750x1334, 222K)

ERIKA

Attached: 0080.png (1023x667, 330K)