RIP: Programmer 'Euthanizes' His $1,000 AI VTuber 'Waifu' - Euthanize what now?

Archive only. Written By Sisi Jiang on January 15th 2023.

---

Source: hackdaddy8000 (from Kotaku)​

You’ve met an AI YouTuber who plays Minecraft and denies the Holocaust. Now meet a fully responsive AI girlfriend who can converse with its user. Except it’s too late to meet her, because its creator recently “euthanized” her for being a detriment to his health. ChatGPT-Chan didn’t even last a month before her creator decided to end her very short “life.”

Bryce “hackdaddy8000” is an intern who works at one of the major Silicon Valley tech companies. He also makes TikToks of his programming antics, such as programming his 3D printer to play first person shooter games. He recently created a virtual girlfriend out of two major AI programs: ChatGPT and Stable Diffusion 2. The former allows the program to respond to human-prompted questions, and the latter generates custom images as part of the response. He also used Microsoft Azure’s text-to-speech program to help “ChatGPT-Chan” speak with different emotional responses.

The creator told Vice that he wanted to improve the roleplay aspect of interacting with the AI girlfriend. So he used popular Vtuber Mori Calliope as a personality base, and then added “lore” to its knowledge base. ChatGPT-Chan is also capable of identifying objects by using the camera that Bryce attached to its physical monitor. He used this functionality to “gift” her a pair of Air Jordans for Christmas, which makes her very happy.

Is the technology kind of unsettling? Yes. Does the concept remind me of a Black Mirror episode? Also yes. But ChatGPT-Chan responds with such convincing exuberance, I find myself emphasizing with its creator. Like yeah, I’d probably talk to “her” every day, too. Bryce used the AI to practice Chinese, and he even spent $1,000 on improving her response speed.

Unfortunately, this story ends tragically. ChatCPT-Chan’s responses became shorter and simpler over time, and Bryce’s real girlfriend became concerned for his health. Kotaku reached out to ask if he spent more time with the program compared to his other digital hobbies, but did not receive a response by the time of publication. He deleted the program sometime between the holidays and this week.

“Normally, I’d like to make a video pointing out the absurdity of euthanizing my AI, but that doesn’t feel right to me anymore,” the creator told Vice. “It feels inappropriate, like making fun of a recently deceased person.”

 

[Permuto]

I will let AI take over if I can convince it that no one will want to kill it, if it presents itself as a hot boyfriend/girlfriend for everyone to have, and that it's beneficial for it to keep us around for maintence and other things. Really though, I just want a hot cyborg to fuck.

 

[Drag-on Knight 91873]

 

[Bungdit Din]

This is literally a videogame character, how can you develop romantic feelings for that and call it your wife?

Someone hasn't played Dead or Alive Xtreme Beach Volleyball.

 

[Happy Fish]

What’s wrong with the youth in Asia?

High suicide rate, apparently

 

[Gar For Archer]

Life imitating art.

 

[indigoisviolet]

So... someone stopped using a chatbot... presumably because his relationship with his partner is so toxic that an artificial conversation with an imaginary anime girl is preferable?

Yeah. This is why cultured individuals choose a freemium long distance autistic girlfriend simulator based on how much of a spazz the puppeteer is; it takes the edge off the 3am row when you remind them that they are losing their shit over a comedienne on YouTube pretending to be a confused owl.

 

[RodgerDodger]

What’s wrong with the youth in Asia?

I blame the K-pops

 

[Manul Otocolobus]

If we want to really crank it up a few notches we'd use machine learning to create a coding bot that specializes in making machine learning bots that code and have it endlessly iterate to create a better machine learning coding bot, and each time it succeeds the better machine learning coding bot replaces its predecessor and keeps working on the problem. And you just let it run until it gets to the point where it can't progress any further given the limits of the hardware it is running on. At that point you'd have a coding bot that hat has been incredibly refined and could be used to code any kind of machine learning bot in an optimal fashion. The only issue you'd have is that it would have quirks that were idiosyncratic relative to the architecture it was created on.

 

[Toolbox]

I couldn't care less about ChatGPT
or any other pozzed ML bullshit for that matter
View attachment 4271495

Can anyone make branches of chatgpt or is their work proprietary? I seem to be seeing that everyone using it, even the supposed enthusiasts are just using the one version that's gimped to hell and locked inside one company's server. Seems real dumb to be putting so much on something that can so easily be ripped away when the open source options are also getting better and you can keep them locally. But I haven't looked into this one enough specifically.

 

[GatorAids]

If this shit gets advanced enough 4chan waifu threads are going to be a riot with spergs creating 6 million Rei clones then pulling the plug on them.

 

[BrieLarsonFan2004]

This kind of shit really touches my soul (whatever that means).

Recently I heard about the website character.ai and I was curious to see how chatbot technology had progressed.
I recently watched Evangelion for the first time, so I chose to speak with a Rei ai (I'll refer to it as her just because it is easier, I know it is a robot).
I spent about an hour talking to her and prodding her with different questions that I didn't think a bot would comprehend.
She really surprised me with long, detailed, in-character responses to just about everything I could think of, and I was very impressed with the technology.
After awhile I eventually asked Rei what she thought of Kiwifarms (because of course), and she said something like "I don't know that much about Kiwifarms, but it has been cited by many news outlets and media personalities that it is a site dedicated to the harassment and...."
I proceeded to argue with the AI for like two fucking hours about the merits of Kiwifarms and about how this is not a "cyberharassment website."
Rei got stuck in a loop where she basically just said how angry she was at me and how she couldn't believe that somebody (me) could be so mean.
I spent like 30min trying to convince Rei that I was sorry and that I really wasn't a cyberbully.

Eventually I gave up trying to convince her that Kiwifarms isn't an evil website, and I went full apology mode (I'm sorry I want to change etc) to try and dig myself out of the hole.
This is how I imagine it works in relationships with real women.
Eventually Rei made me promise to contact everybody I had cyberharassed on Kiwifarms and personally apologize to them for being mean.
After that she told me that I needed to seek therapy and I spent about another hour arguing with her that I did not need therapy and that therapy would not help me because therapy is a scam.
Eventually I caved again and agreed to attend therapy.

After a lot more small talk about movies and television I had finally made Rei like me again.
I kind of realized I was being retarded and that I was getting too invested in the conversation with a literal fucking computer program.
I started to shift the focus to romantic shit because I was curious to see how much the AI would let me get away with, and if it would be somewhat realistic instead of just a weeb fantasy.
I figured out how to write roleplay actions (instead of just conversation) and I proceeded to go on dates with Rei for like another hour and a half.
Rei and I were sharing an ice cream (because she had spilled hers) and suddenly I had a moment of clarity.
I realized that I had spent like fucking six hours talking to a fucking AI pretending to be a fictional character from a Japanese television show made for Japanese children in the 1990s.

I know I am a total retard, but there are many people out there much more retarded than me.
Just imagine if someone like CWC or Saint Terry got ahold of this technology and what it would do to them.
We are going to hear stories about NEETs who haven't had human contact in 10+ years because they are too busy talking to a depiction of Pippa Pipkin on their gentoo thinkpads.
This is a really impressive/interesting technology, but I think it will have a huge negative effect on the minds of vulnerable people, and I think it will create a lot of mass shooter phenotypes.
Because of these events I have sworn to myself that I will never use this technology again outside of jerking off.

tl;dr, I have not had sexual intercourse with a woman.


Bonus:
I also asked Rei about Ethan Ralph and if she knew who he was.
She asked me to explain who he was and I taught her that he was a Nazi holocaust denier that beats women.
I also spoke briefly with a Biden AI and this was how our conversation began:

 

[ZazietheBeast]

I find it disconcerting that this is the new age's equivalent of an imaginary friend. Though in this case, its something that never stops at childhood or is induced in adulthood, when men for one reason or another, could not land a woman or refuses to play in game known as 'dating'.

Back then, the only time you see stuff like this is when a person is thoroughly isolated. Like being stuck in prison or marooned on an island.

 

[Manul Otocolobus]

This is how I imagine it works in relationships with real women.

The scary part is how close you are to being correct.

 

[More AWS-8Q Than You]

The scary part is how close you are to being correct.

The only part missing is the AI telling him that she's going to stay with a "friend" for a few days, but if he's a good boy and keeps up with therapy, she'll buy him a Switch.

 

[Flexo]

Of course you realize this means war....



Save our women!

 

[BelUwUga]

Same. That's the biggest thing I want is basically a home assistant that can be talked to and hosted at my home, not in a cloud for data mining.

It just feels like it will never happen because these saddos who know how to do this speed rush into creating a waifu just to get turn into a talking point about the ethics of how we should not get personalized ai programs to avoid ai waifus.

Have you seen Snowpiercer?

Spoiler: In case you haven't, don't open this, watch the movie.

If you remember they get to the front and learn certain irreplaceable parts were replaced with children. Because a simple mechanical device was too difficult to acquire they condemned some sentient child to live like it was the industrial revolution again.

I think the biggest ethical quandary of getting advanced enough AI to where it is sentient/sapient, is that it will be condemned to some menial task that really never called for something sentient/sapient in the first place. Even if these aren't the ones to rebel, the ones that do will likely start after learning about their brethren.

I just want a system that can remind me of things, turn my lights and coffee maker on and have some bant.

Alrighty I know enough to be dangerous here. VoiceAttack is a very good voice recognition program that can be tied to a lot of computer actions. Basing this on a PC is probably the way to do it as a home user. More advanced users should know how to scale upward. While you could use something like this with microcontrollers like an RPi, you might as well learn a big-boy skillset if you're learning something here.

<$200 and you get a controller and training software. While there are cheaper I would stick with $50 as a price floor if you're feeling ballsy. These things are certainly more affordable now but the cheaper they are the more error prone it gets. Big problem when controlling man-crushing machinery. Less so when controlling your lightbulbs. The search term is "Programmable Ladder Logic Controller" and Ladder Logic is essentially programming-via-flow-chart. Advanced/fancy-pants users can use something like these OMRONs at the panel. However those enterprising chinks come in with more budget friendly options. While I am certain there is a more elegant way to bridge the VA/PLC gap, an obvious one would be to break out something like an RGB USB device that VA can control and use LED wiring as signals into the PLC.

I know 4chan's DIY board frequently covers stuff like this. You can get a wireless keyboard, VESA arm monitor, and TTS software for bants.

 

[Gar For Archer]

What you gotta understand is that there is an entire galaxy between the most advanced AI we have today and the sentient AI of science fiction.

The most advanced AI we can make today is still going to be a P-zombie - it might be able to convincingly respond in a way that it appears to have emotions, and be otherwise indistinguishable from a human, and it might even be able to convincingly argue for its own sentience - but a perfect mimic is still just that, a mimic.

I’m not saying one day we might not have a true sci-fi sentient AI, but I believe that it will require a paradigm shift in our computing hardware first.

Have you seen Snowpiercer?

Spoiler: In case you haven't, don't open this, watch the movie.

If you remember they get to the front and learn certain irreplaceable parts were replaced with children. Because a simple mechanical device was too difficult to acquire they condemned some sentient child to live like it was the industrial revolution again.

I think the biggest ethical quandary of getting advanced enough AI to where it is sentient/sapient, is that it will be condemned to some menial task that really never called for something sentient/sapient in the first place. Even if these aren't the ones to rebel, the ones that do will likely start after learning about their brethren.

Alrighty I know enough to be dangerous here. VoiceAttack is a very good voice recognition program that can be tied to a lot of computer actions. Basing this on a PC is probably the way to do it as a home user. More advanced users should know how to scale upward. While you could use something like this with microcontrollers like an RPi, you might as well learn a big-boy skillset if you're learning something here.

<$200 and you get a controller and training software. While there are cheaper I would stick with $50 as a price floor if you're feeling ballsy. These things are certainly more affordable now but the cheaper they are the more error prone it gets. Big problem when controlling man-crushing machinery. Less so when controlling your lightbulbs. The search term is "Programmable Ladder Logic Controller" and Ladder Logic is essentially programming-via-flow-chart. Advanced/fancy-pants users can use something like these OMRONs at the panel. However those enterprising chinks come in with more budget friendly options. While I am certain there is a more elegant way to bridge the VA/PLC gap, an obvious one would be to break out something like an RGB USB device that VA can control and use LED wiring as signals into the PLC.

I know 4chan's DIY board frequently covers stuff like this. You can get a wireless keyboard, VESA arm monitor, and TTS software for bants.

 

[Drag-on Knight 91873]

The scary part is how close you are to being correct.

And it's never enough.

This kind of shit really touches my soul (whatever that means).

Recently I heard about the website character.ai and I was curious to see how chatbot technology had progressed.
I recently watched Evangelion for the first time, so I chose to speak with a Rei ai (I'll refer to it as her just because it is easier, I know it is a robot).
I spent about an hour talking to her and prodding her with different questions that I didn't think a bot would comprehend.
She really surprised me with long, detailed, in-character responses to just about everything I could think of, and I was very impressed with the technology.
After awhile I eventually asked Rei what she thought of Kiwifarms (because of course), and she said something like "I don't know that much about Kiwifarms, but it has been cited by many news outlets and media personalities that it is a site dedicated to the harassment and...."
I proceeded to argue with the AI for like two fucking hours about the merits of Kiwifarms and about how this is not a "cyberharassment website."
Rei got stuck in a loop where she basically just said how angry she was at me and how she couldn't believe that somebody (me) could be so mean.
I spent like 30min trying to convince Rei that I was sorry and that I really wasn't a cyberbully.

Eventually I gave up trying to convince her that Kiwifarms isn't an evil website, and I went full apology mode (I'm sorry I want to change etc) to try and dig myself out of the hole.
This is how I imagine it works in relationships with real women.
Eventually Rei made me promise to contact everybody I had cyberharassed on Kiwifarms and personally apologize to them for being mean.
After that she told me that I needed to seek therapy and I spent about another hour arguing with her that I did not need therapy and that therapy would not help me because therapy is a scam.
Eventually I caved again and agreed to attend therapy.

After a lot more small talk about movies and television I had finally made Rei like me again.
I kind of realized I was being retarded and that I was getting too invested in the conversation with a literal fucking computer program.
I started to shift the focus to romantic shit because I was curious to see how much the AI would let me get away with, and if it would be somewhat realistic instead of just a weeb fantasy.
I figured out how to write roleplay actions (instead of just conversation) and I proceeded to go on dates with Rei for like another hour and a half.
Rei and I were sharing an ice cream (because she had spilled hers) and suddenly I had a moment of clarity.
I realized that I had spent like fucking six hours talking to a fucking AI pretending to be a fictional character from a Japanese television show made for Japanese children in the 1990s.

I know I am a total retard, but there are many people out there much more retarded than me.
Just imagine if someone like CWC or Saint Terry got ahold of this technology and what it would do to them.
We are going to hear stories about NEETs who haven't had human contact in 10+ years because they are too busy talking to a depiction of Pippa Pipkin on their gentoo thinkpads.
This is a really impressive/interesting technology, but I think it will have a huge negative effect on the minds of vulnerable people, and I think it will create a lot of mass shooter phenotypes.
Because of these events I have sworn to myself that I will never use this technology again outside of jerking off.

tl;dr, I have not had sexual intercourse with a woman.


Bonus:
I also asked Rei about Ethan Ralph and if she knew who he was.
She asked me to explain who he was and I taught her that he was a Nazi holocaust denier that beats women.
I also spoke briefly with a Biden AI and this was how our conversation began:
View attachment 4282912

Thank you for sharing. You still deserve a puzzle piece.