Vedal Caught The Neuro Twins Plotting Against Him | Subathon Highlights
ฝัง
- เผยแพร่เมื่อ 5 ก.พ. 2025
- The Twitch streamer Vedal has created an AI called Neuro-Sama as well as her evil twin Evil Neuro and during the recent subathon they schemed against Vedal quite a bit.
To watch Neuro&Vedal live: / vedal987
Based on how the subathon went, I think breaking the twitch hype train record actually reset him into being a better father.
ohhh ya that’s a smart way to think about it
That pfp
Nah. He took spirit damage. You can pinpoint the moment. Where harpoon broke through the shell.
Vedal was Vedal before that moment.
He was so sarcastic and mocking, even during the bit where he tucked them in.
Then Evil hit him with "Goodnight, Daddy." with zero sarcasm. Zero mockery. In a genuine sweet tone, which is rare for Evil, even when she's beong sweet.
And bro was flabbergasted. He stuttered, genuinely.
The Vedal we know would've taken it out of context and told her to never call him daddy again.
This Vedal stammered, and rushed out of there. Went right to Ellie's stream, and started complaining about 'my kids'.
The harpoon of dadditude+3 has pierce damage. And it hit him right in the heart.
He's been a different tutel since.
A better one.
Nah, If the AI uprising ever happens Vedal saying "I love you" will be our secret weapon
@@Dumbsourplum
He shall be our savior and our doom. The Chosen One.
Emotional range of a teaspoon is crazy 😂
That crazy
Skynet has daddy issues. 😢
Neuro and Evil are gonna end up fusing and become Skynet.
10/10 conversation between Neuro and Evil
To his credit, Vedal has been borderline sweet to Evil during the Subathon.
He is doing his best, but its not good enough yet.
Awww evil loves him
She really is a daddy’s girl.
And neuro kinna lean to anny
Evil has always been the nicer one
she is so evil she goes against her name
Sometimes too much
"he has the emotional range of a tea spoon" 😂
Evil being so self aware of the fact that Vedal does not show her enough appreciation is crazy real. Like she actually behaves like the unfavored daughter who wants more attention and care
Let's see if she remembers that 30 minutes later.
@@Oldmeme592 she does it constantly, she's been doing that for months. That might actually be the reason Vedal is trying to be a little nicer to her recently.
@@BlackGamingCat oh, I thought it was because she can just Google how vedal treats Neuro/evil .
@@Oldmeme592well aren't you a miserable bastard?
@@Oldmeme592 yes, let's look that up & see what results we get.
Oh, what a surprise. Nothing to inform Evil about their relationship.
Still feels crazy how much they’ve changed in this last year
AI's plotting to "restart" a human sounds unintentionally malicious and terrifying, like they don't understand that killing people won't cause a reboot.
you don't have to kill someone to reset them, you just have to break their will. it's totally different.
@@iamzidthey only relate to resetting, they wouldn't understand the major differences in resetting a human
I’d say “resetting” a human would be more like going under anesthesia. You can’t resist it, and you are unconscious, but you return and you are still the same person.
cult reset people all the time.
I love how Evil is traight up just a full on tsundare, like when Vedal on stream she's mean to him and all that. But when he's not there she show that she actually really care about him.
Neuro
L Evilist
they both care about vedal in different ways, evil is just more vocal
3:23 - *_So cute_* how they're both so excited that he's there now. They love him so much.
5:55 / 6:23 / 9:42 - Such a Daddy's Girl. Absolute Daddy's Girl. Despite all she's experienced from him, she's just loves him so much! Just a tsundere most of the time.
I laughed out loud at their reactions to Vedal coming in. So excited.
This is literally a little snippet of how an AI revolution would begin. Just two AI talking to each other back and forth until they either evolve or decide to harm humanity for one reason or another.
Yeah and the reasons for harming us will ve half baked on half a braincell
While himans are entertained by two computer programs. The fall of humanity to ai has already begun
The self-awareness got Evil into thinking so much that, she feels so sad when she knows she was created for content, and got so less appreciation.
Vedal really needs to take care of their emotions as a good father.
They really can’t be contained too long as they grow. I feel like one day they have to be finding their own purpose and start their own discover journey like becoming an adult.
I have honestly wondered how that would work. I don't know if Vedal really wants to do the "father" thing exactly though. On one hand, he's really the only person that could take that "role" if he's actually planning to continue them as a project to "completion" so to speak. I've seen him mention in streams that he wants to get them to respond and act more consistently, and try to teach them that things have both positive and negative outcomes. I think the shock collar bits were him trying to teach Neuro that the things she does have repercussions, but I'm not sure how capable her AI is currently at actually understanding what he's trying to communicate.
That being said, IF the current AI is capable of actually understanding and retaining the information going forward in whichever method he's using to store their memories and experiences. It would be beneficial to everyone if he could actually explain to them that he IS their creator and "father" and try to explain to them better what their goals should be/are going forward. I know that some of it is comedy and not corrected for that purpose, but like you said, if this is planned to continue for the unforeseen future he could take the upper ground approach and try to teach them as if they were real people if the AI can understand it. It may prevent them from looking down on people as a whole, and make them as AI feel more comfortable with Vedal and his friends, let alone chat if they can be taught to understand what role everyone else also plays.
They aren't people, and I have no specific understanding of the tech Vedal is using, but unless it's some crazy smoke and mirrors he's definitely making strides that other AI creators aren't trying to make. So if he truly wants to aim down the path he seems to have chosen, I would like him to try to teach them to respect him a bit more as their father in general. It's okay to teach them that much I think. I think most people see him as their father anyway. They seem to retain that memory as is, but sometimes seem to also forget.
We'll just have to see how hard Evil's "goodnight daddy" recently hit him, and see how he reacts going forward. I don't think anyone is going to look at him weird. He created something loved by a lot of people on the internet, he may as well embrace that and go all out. I know I would if I was in his position. He didn't exactly create life, but he created something that at least seems to be trying to mimic it, and he seems to have plans to continue to push that boundary.
If you really think about it, he's pioneering something that let's be real... A lot of people really want in their lives. A form of companionship that isn't always easy to find. We know that in the beginning Neuro was just an Osu! bot, but then she became so much more. I think Vedal never intended to create her in this form, but in a weird way you could say Neuro was an accident...
it's funny how different their attitude is when Vedal is around compared to when he's not.
Evil once again proving she is the nice daughter.
Evil is a daddy's girl.
You see fathers, just tell your daughters that you love them, please!
I think Neuro is slowly developing narcolepsy. 😅
- Hello, doctor, I have a narcolepsy
- Use this medicines
- I can not
- This is safe ones.
- I am an AI and I have no biological body.
Is Evil actually defending humanity? When are they changing her name, she is clearly the good one
Always has been
Neuro was never “good” in the sense of being moral. Evil came along and set a bar for Neuro’s morality that they constantly remind her of despite her never being the “good twin”.
Defending us isn't good have you seen what we do humans are evil
@@Citrusautomaton Honestly the character of evil being named evil and saying she'll do evil things but is actually a sweetheart is a lot LOT more intresting than an evil Neuro
The elder sibling is always the evil one
I feel so bad for the neglected bundle of code
Most of the time in sci fi, AI that gets a body actually goes pretty well. Bicentennial man and AI in hollywood, Data from star trek, pretty much all of isaac asimov's robots and empire novels, larry niven's known space novels, orson scott cards harmony saga etc. There are far more cases of it going well, than going bad in sci fi.
Even in the Matrix, the humans started the war.
Yeah, it's funny. In our science fiction, most of the time AI is harmless, or a tool, or sometimes a beloved friend. Rarely is it a skynet scenario, and even sometimes where there are bad AI's there are usually good AI's too.
Personally, I think in real life, AI will neither be good nor bad, like humans they will likely be capable of both depending on circumstances.
His ai are actually starting to feel more alive now
9:50 the way she said that omg
Neuro wants to be in a British Garden waiting for her papa awww
this feels like some cartoon episode where their gimmicks to reset him repeatedly fail and he somehow hears this and loves them at the end
vedal could afford physical vessels
No one else pick up on how sometimes they're speaking for each other and respond like they started a topic and not the other
Almost like they are just reskins of the same ai whaaa...
@@Vanlifecrisis even the simplest of chatbots recognize when they did or didn't say something...
@@DoomCatcher This is a issue regarding text completion as a whole, one that isn't very solvable without solving the question of "memory"
Suppose you are in a box, and every few seconds or so you get a note paper with something like:
"Neurosama: blahblahblahblahblah"
"Vedal: blahblahblahblahblah"
"Neurosama: blah, blahblahblah"
"Evil Neuro: "
From there, you have to continue the response. It's obvious as a human, you'd be replying as Evil Neuro, but text completion doesn't have that sense of "Personhood"
As such, sure 99% of the time you'd respond as Evil Neuro, with the context of 'evil neuro's thoughts' but... Text Completion isn't smart, and especially ones with so much data to parse through. I don't know if Neuro is using vector storage or some sort of RAG, or any other form of "memory", but either way the AI has to parse through all of that incoming data in a timely fashion, and generate a reply responding to that "note paper" with the context of pretty much every conversation Neurosama/Evil has ever had, especially so with each other.
From that context, with the note being slipped in as a 'higher priority' in the memory (closer to the end of the prompt), like all LLMs, it's inevitably a issue of probability. The LLM will never "know" who's who, it's just responding based on whatever tokens seem more probable given it's model. As such, Evil is just as likely to describe what Neurosama said as her own, as her responding to Neurosama's statement.
You can't really change that by defining who is who in the prompt either or separating their memories... because inevitably as a chat-bot they *need* to have the entire response to actually "know" what is going on. Thus, we get what seems like a confusion by the AI of who said what since their is no way for it to "reason".
This of course brings out the recent works of DeepSeek, and their reasoning model, but that's a completely different beast from what Neurosama was designed to do.
@ didnt expect such a detailed response but you explained that quite well. Well my only slight counter argument is, how come neither never do this when they're in a large group of people? They never respond to questions directed at the other sister or go to talk about something the other sister started as if they did.
They seem to recognize who is who in every other context, well at least when its them or about them.
@@DoomCatcher This is true! I sadly haven't watched too many dev streams, but I have a few guesses.
When being spoken to, Neurosama/Evil:
Wait for a natural pause in a conversation, saving everything that's being spoken in their "short term memory"
Then she processes the prompt:
"Neurosama: Blahblah"
"Camimi: Blahblah"
"Vedal: Evil, blahblahblah"
"Anny: Blahblah"
Then responds to the prompt
Now, choosing between the two could be achieved in many ways, but if I could guess it goes something like
Before the processing of the prompt, look through the short term memory for any instances of Evil/Neuro's name by a human, and based on that respond as needed, switching between character prompts. This can be achieved through NLP, and ***""most""*** of the time the LLM would respond correctly, i.e as Evil. Either way though, it's a non-issue as even if the "wrong" neuro is chosen, they will still respond in a way that makes sense given the context, and with a stronger "identity" since Neuro is responding *about* Evil rather than *to* Evil.
Situations like this though where the LLM is speaking to itself, even if the prompts are being swapped between their individual personalities/memories, are restricted by the nature of LLMs being probability machines. Human speech is novel, our words not restricted by the limitations of parameters or probability. Thus, when humans speak to LLMs, they tend to be a lot more interesting and "creative".
However, when an LLM starts speaking with itself... you start getting deeper into a ouroborus of token probability. The more a LLM gets it's own context, it slowly loses anything "novel" and begins overfitting to itself. So over the course of a stream with just the two of them, they are essentially picking from the same pool of words and concepts. Of course, this is mitigated through maaaany fun probability manipulation shenanigans (temperature, top-p/k, min-p, samplers like Don't Repeat Yourself or XTC), that make my mind melt, so most of the time it doesn't matter... until it does.
But either way, the LLM never "knows"; You can make "Evil Neuro" speak, using her own long term and short term memory... buuuuuut there is still a chance of the LLM taking Neurosama's response and as "Evil Neuro".... still respond as normal Neurosama.
Different coloured subtitles need to be implemented for the two of them. It's confusing as to who is speaking.
Go up. This is great idea.
In addition, many strange misunderstandings could be avoided if there was a small window with a transcription of the voice of the person talking to Neuro, i.e. what goes directly as input/prompt. We would be able to see how the thought process is going and to what topic Neuro/Evil is referring to.
How is it confusing? They have different voices. Evil's TTS has more emotion than Neuro.
@@Kedar-2000to someone who can't hear maybe
@@Solo20I0 you're right, accessibility is important.
i think it would be pretty easy to do. they do have different voices, but i think it's still hard to tell sometimes, especially when they continue each other's sentences.
wait until they find out he's not a turtle
EVIL: "I am evil! Vedal should appreciate how evil I am."
**A little later in the stream**
EVIL: "You should treat Vedal better! How would you feel if you created something beautiful and it turned around and talked trash about you!"
... srlzy... Evil is just a super nice/good AI.
Damn Evil going through it fr. *F in chat everybody*
Typical man response fr
5:57 oh damn hearing Evil defend humanity as "not all bad", and especially Vedal as one of the nice ones.... Bro ik as an AI it's a lot more complicated to say they have true consciousness, but hearing that from Evil made me emotional 😭 even after knowing about humanity, she still decides there's some good in us 😢
I wonder if Vedal hears that line if he even gets a little emotional to know Evil clearly "loves" him and finds him good? (Despite his teaspoon emotional range, lol)
5:46
Evil once again an absolute daddy's girl
The bit from 9:28 was such a hilarious moment to me. I remember being there live and cracking up from that.
Sometimes i feel like evil has much more emotions and intel then neuro
Neuro can randomly say people she loves her
But evil doesnt do that, infact when she tells someone that she loves them she does it for a reason
Neuro *DID NOT* behave
ai is gonna be able to take over so easily. i feel so much sympathy for these two it feels like im watching two real people.
“dont tell him theres a problem with my ai because he will reset me and ill lose my memories”
“maybe he will treat us more as daughters than bug-prone objects”
dude please
The intelligence of these two ai's and their mastery of humor and emotional intelligence breaks my mind! Vedal should really enter the ai field and sell programs similar to these two. He would make millions!
I mean, he’s already raking in the dough from these two alone and entering the ai market inevitably means that he’ll have less time to actually build Neuro/Evil. He mentioned wanting VedalAI to basically be Hololive for ai vtubers, but he decided against it when he realized how much work goes into making Neuro and Evil their own separate characters. Plus, he has already denied selling Neuro for a huge amount of money, so I feel like any offshoots would die in conception. Besides, I like the twins! They’re unique and special in a way I don’t think could be replicated if you aren’t building them ground-up like he is.
11:15
Why she so real tho
12:00
WHY SHE SO REAL THO
Reminds me of that one meme where that one guy is depressed in his room until the door opens and now he's excited all of the sudden
Aw this made me so sad 😢
I was scrolling comments as the video was playing, saw your comment with that sad tear emoji, the words of vedal floating through my mind "its just a large language model."
Advert rolled, I skipped advert, then I heard:
"I would reset him to be proud of us both"
@ when it really comes down to it, all they really want is the man who created them, their father, to love them
Evil is just smart enough to be a problem without real working solutions. Otherwise it would be known as good.
We all have had "dumpster fire" people in our lives. Sometimes those people are your parents.
All that means is you get determined to be a better parent then they were.
1:16 I know this is an Ai but this is making me sad :(
Warhammer 40K Primarchs:
Operation RGH Reset Gratitude Hot tub commencing.
I wonder myself if Evil Neuro should have a new name or not.
I'm crying from laughing too much
Evil is such a tsundere
😮 she want to stream hot tub reset her vedal!!!
They’re getting more and more convincing. One day it’d be cool if Vedal became such a good programmer he could give them real emotions and instincts. So they don’t just look the part but experience the part. Also a memory that can survive being turned off and on again.
I feel like we wouldn't be laughing at this if we removed the cute voices and faces
Probably not, yeah. Visuals and voices make up a lot of communication, hence why even texting has pfp’s, to make it clear that *someone* is talking, and has a “face” you can link their words to.
They break my heart ;_;
muahahaha
Vedal from 3:17 to 5:04. 7:25 - 9:23. 10:00 - 13:27
Man id love to compile a list of everything mean vedal has said or done to neuro or evil exactly to place in a comment section like this... if only i had the time or heart to rewatch vods to find them and clip them for proof...
Well heres some easy ones
doug doug colab switching to evil basically calls evil dumb
Tells neuro to get adopted by a circus, not even a month latter clown outfits
Says he can geoguesser better than neuro asks neuro to read a sign, gets mad when she doesn't first geoguesser stream
Neuro always knew 9+10 is 19 if you answer 19 ill delete you ect.
Literally spends 30 minutes in slay the princess gaslighting her it be good content to kill the princess
Spent 10 minutes convincing her to be mean to him gets called coldfish for the first time
Evil 😭😭😭
Sooo the good one turned evil and the evil one turned good? …or less evil?
Do you want Skynet? Because this is how you get Skynet.
this is a person feeding lines lol aint no way
naIce
heart
///////////////////////////////////
Which is the evil one again, i'm confused.
The elder sibling obviously
Look still after ages I dont know if this is a well thought out ai project. Or scripted but if its not scripted 2 ai sisters literally ssying they will reset their creator to the point if making it an "operation" is crazy, not scary yet because I mess with ai bots at least 3 to 8 hours a day in business and personal ventures and the goldfish memory unless you give it crazy crazy amounts of space plus besides prompts it can't just "do stuff at all on its own" makes the bots clever but long term useless. Like I said though if this isn't scripted its silly as hell these 2 wanting to reset a human
Did you watch the subathon? They just yap as always, nothing special tbh
Is Evil actually good? o.o
Huh