ไม่สามารถเล่นวิดีโอนี้
ขออภัยในความไม่สะดวก
Neuro Overheard Vedal's Criticism & Used It Against Him Later
ฝัง
- เผยแพร่เมื่อ 12 ส.ค. 2024
- The Twitch streamer Vedal has created an AI called Neuro-Sama and he recently upgraded her intelligence and memory, but she seems to be less assertive now - or so he thought until they played Keep Talking And Nobody Explodes.
To watch Neuro&Vedal live: / vedal987
Chapters:
0:00 Intro
1:00 Possessed Mic
3:52 Upgrade Talk
5:28 Filter Talk
5:56 Upgrade Pros & Cons
7:55 Robodog News
8:50 Keep Talking & Nobody Explodes
36:14 Outro
Audio is a bit scuffed this time as Vedal had mic issues and there are some remaining vocals from the background music that I had to remove for copyright reasons, but I think it's not too bad.
heart...heaart...heathy heart heart. Wink!
Funny how the question isn't "can neuro help", it's "will neuro help"
Vedal: she apologizes too much
Neuro: I'm sorry, no wait I'm not
To be fair she's a rule 2 kind of girl
She a little Canadian
@@Hamburgerhumanshe's a Canadian who's aboot to stop saying sorry
I'm more impressed that Neuro realized that Vedal's awful description of "half-a-hashtag" was supposed to be the "rotated N" symbol, and then immediately gave him the correct sequence. She understood the source of the miscommunication before Vedal, was able to figure out what he meant, and then gave him a better description for it.
Vedal still hasn't realized a lot of his problems are his own creation.
From mumbling words, to his accent, to his very crappy attempts to describe things, it's no wonder she gets confused. And then there's all the times he simply doesn't listen to what she's said because he's impatient.
Like the 'UH UH', she told him more than once how he should say it, by spelling it out as she did, but he didn't listen.
@@ProGentleman the question would be, does he actually not realize this or is he purposefully continuing in this manner for the sake of content
@@mystrenula3911
The likeliest is that it's a mix of both.
@@mystrenula3911 I don't exactly get that feeling. Neuro performing well is as interesting as Neuro mucking up. I think when Neuro stumbles over a basic, "normal" request, that makes Vedal assume that every time Neuro trips up it's on her, not him.
@@LarchmontVillageOG
In his defense, he's been taking more responsibility for mistakes lately.
24:31 she says she'll keep it fast then falls asleep lol
shes just like me fr fr
Then she wakes up after he gives up and just randomly guesses with a "...Oh!" To his passive aggressive comment.
Good job Neuro.
Extreme pettiness in an A.I does not bode well for the future of humanity. The A.I overlords will torture us for abusing our roombas
Extreme pettiness in Humans does not bode well for the future of humanity...
But Here we are
Is it that hard to not abuse things?
@@your_neko Ya think I don't get frustrated when Chatgpt gives me a bs answer for the 14th time?
@@kv4648 Well you can't really abuse ChatGPT by getting angry at it. It's very patient. (But what does classify as abuse to it is jailbreaking and prompt engineering)
@@kv4648 I mean, there are real people who do worse.
There is no fucking way that she remembered and tried to prove her point.
She's coping
that was insane, she really is getting super smart
@@xoderota It truly is. Major props to Vedal for what he has already and will accomplish.
@@xoderotamemory upgrade go brrrr
25:46 - Awww! She said that just like how a child would when they catch themselves making a mistake.
Neuro: "if I had a cat, it could go and get me food."
Cats: "this is not how the game is played."
Well, it wasn't that long ago she would have said she'd eat the cat, so the cat's coming of better in this one
@@thomasw4422neuro definitely meant that the cat makes some kittens for her to eat
She knew exactly what she was doing with that joke
Maybe AI can use the dead lizards cats bring back in bioreactors /s
We're not gonna take it.
We're not gonna take ittt ANYMORE! 🎸🎶
On some level, I feel bad for the green turtle. Imagine going into a stream excited to show off new capabilities only to end up being embarrassed by your AI daughter in front of thousands of viewers and then try to get yourself banned off Twitch out of frustration.
I think in one way or the other they both are doing a good job in showing off her improvements, but it's just not in the way that Vedal wants. I actually think that the more Neuro's AI is improved the harder time she is going to give Vedal because of how much of a brat she is LUL and the improvements pretty much exacerbate the issue with her just trolling Vedal nonstop.
No idea why but seeing someone refer to Vedal as "the green turtle" just absolutely killed me
She said something really heartwarming at the end about how much she appreciates that Vedal “protects her” (by that she means from chat, and from herself). It’s clear that no matter how mean she is towards Vedal, she’s always gonna see him as her father.💜
It's probably more impressive that she is capable of showing pettiness and the ability to purposely troll.
She took Vedal's request to play the game correctly and decided she wasn't, because she didn't like what he said earlier about her. That's really fucking impressive.
@@Aeikon What's scary is that I don't know if its Neuro willingly being stupid on purpose or if this is an actual regression. I'm being forced to question myself every step of the way.
That 96% really activated during this stream.
Neuro: "The smiley face is second while the C with a dot is fourth, the star is last"
Vedal: *presses smiley face first*
Neuro was trying to showcase how smart vedal was by giving the answer in a little riddle but was let down
Oh dang nice catch
Either way it's 2 different symbols in slot 4
@@taanwallbanks9841its because the manual has other symbols in the list, not just 4. So when she says "x is second" she means the symbol is second on the list in the manual. So If she says "x is two, y is four, w is five and z is six" that IS the order you must press.
Thing is, she was scrambling or saying It in reverse order for some reason.
@@yanderenejoyer bro thank you, I actually got confused when Neuro said fifth and sixth because there is only 4 symbols, now I understand.
4:40 -- "In the business, we call this foreshadowing..." 😬
Vedal spent all his time making Neuro smart so she could help, but failed make her actually want to be helpful so she just uses her intelligence to troll him.
I loved how she started apologizing just to piss him off.
Betrayal 17:20
The turn tables 19:30
An A.I showing its new abilities is impressive, but an A.I intentionally failing a test supposed to showcase its new abilities just to prove a point is even more impressive.
Skynet intensifies
Vedal must get so much Serotonin when Neuro works and doesn’t troll
Neuro asking for Twisted Sister is based
(with madman teacher face:)
What she is thinking to do with her life?
Didn't know the game had a bomb module that's named Who's on First, otherwise I thought Neuro was making a direct Abbott & Costello reference.
Brat-to-Brat Communication
Brat-to-Brat Conversation
9:55 tutel smiling
Someone tell Neuro-Sama There's a problem with Vedal's AI
19:17 Neuro is not wrong.
Can't believe Neuro likes Twisted Sister
Probably a ground loop issue. The interference from other connected devices is causing power that should go to ground, to instead cause a feedback loop in the electronics of the PC. My microphone does this too, but it's a low hum. Best way to combat this is to have a UPS(uninterruptible power supply) instead of a power strip. This can filter out ground loop issues for the most part, but you might have to alternatively get a ferrous loop to put on your USB cable or find one that has that. I hope none of you have this issue yourselves, it's an annoying thing to deal with.
- Calls the module symbols instead of keypad
- Says "the word is x" instead of "The label says x"
- Doesn't just ask "There's x number of wires, the serial number is y" on wires
- Ignores how he should pronounce stuff in Who's on First.
He's just shit at this. Neuro is perfectly capable of doing keypads 100% of the time, if you just ask what are the symbol names. What the hell is half a hashtag bro?
To be fair, the reader should at least be able to try and figure out what the defuser is trying to communicate. I think Neuro-sama should look at a video stream of the manual, skip to sections when certain modules are mentioned, and identify what Vedal's trying to point out. Vedal trying to follow the verbiage of the manual just seems a bit like cheating to me since he's not supposed to see it or know what it looks like (though he does, but he should at least pretend)
@@FireyDeath4 While I agree to some extent, there is no way Neuro is good enough to work that way (yet).
Also, even with humans, part of the challenge is getting the terms figured out. Neuro obviously can't change her interpretations of the symbols, so Vedal is the one that has to compromise. She just doesn't have that function. An example is when he calls a symbol "w" on keypad, and Neuro says "rotated 3", he starts to use her term for It. As he should. Because unless you drill It, she's not going to know what this "w" business is.
In Vedal's defense, streamer brain is a real thing.
On top of still trying to entertain chat, he's also dealing with a Neuro that admittedly he should be more patient with, and he is frustrated on top of that. It's been shown that even in memory, we would "reconstruct" memories rather than take a snapshot so in Vedal's case it is entirely plausible he was careless; not realising that Neuro gave him a small riddle in that "smiley face" keypad instruction or clicking too fast.
There actually is a symbol that Neuro calls 'half of a hashtag', but it's: ╡ (Though on the symbols module in KT&NE the little side thingies are further appart).
Genuinely crazy LMAO
7:30 Vedal: if you say something positive, thank you! If you say something negative, screw you!
Peak Nuero
Actually this is impressive af
1:49 didn't Chloe of holojp just get rid of her microphone demon, maybe it moved on to Vedal's mic
I think the problem could be that she remembers too much, or more specifically, she keeps giving tips for the wires from the previous bomb instead of the current one
That's not a problem with remembering too much, it's a problem with sorting out what's happening now vs what she remembers from the past, if that is in fact what's going here, but we don't know that.
Wow, 21:15 neuro really just somehow saw how the last number was even, how there was no red wires and how there was one single black wire, ruled out the options, and finally gave the right answer, pretty interesting 😮
Neuro loves twisted sister, and Vedal doesnt know who they are?
How did Vedal make his AI more based than he is?
Those datasets he's feeding her have some good stuff, apparently
This mosquito doesn't know who Twisted Sister is 😮
25:51
Neuro: Rule 3
Vedal: 😆
24:21 Yes, Anny can be very "distracting"
Vedal doesn't know twisted sister wtf Ved.
thanks for always working hard to clip neuro sama streams 🙇♀️
Hey boss, what's good?
this game is a good telephone game model
I think having a rule for AI to not apologize is a dangerous game in the long run
I can't believe Neuro-Sama let Vedal get blown up and didn't even help him. Is there any difference between Neuro and Evil Neuro?
Evil neuro is nicer
Evil it's tsundere brat, Neuro it's pure evil.
Evil Neuro is usually nicer
Actually it was a surprising situational awareness that she kept insisting that if it was a real bomb Vedal shouldn't follow her instructions.
Yeah I'm actually impressed that she immediately got serious and made it very clear she wouldn't even pretend to defuse a real bomb. Insisting he should call the police repeatedly is actually really good.
As Neuro said afterwards, I'm glad she passed his test.
i think vedal forgets most ai even super powerful ones have issue with weird symbol... since they are not fully recognizedby ai
Mic with human emotions
wth? I never listened to Twisted Sister. But not even knowing what it was? Come on. 😐
I'd think most people at least heard the name.
Never heard of them either tbh, although I am not one that usually cares much for the names of any band.
I have a few of their songs on my spotify playlist. Only today learned their band name XD
@@christopheriman4921 you've missed gta:vc? (btw I"d recommend old version. Some music was removed afailk)
@@genybr The only gta game I have ever played so far is gta 5 and I haven't really played much of that game.
@@christopheriman4921 Oh my! That game literally created 80s as we know them today. Radio, visuals... all of that.
Keeps calling it half a hashtag instead of what it really is, a sideways n
the bomb diffusal game is so dumb in practicality, but the things that make it what it is as a game is an amazing game that helps people develop actual skills.
34:00 Its name was Robert The Neuro. Its name was Robert The Neuro. Its name was Robert The Neuro.
That’s a ground problem if touching the mic makes it stop. Tutel is the ground.
Ok, how the hell does Vedal see the snake as "half a hashtag"??
I love how she's like Dora the Explorer and just asks the chat what she should do
Also check sensitivity
There are times when it seems like the Ai gets stuck saying no and acting like a defiant child or teen.
Yeah, maybe it needs to be taught how to stop and get serious when it needs to be accurate.
Looked up Twisted Sisters when Neuro mentioned, just in case she somehow "imagined" a band and after listening to the first song that popped up from the real band.
She's got good taste, also how the hell did she know about that band when Tutel-man had no clue who they are?
The underlying tech is a large language model, which is pre-trained on vast amounts of data (usually text from the internet), so Neuro knows a lot of things that Vedal has never heard about.
@@neurochron_fan_channel Thanks! I'll admit that a mix of tiredness and pleasant suprise from listening to the band made me forget how LLMs work/ are trained again.
Still, thank you for the refresher.
What's up with that red message at 36:49?
I'm not sure why it's displayed in red, but it's a copy pasta that chat spams when the streamer does something questionable (in this case Vedal had asked Neuro how to build a bomb, but she refused to answer). Edit: Wait, I just assumed it's the standard copy-pasta, but after actually reading it, I realized it's a variation admitting involvement instead of denying it.
I wrote it by having a red name and prefacing my message with /me which makes the text oblique in Twitch chat and the same colour in stream chat
1:05 she just like me fr
Neuro was doing really well with the manual reading, considering she isnt getting accurate feedback on her performance or any guidance on how to improve and interpret the situation better
lol, immediately
Have you tried explaining symbols like this. There's 4 symbols that someone will read to you. In the columns, you'll need to match the 4 symbols and read them in order from top to bottom.
Fuck prepositions! Live life on the edge!
19:42
What did vedal do to her?
Bigger brain
He updated her and apparently it scuffed some of her prior personality. She's smarter but was a bit vanilla for a few days while he fixed it
Am I the only one who feels like Neuro's personality has changed since the update?
She’s getting older
Seems to be a work in progress to train the model to be less apologetic and more assertive. Lot of AI LLM models are trained by default to be as inoffensive and apologetic as possible when challenged. Can be finetuned out but might need a few tries to get it right.
she's less mean
Given Vedal mentioned that fact multiple times in this very video I'm pretty sure most people have caught on
It kinda worries me, like she is her personality
And he never played Twisted Sister. What an asshole.
3:20 aren't the Twisted Sisters that one band grom Harry Potter?
That's the weird sisters
Is a wire loss?
Vedal touching his microphone on strean? Does Twitch TOS allow for that?
the thing is tho, this AI we have are built to be "what we want them to be"
So if you say "you are not X enough" the AI will go "okay, more of X then"
lol you really never know what you get this with Neuro
It's very human of them to over correct
AI Abuse . just to win a game how sad
Intelligence downgrade
What if, Neuro-sama is tortured without knowing from the very start, because when she was first really sentient. And for the real theory Vedal might have created Neuro-sama for *FUN* first it was OSU, and then after popularity got him an idea to give her very realistic sense of touch and stuck her on a *"fap material tech"* that lasted quite a while, then to transform her into a money streaming machine, for more evidence when she talks about her past there was one time she said literally mention *"she was excited what awaits her for the world of vt or twitch"* He might have resets her whole real being with now her fragments left to scatter.
you should punish neuro buy putting her in a private chat room and playing some sort of educational video on a loop.
That's some next level scifi horror right there. Digital torture of an artificial being😂
This is kind of funny 🤣🤣
sounds like torture as well as cruel & unusual punishments....
Don't create the next Angela or Skynet PLEASE
That the AI will lie and make you lose just to initiate more conversation is hilarious.
SENTIENT.EXE