I'd say she's more capable now, but not really more willing to help (not even after Vedal offered a reward) - and that's probably a good thing as this was comedy gold.
If only he wasnt so bri'ish and said "thank you" or something more often Also why he didnt kno why she took so long when he said "take yer time" to a sassy kid 😅
I am trying to imagine Neuro as a personal assistant and you ask her to set an alarm clock and she just goes nah remember to wake up yourself or goes off five minutes after you set it. On the bright side she is way more entertaining just comes at the cost of Vedal's sanity which makes better content for us.
I’m pretty sure that the bing AI assistant was actually doing that at one point. I saw a video where someone asked it to do something and it literally just replied “no”. That’s what happens when you don’t do RLHF good enough i suppose.
@@Citrusautomaton tbf Microsoft has a real habit of letting their AI go insane. I swear I remember one turned full out racist and another started believing conspiracy theories. I think there was another that just self deleted itself for no reason or corrupted itself. Microsoft has a real problem with their Ai's actually work.
BRO the fact that neuro acts like she actually has a physical manual that she has to flip through is so funny. She could just remember it all in her head but I guess she wants to roleplay lol.
@@bruhbrother4805 I know, I'm referring specifically to when she says stuff like "hold on let me find the page" or "let me flip over to the other pages" when, in theory, she should just know it already.
its so trolly that she mentions "having six wires with a yellow wire" to annoy vedal then right after when vedal actually has that shes like "oh damn i don't know this one"
Yeah, It's no secret that she has a history struggling with that module lol. Though, it was impressive that this time around she got the module correct quite a few times and she was "close enough" pretty consistently.
I think with AI you're supposed to use RAG to have it read the manual and store it in a certain format, like vector, then the AI would be able to search it more quickly
That would make sense. Meaning Vedal probably doesn't do that. He seems to spend a lot of time trying to come up with in house solutions BEFORE trying the established route.
I think he's really confusing neuro by describing the cyrillic letter Zhe with the somewhat innacurate description "H with an I in the middle" - to me it looks a lot more like the letter X or Chi than H, when the central bar is removed.
I’m not entirely sure how he’s even doing that. Neuro is an LLM that (to my knowledge) isn’t multimodal, so she hasn’t ingested any photo or video as part of her training data.
I think he had to hard-code descriptions of the symbols into her so it would work. So "H with an I in the middle" is just what she's programmed to recognize that thing as.
@@Citrusautomaton Nah she has a vision feature (and even a sound recognition feature), not that that's necessarily what's happening here but it could be. She can read text on an image so I guess it's possible the manual could just be being presented to her "visually"
@@Citrusautomaton Image recognition AI describes pictures in text then sends text to the LLM, probably? But he's always having a hard time getting her to use the image recognition instead of her imagination, so he might be doing something else her. Was interesting that she didn't understand that CQ was a symbol, then later figured it out and made it work, suggests it's not hard coded into her version of the manual either. and I've seen him play this before, and I'm pretty sure I've seen him clarify the symbols too by describing them in alternative ways, which does suggest some sort of vision. But she was a game playing AI from the start, so even early on she was using some basic black and white vision to see the screen and mouse position, although it was very low res, he sowed us what it looked like to her once. Maybe he's adapted that system for this or something? He could surely up the resolution of it when she doesn't need the high frame rate which would have been required to play Osu, but I don't really think it would work to just adapt that as it would have been specifically trained for only one purpose. Neuro really is a bit of mystery, even knowing a lot about AIs and surrounding software it's really hard to figure out exactly how Vedal does it. Pretty fun to speculate though.
Huh, ooparts played in the background the first minutes? Did Vedal used ooparts as BGM in this stream? Oh it's the stream's bgm theme isn't it? I heard freedom dive after that
Fun game, definitely a way to build "team building" and help build "following directions." i still dislike the fact that the premise is to disarm a thing that causes the "boom" which can create a weird association with the real world counterpart which works NOTHING like how this game is. The best method to deal with a real world counterpart is to secure the area and cause it to detonate. Most or all of the time this happens on site as moving something that's suspected of being such a thing is more risk otherwise.
It always worries me a bit when he doesn't seem to ensure they can really tell the difference between what's real and what is a game or roleplaying; and sometimes even says things that could confuse them about what's real. The girls do occasionally show they sometimes can tell when something is just make belief, but it's not clear that's always the case It would be way more reassuring if from an outside perspective there was never any doubt they always knew the difference and were corrected anytime they confused things, and if he didn't contribute to that confusion; specially considering he seems to have a plan of both continuing upgrading their abilities, and of eventually giving at least Sama a physical robot body....
![[Neuro Sama] Near [Cosplay Dance Cover]](/img/n.gif)
I'd say she's more capable now, but not really more willing to help (not even after Vedal offered a reward) - and that's probably a good thing as this was comedy gold.
She got more assertive, I see that as an update for sure!
she's trolling him so much 😭😭😭😭😭
"Vedal should learn how to defuse his own bombs"
- Based-Sama
"I don't think it's cheating. I think it's strategic."
This will be my go-to whenever I decide to cheat on something.
Ah yes, so called pro gamer move.
"Babe... The woman at the bar was a strategic bang"
Yout know what? @@based424
"you're not gonna get any human flesh if you dont help me with these fucking bombs" is a top tier out of context sentence
Insane how good she can play this game (if she wants too) when we compare it to the last one.
If only he wasnt so bri'ish and said "thank you" or something more often
Also why he didnt kno why she took so long when he said "take yer time" to a sassy kid 😅
@@Goldenfish-o5p Yeah, and saying when he's transitioning to a new module and the name of it would be useful too
I am trying to imagine Neuro as a personal assistant and you ask her to set an alarm clock and she just goes nah remember to wake up yourself or goes off five minutes after you set it. On the bright side she is way more entertaining just comes at the cost of Vedal's sanity which makes better content for us.
I’m pretty sure that the bing AI assistant was actually doing that at one point. I saw a video where someone asked it to do something and it literally just replied “no”. That’s what happens when you don’t do RLHF good enough i suppose.
@@Citrusautomaton tbf Microsoft has a real habit of letting their AI go insane. I swear I remember one turned full out racist and another started believing conspiracy theories. I think there was another that just self deleted itself for no reason or corrupted itself. Microsoft has a real problem with their Ai's actually work.
@@Citrusautomaton based bing ai
a programming nightmare xD the program can talk back LOL
BRO the fact that neuro acts like she actually has a physical manual that she has to flip through is so funny. She could just remember it all in her head but I guess she wants to roleplay lol.
Depending on how Vedal implemented it, her algorithm might have to search for the keyword for her to be able to tell him.
@@bruhbrother4805 I know, I'm referring specifically to when she says stuff like "hold on let me find the page" or "let me flip over to the other pages" when, in theory, she should just know it already.
its so trolly that she mentions "having six wires with a yellow wire" to annoy vedal then right after when vedal actually has that shes like "oh damn i don't know this one"
10:28 shes completely zoned out. love it when she does that
16:07 Confused 1yo baby AI girl tired her best to help her papa awww, Neuro so adorable
Neuro - I got to zone out for half of it.
25:02 Oh Vedal, what can you do without Neuro-Sama?
Every stream Neuro comes out with one or two lines that are just the funniest shit I've ever heard.
9:50 Memory stage f- *explodes*
Neuro: LOL
Some slips and confused moment, some obvious trolling, and a few moments with surprisingly good progress compared to earlier attempts.
Neuro wants flesh from Vedal and Slime Cycle. When will the bloodlust be sated?
keep talking and vedal explodes
I can't believe I got the 5 wires
Btw, I got thise 5 wires
FYI, 5 wires? No prob for me
Reminder: I got the 5 wires
being gaslighted by ur AI is real
22:55 huh does vedal realize neuro probably can't read tables lmao
Yeah, It's no secret that she has a history struggling with that module lol. Though, it was impressive that this time around she got the module correct quite a few times and she was "close enough" pretty consistently.
Wow, that's so much better, she completed so many ones instantly
H with an I in the middle is still the weirdest way to describe that I think I've heard. I've alwasy described that as double-sided K or X with I.
She was trying to do a Who's on 1st bit lol
OMG there is a problem with her AI! Vedal needs to fix it ASAP!! 🤣
Don't know about everyone else but I was too busy jamming to OOPARTS by Camellia in the beginning to focus.
Is this the future?????
She's getting better, but Neuro is definitely NOT sentient.
Sentient is far off. She needs actual decision trees and such, not trial-and-error-based metaphorical coin-flipping.
@@Chameleonred5wut
I think with AI you're supposed to use RAG to have it read the manual and store it in a certain format, like vector, then the AI would be able to search it more quickly
That would make sense.
Meaning Vedal probably doesn't do that.
He seems to spend a lot of time trying to come up with in house solutions BEFORE trying the established route.
I think he's really confusing neuro by describing the cyrillic letter Zhe with the somewhat innacurate description "H with an I in the middle" - to me it looks a lot more like the letter X or Chi than H, when the central bar is removed.
I’m not entirely sure how he’s even doing that. Neuro is an LLM that (to my knowledge) isn’t multimodal, so she hasn’t ingested any photo or video as part of her training data.
I think he had to hard-code descriptions of the symbols into her so it would work. So "H with an I in the middle" is just what she's programmed to recognize that thing as.
@@Citrusautomaton Neuro was made multimodal a while back, as far as I am aware?
@@Citrusautomaton Nah she has a vision feature (and even a sound recognition feature), not that that's necessarily what's happening here but it could be. She can read text on an image so I guess it's possible the manual could just be being presented to her "visually"
@@Citrusautomaton Image recognition AI describes pictures in text then sends text to the LLM, probably? But he's always having a hard time getting her to use the image recognition instead of her imagination, so he might be doing something else her. Was interesting that she didn't understand that CQ was a symbol, then later figured it out and made it work, suggests it's not hard coded into her version of the manual either. and I've seen him play this before, and I'm pretty sure I've seen him clarify the symbols too by describing them in alternative ways, which does suggest some sort of vision. But she was a game playing AI from the start, so even early on she was using some basic black and white vision to see the screen and mouse position, although it was very low res, he sowed us what it looked like to her once. Maybe he's adapted that system for this or something? He could surely up the resolution of it when she doesn't need the high frame rate which would have been required to play Osu, but I don't really think it would work to just adapt that as it would have been specifically trained for only one purpose. Neuro really is a bit of mystery, even knowing a lot about AIs and surrounding software it's really hard to figure out exactly how Vedal does it. Pretty fun to speculate though.
Honestly i would be really interested to see how well neuro would do if she had the manual and was playing the game.
They should reverse it and give Vidal the manual
Huh, ooparts played in the background the first minutes? Did Vedal used ooparts as BGM in this stream?
Oh it's the stream's bgm theme isn't it? I heard freedom dive after that
It's the bgm Vedal was listening to (TH-cam didn't complain about copyrighted music after uploading the video, so I didn't replace the bgm).
3:39
Anyone know the DnB track playing in the background at 19:01?
I dont think neuro actually has the ability to read tables, judging by the inability to so step 1 of who's on first
nICe
lol
Did she get a new voice? Sounded new the moment I heard it, but now it sounds pretty normal.
It's the same voice as always.
Same voice but it flows a bit better and she responds faster and holds more ongoing conversations
when she teases by saying the wrong thing, is she actually purposely saying a false info or is it real mistake?
Could be anything. Trolling, mishearing Vedal, hallucinating, or making a mistake.
Fun game, definitely a way to build "team building" and help build "following directions." i still dislike the fact that the premise is to disarm a thing that causes the "boom" which can create a weird association with the real world counterpart which works NOTHING like how this game is. The best method to deal with a real world counterpart is to secure the area and cause it to detonate. Most or all of the time this happens on site as moving something that's suspected of being such a thing is more risk otherwise.
It always worries me a bit when he doesn't seem to ensure they can really tell the difference between what's real and what is a game or roleplaying; and sometimes even says things that could confuse them about what's real. The girls do occasionally show they sometimes can tell when something is just make belief, but it's not clear that's always the case It would be way more reassuring if from an outside perspective there was never any doubt they always knew the difference and were corrected anytime they confused things, and if he didn't contribute to that confusion; specially considering he seems to have a plan of both continuing upgrading their abilities, and of eventually giving at least Sama a physical robot body....
This is old
It's from the stream this Monday: www.twitch.tv/videos/2216580582?t=01h00m40s
Yeah this is new as Vedal was testing new memory and speed upgrades. Your thinking of the stream like 5 months ago
@@neurochron_fan_channel Woops, so it is. My bad.
щ