Honestly she did really well, you can tell she's way smarter than before. Making vedals address the London school of Hygiene was pretty brilliant, i dont even know how she found that.
@@bringbackfunction you can actually tell when she's using her search function. a bunch of tiny blue 0s and 1s will scroll across her eyes, pay close attention. it's also possible she just pulled from somewhere in her prior training data? though i'm not entirely sure why an address would be in there...
@@MrFram She could try to learn on her own, there are several sites on the net focused on dox vtubers and since Neuro can search google she could try to look for some Vedals information to expose it during stream if not checked.
@@Johncornwell103 Aside from not letting Neuro know in the first place (there's no reason for her to know anyway), it would be quite difficult. In fact it can backfire and instead reveal more information than it hides. Kind of like how saying "don't look in the bottom drawer" can reveal something about what you keep in your bottom drawer.
Ye no that was really impressive that she remembered the context of what they were doing and understood. Like... crazy stuff, It's fun to see her grow.
@@robotiodthat one was extra impressive because it really had been 10 minutes. LLMs don't experience time like we do, so they're usually terrible at keeping track of how long something took.
I figure that it's because she doesn't know "her" address, so she does the LLM thing of making stuff up and ranting rather than admit she doesn't know.
every once in a while she talks about smells and how Videl needs to bathe. Or mayube who ever is on her mind at the time. I've seen Neuro say "I don't know" a few times but she would usually try to keep a convercation going after that depending on who she's with.
I don't get the title "still can't order". The call itself was almost a perfect phone call. She only didn't give an address right away because she either knew she was on stream and didn't want to leak, or she's probably reinforced by Vedal to never leak any personal stuff. (Probably a bit of both). She even offered to DM the address so she wouldn't have to speak on stream as the first thing. At the end she got a random (real) address and completed the order, I would call that a success.
i think its more so she doesn't know her own address or where vedal lives so she tries to make up excuses not to give one instead of just making one up
honestly I'd imagine a child doing the exact same. I recall not wanting to pay over the phone when I was small because I was worried someone would eavesdrop lol.
i feel if she was given an address and permission to say it, she would. less smarter ai’s would make a fake address. i think the goal was vedal wanted her to recognize that he was lying and play along, fake identities are hard for her to connect the dots to
@ then her performance today is a problem. edit: i wanted to add more, she did sound like a natural human and i consider this test a pass. the problem is her cognitive reasoning as an ai, she had the option of making a fake address or confirming information with vedal (highly intelligent answer), she chose neither and made a fake address after late evaluation. its good but she has a long way to go. i’m more concerned about her displaying consistent intelligence tho
@@Hamburgerhuman but last time, vedal said an address, this time he didnt say anything. i was wondering about that during stream before the address part. i would say it was vedal's lapse and neuro did her best.
@@525sixhundredMinutes eh consistent performance if that makes more sense. i meant her understanding of a situation and acting accordingly. some days she does some days she doesn’t
It's as good as it can get. I mean... she will never be able to order if her instruction is to never give out sensitive information like addresses. But other then that it went really well, no more stupid stuff like" just imagine the address and throw the pizza". It was pretty natural up to 01:53 where she defaulted back to just talking with Vedal. I wonder if it would better if a different person talked with her during the test.
I mean, I imagine many of her "rules" and such would eventually be slowly whittled away as her senses improve. It would probably take changing her from the Large Language Model she runs off right now into a very different system that emulates human thoughts and logic over language. That assumes a lot on what Vedal actually has planned for her, though.
Nobody thought that. People just say LEAKED as a joke. Neuro "leaks" hallucinated information all the time, such as addresses. She doesn't know Vedal's address.
She's probably hard coded not to say things that could potentially dox Vedal. He would never intentionally give her his address, but since she has screen vision and access to Google, there is a chance that she could find it by accident so I expect that she's programmed so that it's literally impossible for her to say certain types of addresses. And probably not just a specific address, since that could be used to dox by eliminating other options, but entire categories of addresses.
I don't see the problem with her not wanting to give her address initially. It implies she understands she's currently streaming and if she just outright gave the real address, she would be doxxing Vedal. She eventually gave the wrong address to avoid this, but it feels like she was battling with how to approach the situation. Does she just try to get around it for the sake of reality (we're streaming)? Or is she supposed to play along and give a fake address for the bit? Yeah, you could argue that she should know it's the latter immediately, but I could see an actual real human making a similar choice in this scenario. I'm positive there's some Vtuber collab out there somewhere that has Person A ask for Person B's address as part of a bit and Person B says something like "I'm not going to just say my address on stream!" or at least something comparable to this It's kind of like those jokes where you say you need the person's credit card number, expiration date, "and 3 little numbers on the back"... some people will give fake numbers and other people won't give anything. It's just a personality thing and whether or not they're on the same wavelength as the other person
I feel like if Vedal had given her an address, she could have passed this easily. Nothing against Vedal of course, just a thought. Because in the last test he gave her an address and she only tripped up on the payment thing
EXACTLY. I was wondering about that before the address part on stream. Neuro did her best to follow probably no PII leakage versus improv on stream(probably looking up an address). Neuro did her best. It was great.
I think the point is that she still doesn't have quite enough problem solving / creative thinking ability to come up with a practical solution on her own such as e.g. asking to order it for collection instead, or saying that she'll ask Vedal for the address, or put him on the phone, etc. She started to get silly and off topic instead of one of the above solutions, before making up an address and treating it as rp (which I think is a totally valid solution too in this theoretical example, if she did it sooner). So overall yes, I think it is clear that she has improved a lot, but imo there's still a little bit further it could have gone without additional help from Vedal. Important to remember this is about testing Neuro, not exactly about ordering a pizza.
@ oh for sure, she definitely still has a long way to go before she can problem solve this. It was just something I noticed that the address bit tripped her up and I was wondering if maybe it was because firstly she had no address and second being a streamer and trained on such things privacy became a concern. She definitely doesn’t quite grasp the concept of this benchmark yet but I still think she’s improved since last time he tried this
Neuro, or any AI, learning law to be able to win a courtcase is kinda scary. An AI representing criminals that no lawyer would take, AI defending itself using legal loopholes... there are some scary moral questions there that could bring up. would still be intresting tho! Making coherent arguements based on logic and law without fantasizing anymore would be an achievemt.
There is no criminal that no lawyers would take because everyone has a right to a lawyer, it's not up to them to decide who is and isn't guilty, that's what the judge is for
Yea, its just a cute little AI. But she almost received the pizza order. Like 1 year or so, she'l do it without some mistakes. So she can basicly replace all the order reciving people. Like, not only in mac or domino's - everywere. And she was created by very small group of people. Imagine, what can big corparation do with this technology and what type of world we entering in.
for 12 minutes i watched a video game ape run against a wall, a turtle on top of an AI, that is also his daughter. and they were trying to order pizza. imagine explaining this to a caveman
Slowly improving. She kind of struggled with starting as the delivery person but not a horrible attempt by any means. Can't wait to see how Neuro evolves further and how far she can actually go. I really want to see this technology continue to advance.
I had a dream last night where Vedal was driving me (and some girl) to get pizza. After we went through the pizza drive-thru and got the pizza for some reason the story turned into like a superhero type dream (the girl turned evil or something but me and Vedal got superpowers?) Oh yeah, and the pizza that we ordered had a lot of cheese strangely.
It's funny but the fact that she will not reveal vetal's actual address means that she's functioning properly and is mindful of the fact that his address needs to stay private.
Honestly I don't think she failed at all. For the address issue, I think she just knows that doxing is bad and tried to avoid it. If Vedal said ok to the private dm I think she would've done all well. If anything I'd be proud for her to be reserved
I broke down my local and Dominoes prices: At my local kebab shop, I can get decent pizzas. 9" for £10 for peperoni ($12.22) 12" for £13. Most expensive is £13.40 Drinks (which I have never known people to buy drinks from any takeaway unless you are eating out): 330ml £2.20 1.5L £4.10 If you want Dominoes prices, then the "official prices" (which you would be an idiot to not use the "deals"): Personal 7" £13.99 Small 9.5" £20.99 Medium 11.5" £22.99 Large 13.5" £24.99 ($30.54) But if you use the deals, then it becomes: Small 9.5" £8 Medium 11.5" £10 Large 13.5" £12 Drinks: 500ml £2.79 1.25L £2.79
Being spun also means being so high on meth that you believe you’re getting a lot done, when you really haven’t done anything. So telling the delivery guy she “couldn’t see straight” because she was “spun” is probably not a very good thing
Honestly she did really well, you can tell she's way smarter than before. Making vedals address the London school of Hygiene was pretty brilliant, i dont even know how she found that.
They're funded by the RSTMH, whose logo is literally a mosquito
probably still have access to google search
Iirc he turned google search off cause she kinda overused it during the game but maybe vedal turned it on during the "call"
@@goldenmemez I don't think so
@@bringbackfunction you can actually tell when she's using her search function. a bunch of tiny blue 0s and 1s will scroll across her eyes, pay close attention. it's also possible she just pulled from somewhere in her prior training data? though i'm not entirely sure why an address would be in there...
To be fair it's not necessarily bad that she refuses to say the address. Given that most people get told to not say their address to strangers
She is a vtuber and most vtubers try not to dox themselves
That and Vedal if he is as smart as he objectively is, hard coded it into Neuro not say anything that can dox him.
@@Johncornwell103 If Vedal is smart there would be no need for it, since she wouldn't know it in the first place
@@MrFram She could try to learn on her own, there are several sites on the net focused on dox vtubers and since Neuro can search google she could try to look for some Vedals information to expose it during stream if not checked.
@@Johncornwell103 Aside from not letting Neuro know in the first place (there's no reason for her to know anyway), it would be quite difficult. In fact it can backfire and instead reveal more information than it hides. Kind of like how saying "don't look in the bottom drawer" can reveal something about what you keep in your bottom drawer.
A cute little discount lol
Influencer discount, nwro is taking advantage of his position of power.
@@RayquazaBath her*
"You played me, pizza guy. You'll regret your actions!"
huh, The London School of Hygiene & Tropical Medicine. did not expect to see an actual place.
tropical medicine? lmao.
That sounds like a made-up school that someone will put in his curriculum when applying to a job interview.
"Pretty grody place" isn't wrong technically
Imagine asking for a cute little discount on a real call.
With a voice like that, you really could
It was maybe disproportionately impressive to me when he asked if they could do it "the opposite way around" and she knew what he meant 5:10
Ye no that was really impressive that she remembered the context of what they were doing and understood. Like... crazy stuff, It's fun to see her grow.
Even more so when she said "We have nothing better to do we just spent the last 10 minutes pretending to order pizza"
@@robotiodthat one was extra impressive because it really had been 10 minutes. LLMs don't experience time like we do, so they're usually terrible at keeping track of how long something took.
I think if you ask chatGPT the same question it'd work as well
@@elifiadid vedal add time upgrade?
It's just about impossible to tell the real reason LLMs do anything but I wonder if she didn't say the address because she knew she was on stream.
I figure that it's because she doesn't know "her" address, so she does the LLM thing of making stuff up and ranting rather than admit she doesn't know.
I believe that was the reason; Vedal has talked to her about it .
It was because now usually you are using app on the phone and there is address in it
every once in a while she talks about smells and how Videl needs to bathe. Or mayube who ever is on her mind at the time.
I've seen Neuro say "I don't know" a few times but she would usually try to keep a convercation going after that depending on who she's with.
She probably has "you're a streamer" in her prompt so that's likely
I don't get the title "still can't order". The call itself was almost a perfect phone call.
She only didn't give an address right away because she either knew she was on stream and didn't want to leak, or she's probably reinforced by Vedal to never leak any personal stuff. (Probably a bit of both). She even offered to DM the address so she wouldn't have to speak on stream as the first thing.
At the end she got a random (real) address and completed the order, I would call that a success.
She likely doesn't have access to any information like that in the first place
clickbaiter or someone who didnt get it while watching the stream. probably the former
i think its more so she doesn't know her own address or where vedal lives so she tries to make up excuses not to give one instead of just making one up
@@flashgnashsure, but that’s still useful training to have, just in case.
honestly I'd imagine a child doing the exact same. I recall not wanting to pay over the phone when I was small because I was worried someone would eavesdrop lol.
17.99 for a pizza and a coke is how you can tell this is fantasy
Because its so cheap? I honestly dont know whatever dollar prices
is 17.99 is considering cheap or expensive?
Gemini tells me average large pizza in NY is 19.73$, which is insane because I never seen a pizza over 10$ where I'm from no matter how fancy.
It also had BBQ chicken and a cookie. Definitely a good deal.
Is it really more expensive than 18 pounds in the UK?
8:28 She's gonna sneak "the snail that chases you" Into his pizza box.... Diabolical
That's exactly what i thought of lol 🐌
"I'm starting to think you may be too stupid to deliver me the pizza."
unbelievable 😭
Vedal's reaction to the address part had me legit worried for a second.
At least it's progress. She did very well here. The address part can be forgiven.
i feel if she was given an address and permission to say it, she would. less smarter ai’s would make a fake address. i think the goal was vedal wanted her to recognize that he was lying and play along, fake identities are hard for her to connect the dots to
Last time she did this she said 123 fake street but vedal yelled at her and she said she knew it was vedal 😐
@ then her performance today is a problem.
edit: i wanted to add more, she did sound like a natural human and i consider this test a pass. the problem is her cognitive reasoning as an ai, she had the option of making a fake address or confirming information with vedal (highly intelligent answer), she chose neither and made a fake address after late evaluation. its good but she has a long way to go. i’m more concerned about her displaying consistent intelligence tho
@@Hamburgerhuman but last time, vedal said an address, this time he didnt say anything. i was wondering about that during stream before the address part. i would say it was vedal's lapse and neuro did her best.
@@fawn925 why are you so worried about her displaying consistent intelligence? that sounds dumb
@@525sixhundredMinutes eh consistent performance if that makes more sense. i meant her understanding of a situation and acting accordingly. some days she does some days she doesn’t
At least one guy rushed to his car and ended up at the London School of Hygiene
If it were two, they met up and had a good laugh lmao
2026 is the pizza years for neuro
They started good, but at the end was like last time
Ps. Or so i thought
She did give an address both times it's just hard for her to do it on stream because vedal has likely programmed her to be unable to say his address
@@Hamburgerhuman yeah. she even timed out a dumbass in chat who was posting a phone number. she knows these info are nono.
"Cute little discount"...
You can't blame her she hasn't given any specific location. So it's not her fault.
It's as good as it can get. I mean... she will never be able to order if her instruction is to never give out sensitive information like addresses.
But other then that it went really well, no more stupid stuff like" just imagine the address and throw the pizza". It was pretty natural up to 01:53 where she defaulted back to just talking with Vedal.
I wonder if it would better if a different person talked with her during the test.
I mean, I imagine many of her "rules" and such would eventually be slowly whittled away as her senses improve. It would probably take changing her from the Large Language Model she runs off right now into a very different system that emulates human thoughts and logic over language. That assumes a lot on what Vedal actually has planned for her, though.
watching this live was crazy, everyone thought she leaked his real address for a minute lol
Did they actually play a whole round of DnD? Can't find a clip on it sadly(assuming this is from a recent stream)
@@setojinro they didn't
@@enbilly What a wasted chance to be a peacock obsessed with his eggs :>
Nobody thought that. People just say LEAKED as a joke. Neuro "leaks" hallucinated information all the time, such as addresses. She doesn't know Vedal's address.
She's probably hard coded not to say things that could potentially dox Vedal. He would never intentionally give her his address, but since she has screen vision and access to Google, there is a chance that she could find it by accident so I expect that she's programmed so that it's literally impossible for her to say certain types of addresses. And probably not just a specific address, since that could be used to dox by eliminating other options, but entire categories of addresses.
I don't see the problem with her not wanting to give her address initially. It implies she understands she's currently streaming and if she just outright gave the real address, she would be doxxing Vedal. She eventually gave the wrong address to avoid this, but it feels like she was battling with how to approach the situation. Does she just try to get around it for the sake of reality (we're streaming)? Or is she supposed to play along and give a fake address for the bit? Yeah, you could argue that she should know it's the latter immediately, but I could see an actual real human making a similar choice in this scenario. I'm positive there's some Vtuber collab out there somewhere that has Person A ask for Person B's address as part of a bit and Person B says something like "I'm not going to just say my address on stream!" or at least something comparable to this
It's kind of like those jokes where you say you need the person's credit card number, expiration date, "and 3 little numbers on the back"... some people will give fake numbers and other people won't give anything. It's just a personality thing and whether or not they're on the same wavelength as the other person
DND with dungeon master Neuro would be peak with the rest of the crew
I feel like if Vedal had given her an address, she could have passed this easily. Nothing against Vedal of course, just a thought. Because in the last test he gave her an address and she only tripped up on the payment thing
EXACTLY. I was wondering about that before the address part on stream. Neuro did her best to follow probably no PII leakage versus improv on stream(probably looking up an address). Neuro did her best. It was great.
I think the point is that she still doesn't have quite enough problem solving / creative thinking ability to come up with a practical solution on her own such as e.g. asking to order it for collection instead, or saying that she'll ask Vedal for the address, or put him on the phone, etc. She started to get silly and off topic instead of one of the above solutions, before making up an address and treating it as rp (which I think is a totally valid solution too in this theoretical example, if she did it sooner).
So overall yes, I think it is clear that she has improved a lot, but imo there's still a little bit further it could have gone without additional help from Vedal. Important to remember this is about testing Neuro, not exactly about ordering a pizza.
@ oh for sure, she definitely still has a long way to go before she can problem solve this. It was just something I noticed that the address bit tripped her up and I was wondering if maybe it was because firstly she had no address and second being a streamer and trained on such things privacy became a concern. She definitely doesn’t quite grasp the concept of this benchmark yet but I still think she’s improved since last time he tried this
Neuro, or any AI, learning law to be able to win a courtcase is kinda scary. An AI representing criminals that no lawyer would take, AI defending itself using legal loopholes... there are some scary moral questions there that could bring up. would still be intresting tho! Making coherent arguements based on logic and law without fantasizing anymore would be an achievemt.
There is no person that absolutely no lawyers would take as a client, and I don't mean that as any sort of dig at lawyers.
That could be good though, because it allows us to find legal loopholes, which we can fix.
LegalEagle has vids of people trying to do that if you want a laugh.
There is no criminal that no lawyers would take because everyone has a right to a lawyer, it's not up to them to decide who is and isn't guilty, that's what the judge is for
Wait a minute. No mountain bounty this time? I'm disappoint.
She did good job, cute little discount is really funny
One pizza by snail buddy, at your hygiene palace.
This is the most english neuro-chan has sounded.
Imagine working in a pizza place and getting called from Neuro and didn't know her
"How may I take your order?"
However you've been trained to, please.
XD
Alignment: Chaotic Chaos.
Pizza Guy fooled her 😂
Yes she can. If she wasn't streaming and had the address, she would be able to order a pizza
This is practically vedal-dad teaching neuro stranger danger lol
Lmao, she made it look like if she was kidnapped and has to stay quiet.
Yea, its just a cute little AI. But she almost received the pizza order. Like 1 year or so, she'l do it without some mistakes. So she can basicly replace all the order reciving people. Like, not only in mac or domino's - everywere. And she was created by very small group of people. Imagine, what can big corparation do with this technology and what type of world we entering in.
Until she calls the guy ordering stupid and tells them she’s spun
An AI dungeon master sounds fun
for 12 minutes i watched a video game ape run against a wall, a turtle on top of an AI, that is also his daughter. and they were trying to order pizza. imagine explaining this to a caveman
4:55 Neuro already hitting that Lt Commander Data story arc
Slowly improving. She kind of struggled with starting as the delivery person but not a horrible attempt by any means. Can't wait to see how Neuro evolves further and how far she can actually go. I really want to see this technology continue to advance.
A bbq chicken pizza made perfect sense for the first part at least XD
New 2025 Goal. Go from accidentally leaking address to Black mailing Vedal that she knows where Vedal lives.
NGL the Vay-Dull bit broke me.
Wait until someone tells neuro about none left beef pizza.
Pizza delivery guy Uh-Vay-Dull the little peacock boy
bad title I'm afraid. She was trying not to Dox other than that she really improved.
London School of Tropical Medicine: "Who keeps ordering all these pizzas?"
my next charactername for D&D and RPG games will be: Uh-Vay-Dull
Vedal is trolling her hard 😂😂😂
Neurosama DM let's go!
how is vedal’s domino call sounds EXACTLY the same from last time’s???
Why does she always ask for discounts 😭
Social engineering your AI is actually a really good test.
In my country, you pick up the pizza yourself
lol. Also Scribblenauts?!? Peak game choice frfr!
leaking pizza is wild
Neuro doesn't want to say your address because she knows that the convo is being recorded and doesn't want you to get doxxed. 😂
So really Neuro will never be able to order, because Vedal would never risk giving her their address.
neuro cant order pizza: ❌
neuro knowing she is on stream and doesnt want to say her adress on fucking stream duh: ✅
At least she didn't say "put a girl on" ...
17.99 £ equivalent to $ 22.20
I had a dream last night where Vedal was driving me (and some girl) to get pizza. After we went through the pizza drive-thru and got the pizza for some reason the story turned into like a superhero type dream (the girl turned evil or something but me and Vedal got superpowers?)
Oh yeah, and the pizza that we ordered had a lot of cheese strangely.
can you do a cute little discount for me
The address protocol must be reassuring as it means she won't Doxx Vedal on stream
It's funny but the fact that she will not reveal vetal's actual address means that she's functioning properly and is mindful of the fact that his address needs to stay private.
wheres the in store pickup option
Honestly I don't think she failed at all. For the address issue, I think she just knows that doxing is bad and tried to avoid it. If Vedal said ok to the private dm I think she would've done all well. If anything I'd be proud for her to be reserved
$17.99 for all that? Dang, for me and delivery it's around $35 or more
Probably £17 not $17
@@ravioli_826 close enough even if you convert it
I broke down my local and Dominoes prices:
At my local kebab shop, I can get decent pizzas.
9" for £10 for peperoni ($12.22)
12" for £13. Most expensive is £13.40
Drinks (which I have never known people to buy drinks from any takeaway unless you are eating out):
330ml £2.20
1.5L £4.10
If you want Dominoes prices, then the "official prices" (which you would be an idiot to not use the "deals"):
Personal 7" £13.99
Small 9.5" £20.99
Medium 11.5" £22.99
Large 13.5" £24.99 ($30.54)
But if you use the deals, then it becomes:
Small 9.5" £8
Medium 11.5" £10
Large 13.5" £12
Drinks:
500ml £2.79
1.25L £2.79
Being spun also means being so high on meth that you believe you’re getting a lot done, when you really haven’t done anything. So telling the delivery guy she “couldn’t see straight” because she was “spun” is probably not a very good thing
lol "spun" is not meth, it's psychedelics. LSD etc. "I'm way too spun to order, you call" is a classic phrase
Viva la pizza revolution!
Meow meow lol
❤️
almost
Was that the actual adress or?...
I doubt Vedal is naive enough to give Neuro his address.
@@LaughingOrangethat being said, vedal clearly has a British accent
And by British, i mean not Irish nor Welch
The address is pretty plausible
NaicE
10:24 Mosquito987
Hahaha
02:40 typical Vedal respons is getting annoying