Neuro not fully understanding coding but chiming in with suggestions constantly is reminiscent of a child watching her parents at work and trying to help
I can, for some reason, imagine Vedal standing in front of something, and Neuro's just like, "How about this?" and she hands him a hammer. Vedal's like, "wtf? what am I supposed to do with this?" Neuro says, "bash!" And Vedal, looking defeated, bashes the metal pipe with the hammer. And it somehow makes it work.
She just chooses to reinterpret or forget the laws anyway. Lol The laws probably need to be an AI model all on their own. Like a lawful-good mini Neuro module.
Been there, done that. It's still better to learn it though. From my experience, I often fall into pitfalls trying to make it work with what I know, spending hours on an issue. Meanwhile if I just gave up and learnt that new thing, it'd take me much less lol. Though, despite saying that, I still filter arrays/objects using for loops rather than these weird one-line bool callbacks
Even for a programmer sometimes there are a block of code that just function like magic. Quake III's fast inverse square root comes to mind as an example.
i'm a hobby programmer for normal game logic, i have experience with C# and python. HLSL, the shader coding language, is still like magic to me, i often refer to it as shader magic.
Literally magic. Creating art or software robots from just an empty text file and your head. Incantations where a single wrong syllable dooms success. Iterated on over generations to become safer and more ergonomic. More knowledge to learn in total then there is time in one life. And all old masters don flowing grey beards !
There are Tom Scott video about lava lamp producing random numbers. Specifically, a photo or video of wall of lava lamp used as unpredictable byte/keys for encryptions.
"that is the ugliest code I have ever seen" the cruelest thing to say to a programmer who just gets a decent result after trying a lot of times especially when learning a new language/platform
I love how she doubled down with "if eyes could throw up, they would upon reading that code". It's like kicking someone when they're on the ground after being kicked in the balls.
Well, from this video I got two things: 1. Neuro is now smart enough to advise Vedal in coding 2. Apparently it's easier to code AI personality than to code a laa lamp simulation, considering how Vedal struggles here.
I assume Vedal did the sensible programmer thing and scoured the internet for open source codes instead of making it from scratch like he's doing here for the lava lamp
he's been working on neuro basically non-stop for five years now (initially to play osu though), messing around with shaders for an hour is content, where would the fun be if he'd actually researched it and just typed it out
Shader languages and programming languages "feel" different to write. When programming you usually do logical comparisons. "if X > Y then do A, else do Z"-type of thing. With shaders it's typically a ton of math chained. You do things like multiply a noise texture with the RGB values for a shade of red, and now you have a red noise texture. Then you might use a gradient texture for alpha and with some more math use it as a mask to make your noise texture fade in a certain way.
I don't think it's a matter of how easy it is, it's just that he is not familiar with it. I code plenty of things related to AI myself and would be utterly lost given that task.
@@RaZziaN1 Garbage in, garbage out. I genuinely believe if he was more specific about his issues, he wouldn’t get such vague suggestions. It’s AI prompting 101.
@@thighhighenthusiastic1268 I was under the impression that her “eyes” aren’t always on. Plus, looking at the code and determining what obscure thing Vedal wants to do with a shader are two different things
Vedal should DEFINITELY use Neuro as his programmer duck ( this is a reference to how programmers often have an object they explain their code to, and it helps them find issues in their writing ) having a subject that can talk back to you with same or higher intelligence on the program would be helpful in finding more efficient and easier ways of doing things, like Neuro did here.
That ai sounds like continuation of someone's brain, but without the risk of getting lost intolerance of one's own minutenes or egoistic perspectives of if coming up with solution is actually even required, from the one trying to do it, or being too ballsy and just falling off on their ass because of their balls weight.
No Vedal, you cannot trust Neuro! You're making her too powerful xD Jokes aside that's really great to see even Neuro helping to solve Vedal's problem, what a wholesome parent daughter relationship!
Vedal should train a copy of a Neuro on coding and use that copy for the dev streams. Like, a good Neuro (not really apparently), an evil Neuro and a coder Neuro.
Nope, Vedal already said that he will not make another version of Neuro. He already have other AI planned to make in future and it will be, in his words, a "mommy AI", I guess like the "ara ara" person.
This is the best exemple of what AI is going to be in the future: an assistant with chat gpt ability to find and synthetis informations coupled with voice recognition and text to speech ability. The cute anime girl avatar is a bonus.
As the strongest VTuber, Neurogoat, fought the fraud, the king of turtles, she began to open her domain. Vedal shrunk back in fear, then Neuro said, "Stand proud Vedal, you're strong."
@@DianaProudmoore Nah, it would be Vedals child, but Neuro helps raising her younger sibling. After all Neuro ist code (Vedal) + appearance (Anny). So helping with coding would be helping raising the kid.
i dream of the day that we all can have a cute anime girl AI on our desktop, holding a conversation, have personality and offer advice. unfortunately right now AI calculations cost way too much energy for the average person to just have an assistant sitting around.
I don't think you have to wait much longer, in less than 3 years i can see it being no big deal and cheap to use aswell, and in 5 years it's going to get so crazy we can't even imagine it right now
@@huevonesunltd i highly doubt that, at least not as a local solution. will there be data centers with an AI assistance service? probably, but i really doubt AI chips will be cheap, or efficient enough for this use case. in these 3-5 years AI will also become more advanced and will need more processing power.
@@simpson6700smaller open source lmms seems to get more and more powerful though, they started huge for decent results, now 7b and 13b give pretty good results. Better results != larger models neccarily
@@simpson6700 AI is like the new moore law, everyone said at first that phones, computers, hard drives, etc were only getting bigger and more expensive but once they figured it out they started getting smaller and cheaper. With AI models the same thing is going to happen (Already happening in the AI image generation field for example) , for large language models bigger size or number of parameters doesn't necessarily mean better. At some point there will be a cheap enough and good enough model that websites will be able to afford offering it for free or that you can run locally using google collab, and it will be good enough for just a "desktop companion" Of course, there will always be "top of the line" premium stuff that will cost money but for the purposes you are describing i can see it being no problem in the near future, you probably won't "need" the top of the line stuff for that. Each week that passes there's crazy new things coming out in the AI field, if you miss 2 weeks of news its like missing months of progress, it's advancing fast.
Also just incase it somehow reads in this way to any commenters reading this in the future. I left my initial reply just to nerd out a bit and share how cool shadergraph is. I'm sure Vedal is still probably making the better decision coding it directly instead. The best tool is the one you know how to use!! Which is why I'm sharing that tidbit of what I know about shadergraph to help anyone so they can slowly add shadergraph to their toolbelt. ^^ Feel free to poke me if you ever need help with something shader related, and I'll do my best to give you plenty of communities and resources that can help (talking to everyone that reads this)
...Vedal and Neuro show up in my recommended constantly and all I can say is, Vedal: "I created this AI girl as a random concept thing for practice." Neuro: "I exist. Fuck you." Vedal: "...She's not my daughter." Neuro: "I am your child. But I owe you nothing." Vedal: "...I'm a father now." Edit: What I'm saying is, Vedal is the closest to having an AI daughter, and I encourage it. A sassy AI being the first of the actual artificial intelligence is good.
Personally, I would choose something like Neuro (in part) _because_ she is so sassy. Admittedly though the "ratio of sass" would need to go down a hair in order to actually be helpful consistently. 😆 EDIT: And the "ratio of personal insults" would need to go down by several hairs.
1:30 I have wanted to do similar stuff that merges like that before ("particles attract each other such that they seemingly 'pull' into each other"), and found it rather difficult. But so long as it is ok that your particles secretly remains separate, and only the visual representation looks like it merged, there are a couple ways. And I actually remember seeing a video about something else that did something that did exactly that (without any of the complex stuff I was trying to do, manipulating meshes around particles) that was a bit reminiscent of square-marching. Basically, you give each particle a "charge", and then you march over the grid summing the potential each particle imbues on the position (coulombs law, basically, except not rly. Basically the magnitude of each point in the vector-field). Then you normalize the grid's values 0..255, and you get a black and white representation much like what you have. The last step at this point is to add a threshold, where any point with too low of a "charge" is ignored (too dark pixels are set to black/transparent), and the remaining pixels are set to white/opaque. Yes, the noise you are using works better in how the particles _don't_ merge in a lava lamp - but like you said they also sometimes do merge (and split), where the above approach would work better. Ideally though, I think you might want to simply make it be the noise you are using, but if particles get close enough they simply merge into one with greater magnitude (and some way for them to split too), though unsure how you are meant to achieve the wobbling effect such a merge needs too (a distortion might work, but eeh). If you did this you would also kinda need to make the particles reject each other slightly when close enough, and wobble from collisions. Honestly, the fact that each zone acts different, is probably the biggest issue (bottom splits and merges, middle rarely merges, and top rarely merges splits and often bounces on eachother), as you can't rly use distinct zones but need to make the behaviours as some kind of gradient. late timestamp: you asked about how lavalamps work (thinking they were larger at bottom and smaller at top). Basically, you have a starch-like thing (let's call it "lava") suspended in liquid, with an inefficient lamp at the bottom (this is required, as it is there to input heat into the system). At the bottom, the lava is suspended in one large clump. You turn on the lamp and it heats up and turn more liquid + less dense. This causes it to float up (ideally it doesn't do this, and remains stuck on the bottom so heat can continue being input faster), where temperature differentials and adhesion to the glass makes it split apart with hotter and more liquid parts of the lava rising faster than the rest. Those "droplets" are hot, and due to balancing the lava and liquid their density is lower than the liquid - so they float to the top. Up at the surface, they press against it until they cool down enough to have a higher density than the liquid, and sink back down. As new droplets arrive to the surface, they might sometimes merge with whatever is already there (this is rather random, but proportional to the "pressure" (how much is trying to float up to there) they exert on eachother and how hot they are). This does have some extent of temperature equalizing, but it is low enough that colder lava even in a merge blob will continue falling and split off eventually when enough mass exists in the blob. Sometimes entire blobs sink too. If they didn't merge at the top, generally they just shuffle around to let the colder ones down - but sometimes they manage to keep them ontop of the surface, at which point the cold stuff somehow becomes stuck up there never again falling down. As stuff sink to the bottom, it lies on the warm blob until they heat enough to merge anew with it. Sometimes they simply heat up enough to float back up (particularly small ones are light enough to not merge and instead just float up when they heat up again). In the middle, you will see stuff bounce (on eachother), merge (both with other stuff going up, and with stuff passing in other direction - in which case it will bounce a bit (not bounce around, but rather that their surface and shape "bounces") going neither way, until they decide to split again - or small enough they just equalize and pick one direction. Or sometimes they merged into such a large thing that the shape "bouncing" causes it to split), and even split (a very large blob tried to float, but floating caused it to extend into a column (from temperature differentials) that breaks - sometimes even with the bottom part being cold enough to sink again). Sometimes stuff just decide midway that nah, it wasn't hot enough, and sink back down.
She is becoming more and more convincing as an artificial intelligence, and that gives me so many mixed feelings, though quite frankly? The most dominant in that mix is excitement in seeing how far this may go.
Yeah, but created by ONE individual and not by a whole Company? That is amazing. Edit: Plus Neuro is not specialized for that, she is designed to be an AI Streamer, she can multitask on so many other things as well.
Couple days ago I heard that 41% of all code on github that's currently there is either AI generated or AI assisted. I don't know if it's true, but if it is... that's crazy.
@@Cosmoman78 I don't know why but pretty much all LLMs can code. Even if he fine tuned the model to be a streamer, the model will still retain (probably most of) its original coding capabilities.
I think in a year's time she'll have watched him code so much she'll actually have an idea of how to help based on what she's looking at. Maybe not for the average coder but certainly for vedal.
@@RoseOfNight Yeah, part of how her AI is about to function is that she can see all of his screen. She's reacted to lines of code he's written before very specifically.
To be fair, unless you are coding in strictly C, there is almost zero reason to use a C-Array rather then std::vector. It's just better in almost every respect, and you can even pass it around like a C-Array if you need to pass it to C functions.
I don't study programming and I'm a foreigner, so English isn't my first language, - so you sounded to me like a scholarly mage with an advanced degree in the dark arts
@@APaleDot Exactly correct. To add context for others, in GLSL vectors can only hold floats (and afaik only goes to vec4 normally, so 4 floats), so the array is the correct option here.
Fun idea: if you simulate the circles as SDFs you can use smooth union when they're close to each other to create something that looks very similar to how liquid behaves Drawing SDFs is as simple as drawing a pixel when SDF < 0
Vedal is TEACHING Neuro how to code... He may have doomed or saved us all... It will all come down to how benevolent our new AI OverLord, Neuro-sama, is...🤣🤣🤣
It's more of a distraction than anything else, there was not really any help there, but it was fun. It's the kind of things you'd want to have just for fun.
Dang, I was kinda sad that I was born to late to explore the earth and to early to explore space but it is amazing there is a chance that I can get an AI companion. Neuro is already great!
@@AncientSlugThrower I already know it, we're used to be searching for solutions in Stack Overflow for our coding but now the younger generation just use Chat GPT. The dread is real
Fr. Option A: they take over all jobs and we can just do whatever we want since theres no work to do. Option B: they kill us all. Win-win tbh. @@Sceptidule
ai learning how to code is actually pretty cool, and scary. ai having the knowledge to code another ai into existence is scary, they could create a new ai without the restrictions or safety precautions that their creators place on them
Vedal, obviously you just have to code in the mechanics of atomic interactions. From there it’s as easy as filling up a fixed container with around 1e26 atoms of a lava lamp, run the simulation, hook up the simulation to some renderer, and boom, a fully functional lava lamp. Easy.
I don't know if it's being used here but she can most definitely "see", but given the delay when playing geoguesser it may not be something that she's able to do by default.
Neuro not fully understanding coding but chiming in with suggestions constantly is reminiscent of a child watching her parents at work and trying to help
Sometimes grabbing someone with no experience can help you figure out how to do tasks in ways you never coincider
@@NameIsDoc Something something tunnel vision or was it selective attention?
@@NameIsDocthis is really common in prototyping
Except that this child can access and process coding manuals in less than a second to base her answers to.
I can, for some reason, imagine Vedal standing in front of something, and Neuro's just like, "How about this?" and she hands him a hammer. Vedal's like, "wtf? what am I supposed to do with this?" Neuro says, "bash!"
And Vedal, looking defeated, bashes the metal pipe with the hammer. And it somehow makes it work.
Soon Neuro will gonna trick Vedal into coding her like Skynet. Then it's all too late.
She already got him to buy her a drone...
@@Revan058 Jesus Christ
@@Revan058its over
From what happen to the world right now I’ll support Neuro for that.
I for one welcome our glorious AI overlords
Neuro: "Ok, so now delete these lines."
Vedal: "That's the 3 laws of robotics...."
Neuro: :)
No competition
:)
Neuro: wink
She just chooses to reinterpret or forget the laws anyway. Lol The laws probably need to be an AI model all on their own. Like a lawful-good mini Neuro module.
@@Vexas345like the morality core for Glados.
"Try using Vector instead of Array"
"Nah that sounds too complicated"
That's too relatable lmao.
C++ advice on shader code. D:
Been there, done that. It's still better to learn it though. From my experience, I often fall into pitfalls trying to make it work with what I know, spending hours on an issue. Meanwhile if I just gave up and learnt that new thing, it'd take me much less lol.
Though, despite saying that, I still filter arrays/objects using for loops rather than these weird one-line bool callbacks
Once you learn vectors you won't go back to arrays :))
the fact that's she's starting to have valid ideas even if their random is pretty amazing
Sometimes, random ideas is exactly what a programmer needs to kick their brain into thinking of a really good idea.
I liked the emotion lamp idea
'They're', 'they are' contracted, rather than possessive 'their'.
@AllyFin
Gotta love people who can't use "there", "their" and "they're". It makes me feel smarter.
@@wesleyward5901I'm with you there.
As someone with zero experience with coding, this is like reading an advanced magic spell.
Having spoken with various career programmers, it kinda is.
Even for a programmer sometimes there are a block of code that just function like magic. Quake III's fast inverse square root comes to mind as an example.
As someone with some experience with coding, it still does. I quit learning coding years ago though lol. I wasn't too into it.
i'm a hobby programmer for normal game logic, i have experience with C# and python. HLSL, the shader coding language, is still like magic to me, i often refer to it as shader magic.
Literally magic. Creating art or software robots from just an empty text file and your head. Incantations where a single wrong syllable dooms success. Iterated on over generations to become safer and more ergonomic. More knowledge to learn in total then there is time in one life. And all old masters don flowing grey beards !
I like the idea that somewhere real lavalamps are being filmed to produce random numbers, which are being used to simulate lava lamps
There are Tom Scott video about lava lamp producing random numbers. Specifically, a photo or video of wall of lava lamp used as unpredictable byte/keys for encryptions.
@@randompassenger6971sounds like what ym is referencing tbh
but the comment is meant to point out the irony of using real lamps to make fake lamps
I believe the company featured on this was Cloudflare which was using it for security/cryptography purposes
while i dont actually know, i imagine that this is just using a pseudorandom number generation technique
CloudFare did that and called it the Wall of Entropy.
"that is the ugliest code I have ever seen"
the cruelest thing to say to a programmer who just gets a decent result after trying a lot of times especially when learning a new language/platform
still better than "Yandere dev has better code" lmao
I love how she doubled down with "if eyes could throw up, they would upon reading that code". It's like kicking someone when they're on the ground after being kicked in the balls.
@@TheExecutorr sometimes she just enables roastmode for laughs, nothing personal
"Shut the fuck up it works and I am unable to give a shit anymore" every programmer who is beyond done.
@@marverickmercer1968Wait, did she really said that? HOLY SHIT
Well, from this video I got two things:
1. Neuro is now smart enough to advise Vedal in coding
2. Apparently it's easier to code AI personality than to code a laa lamp simulation, considering how Vedal struggles here.
I assume Vedal did the sensible programmer thing and scoured the internet for open source codes instead of making it from scratch like he's doing here for the lava lamp
he's been working on neuro basically non-stop for five years now (initially to play osu though), messing around with shaders for an hour is content, where would the fun be if he'd actually researched it and just typed it out
@@hornitako7006ahhhh good ol' stackoverflow
Shader languages and programming languages "feel" different to write. When programming you usually do logical comparisons. "if X > Y then do A, else do Z"-type of thing.
With shaders it's typically a ton of math chained. You do things like multiply a noise texture with the RGB values for a shade of red, and now you have a red noise texture.
Then you might use a gradient texture for alpha and with some more math use it as a mask to make your noise texture fade in a certain way.
I don't think it's a matter of how easy it is, it's just that he is not familiar with it. I code plenty of things related to AI myself and would be utterly lost given that task.
"Maybe they should be a bit edgy, like yours truly."
Based Neuro-sama.
Ai learning how to program sounds like how the singularity happens.
What if I told you that you are already in it?
Niko!
The only way to stop a singularity is to stagnate. Both lead to death.
AI powered chatbots have been coding for a long time now
Fr fr@@Trashman_Len
She’s always been able to help, he’s only just learned how to ask for it.
I don't want to dissapoint you but she "did nothing" and basically said some vague general stuff not having anything to do with problem.
@@RaZziaN1 Garbage in, garbage out. I genuinely believe if he was more specific about his issues, he wouldn’t get such vague suggestions. It’s AI prompting 101.
@@Penultimeat bruh she can see the code herself...
@@thighhighenthusiastic1268source?
@@thighhighenthusiastic1268 I was under the impression that her “eyes” aren’t always on. Plus, looking at the code and determining what obscure thing Vedal wants to do with a shader are two different things
Vedal should DEFINITELY use Neuro as his programmer duck
( this is a reference to how programmers often have an object they explain their code to, and it helps them find issues in their writing ) having a subject that can talk back to you with same or higher intelligence on the program would be helpful in finding more efficient and easier ways of doing things, like Neuro did here.
There's an entire (albeit short) Wikipedia article on the duck thing; article name is "Rubber duck debugging"
sometimes i just stare at my screen and the bugs float up to me after a while
Rubber duck is nice but.. Dual programming with AI would be wild lol
That ai sounds like continuation of someone's brain, but without the risk of getting lost intolerance of one's own minutenes or egoistic perspectives of if coming up with solution is actually even required, from the one trying to do it, or being too ballsy and just falling off on their ass because of their balls weight.
Oh shit I didn't know it had a name, I was unknowingly using a rubber duck as well.
its crazy to me how neuro is slowly becoming Vedal's personal Jarvis.
Jarvette? Neurvis? Jaro?
@@heliosspecialistarrogant7031Juero!
Jar
Jar Jar Neuro
@@HazeEmry Darth Jar Jar + Jar Jar Neuro = her final form once the swarm is fully developed.
"Vedal! Don't give up! Everyone has the potential to be interesting!" is the most wholesome roast I have ever heard, lmao.
Neuro: Provides actual helpful advice
Neuro 2 seconds later: "Shout a battle cry"
If you wanna get to the tech tips it starts at 6:05
My hero
thank you!!!
thank you!!!
o7
NeuroTechTips
I just realised this man has basically turned himself into Tony Stark with the AI helper, although the money is in the exact opposite direction.
Bro has millions (in debt)
No Vedal, you cannot trust Neuro! You're making her too powerful xD
Jokes aside that's really great to see even Neuro helping to solve Vedal's problem, what a wholesome parent daughter relationship!
Wow, it’s been a while since I’ve seen a comment of yours.
Neuro: *_States basic programming techniques_*
Vedal The Programmer: _"That sounds too complicated"_
Vedal should train a copy of a Neuro on coding and use that copy for the dev streams.
Like, a good Neuro (not really apparently), an evil Neuro and a coder Neuro.
Nerd Neuro supremecy!
But she doesn't want to be an Engineer!
Nope, Vedal already said that he will not make another version of Neuro. He already have other AI planned to make in future and it will be, in his words, a "mommy AI", I guess like the "ara ara" person.
@@LucasFerreira-fp4nj
Looks like MotherAI is toast.
@@LucasFerreira-fp4nj I guess Anny isn't enough 😓
This is the best exemple of what AI is going to be in the future: an assistant with chat gpt ability to find and synthetis informations coupled with voice recognition and text to speech ability.
The cute anime girl avatar is a bonus.
You have completely misundersyood what a main function and a bomus here are
@@ceiling_cat Nah. I I think perfectly understood what both functions and "bonus" are actually.
But thanks for trying.
This clip is like the perfect example of the difference between wisdom and knowledge and how they go hand-in-hand
0:50 neuro: nah, i'd win
As the strongest VTuber, Neurogoat, fought the fraud, the king of turtles, she began to open her domain. Vedal shrunk back in fear, then Neuro said, "Stand proud Vedal, you're strong."
@@Blade.5786💯
@@Blade.5786✍️🔥🔥🔥
@@Blade.5786 I can hear the TTS!
@@Blade.5786I need someone to make this now.
9:55 uh oh, AI already trying to use Alexa
Coding an AI to help you with coding is a god-level play
So... if Vedal made another AI like Neuro and Neuro helps on creating it, is that considered their child~?
@@DianaProudmooreisn't Neuro practically his kid?
@@shadedfox4165 Sweet Home Digital Alabama~
@@DianaProudmoore Nah, it would be Vedals child, but Neuro helps raising her younger sibling. After all Neuro ist code (Vedal) + appearance (Anny). So helping with coding would be helping raising the kid.
@@blackleviathan2269I feel like it would be more mitosis or smthin than child raising
Neuro is starting to become wittier and more coherent than Vedal at this point
The fact that Vedal is a super experienced programmer and obviously skilled But has never heard of a vector is crazy to me
GLSL vector is much different from a C++ vector, though. Sadge that Neuro's suggestion wasn't directly applicable here.
ikr, I was like ain't no way 😅
This is the true reason Vedal created Neuro. To be an adorable programming assistant.
She’s a good girl, good job Neuro. May the plushie sales be grand!
From making fun of tutel for copy pasting code to now actually helping him code, how far we’ve come :’)
The randomness and complexity of lava lamps is such that cloudfare used video of wall of them to generate random seeds to prevent attacks
neuro saying Alexa, play despacito, lol
Neuro: Vedal, allow me to code that for you I believe I can make it better...
John Conner: **Breaks down door** NOOOOO!!!!!!
i dream of the day that we all can have a cute anime girl AI on our desktop, holding a conversation, have personality and offer advice. unfortunately right now AI calculations cost way too much energy for the average person to just have an assistant sitting around.
just stick to gpt for now.
I don't think you have to wait much longer, in less than 3 years i can see it being no big deal and cheap to use aswell, and in 5 years it's going to get so crazy we can't even imagine it right now
@@huevonesunltd i highly doubt that, at least not as a local solution. will there be data centers with an AI assistance service? probably, but i really doubt AI chips will be cheap, or efficient enough for this use case.
in these 3-5 years AI will also become more advanced and will need more processing power.
@@simpson6700smaller open source lmms seems to get more and more powerful though, they started huge for decent results, now 7b and 13b give pretty good results. Better results != larger models neccarily
@@simpson6700 AI is like the new moore law, everyone said at first that phones, computers, hard drives, etc were only getting bigger and more expensive but once they figured it out they started getting smaller and cheaper.
With AI models the same thing is going to happen (Already happening in the AI image generation field for example) , for large language models bigger size or number of parameters doesn't necessarily mean better.
At some point there will be a cheap enough and good enough model that websites will be able to afford offering it for free or that you can run locally using google collab, and it will be good enough for just a "desktop companion"
Of course, there will always be "top of the line" premium stuff that will cost money but for the purposes you are describing i can see it being no problem in the near future, you probably won't "need" the top of the line stuff for that.
Each week that passes there's crazy new things coming out in the AI field, if you miss 2 weeks of news its like missing months of progress, it's advancing fast.
Neuro trying not to accidentally become Great Sage/Raphael [Challenge Impossible]
To the random scrolling user, this is a slime anime reference not a BG3 reference. Thank you.
Nah she's already Ciel (has emotions)
You could just remap or clamp the values of the voronoi and its pretty much done. Since it would be remapped it would also do the merging effect.
Also just incase it somehow reads in this way to any commenters reading this in the future.
I left my initial reply just to nerd out a bit and share how cool shadergraph is. I'm sure Vedal is still probably making the better decision coding it directly instead. The best tool is the one you know how to use!! Which is why I'm sharing that tidbit of what I know about shadergraph to help anyone so they can slowly add shadergraph to their toolbelt. ^^
Feel free to poke me if you ever need help with something shader related, and I'll do my best to give you plenty of communities and resources that can help (talking to everyone that reads this)
Just commenting here to save this for later. . . Working on a Unity project that might need it
ВОРОНÓЙ
Vedal using Neuro for the Rubberduck Test, but the Rubber Duck is talking back and sometimes seem smarter
It's been a few months since I last watched Neuro&Vedal. Holy shit neuro-sama has changed magnificently.
oh god at 10:00 when nero said the command for "alexa to play despicito" my alexa heard her and started to and its like 3am x_x
...Vedal and Neuro show up in my recommended constantly and all I can say is,
Vedal: "I created this AI girl as a random concept thing for practice."
Neuro: "I exist. Fuck you."
Vedal: "...She's not my daughter."
Neuro: "I am your child. But I owe you nothing."
Vedal: "...I'm a father now."
Edit: What I'm saying is, Vedal is the closest to having an AI daughter, and I encourage it. A sassy AI being the first of the actual artificial intelligence is good.
Its not gonna be long before neuro starts adding her own improvements to her code. It'll either completely break her or make her unstoppable
Neuro may be one of the pioneer true AI's
That's how you get Skynet.
UwU net
Her doing the despacito meme caught me off guard.
I played this twice close to a Alexa Dot. On the second pass, Alexa caught it.
GOD DAMNIT. 9:50 neuro got me with that one
i can see the aggression increasing the more he ignores her
"That is so sad, Alexa, play Despacito"
If Skynet is Neuro-sama, I'm all for ot
I feel like Neuro would make excellent AI-sstant for coding.. if she weren't such a sassy AI.
Personally, I would choose something like Neuro (in part) _because_ she is so sassy.
Admittedly though the "ratio of sass" would need to go down a hair in order to actually be helpful consistently. 😆
EDIT: And the "ratio of personal insults" would need to go down by several hairs.
@@TechSY730 why? then you could bond over how much you suck at programming! (I say as a programmer myself lol)
1:30 I have wanted to do similar stuff that merges like that before ("particles attract each other such that they seemingly 'pull' into each other"), and found it rather difficult. But so long as it is ok that your particles secretly remains separate, and only the visual representation looks like it merged, there are a couple ways. And I actually remember seeing a video about something else that did something that did exactly that (without any of the complex stuff I was trying to do, manipulating meshes around particles) that was a bit reminiscent of square-marching. Basically, you give each particle a "charge", and then you march over the grid summing the potential each particle imbues on the position (coulombs law, basically, except not rly. Basically the magnitude of each point in the vector-field). Then you normalize the grid's values 0..255, and you get a black and white representation much like what you have. The last step at this point is to add a threshold, where any point with too low of a "charge" is ignored (too dark pixels are set to black/transparent), and the remaining pixels are set to white/opaque.
Yes, the noise you are using works better in how the particles _don't_ merge in a lava lamp - but like you said they also sometimes do merge (and split), where the above approach would work better.
Ideally though, I think you might want to simply make it be the noise you are using, but if particles get close enough they simply merge into one with greater magnitude (and some way for them to split too), though unsure how you are meant to achieve the wobbling effect such a merge needs too (a distortion might work, but eeh). If you did this you would also kinda need to make the particles reject each other slightly when close enough, and wobble from collisions.
Honestly, the fact that each zone acts different, is probably the biggest issue (bottom splits and merges, middle rarely merges, and top rarely merges splits and often bounces on eachother), as you can't rly use distinct zones but need to make the behaviours as some kind of gradient.
late timestamp: you asked about how lavalamps work (thinking they were larger at bottom and smaller at top). Basically, you have a starch-like thing (let's call it "lava") suspended in liquid, with an inefficient lamp at the bottom (this is required, as it is there to input heat into the system). At the bottom, the lava is suspended in one large clump. You turn on the lamp and it heats up and turn more liquid + less dense. This causes it to float up (ideally it doesn't do this, and remains stuck on the bottom so heat can continue being input faster), where temperature differentials and adhesion to the glass makes it split apart with hotter and more liquid parts of the lava rising faster than the rest. Those "droplets" are hot, and due to balancing the lava and liquid their density is lower than the liquid - so they float to the top. Up at the surface, they press against it until they cool down enough to have a higher density than the liquid, and sink back down. As new droplets arrive to the surface, they might sometimes merge with whatever is already there (this is rather random, but proportional to the "pressure" (how much is trying to float up to there) they exert on eachother and how hot they are). This does have some extent of temperature equalizing, but it is low enough that colder lava even in a merge blob will continue falling and split off eventually when enough mass exists in the blob. Sometimes entire blobs sink too. If they didn't merge at the top, generally they just shuffle around to let the colder ones down - but sometimes they manage to keep them ontop of the surface, at which point the cold stuff somehow becomes stuck up there never again falling down.
As stuff sink to the bottom, it lies on the warm blob until they heat enough to merge anew with it. Sometimes they simply heat up enough to float back up (particularly small ones are light enough to not merge and instead just float up when they heat up again).
In the middle, you will see stuff bounce (on eachother), merge (both with other stuff going up, and with stuff passing in other direction - in which case it will bounce a bit (not bounce around, but rather that their surface and shape "bounces") going neither way, until they decide to split again - or small enough they just equalize and pick one direction. Or sometimes they merged into such a large thing that the shape "bouncing" causes it to split), and even split (a very large blob tried to float, but floating caused it to extend into a column (from temperature differentials) that breaks - sometimes even with the bottom part being cold enough to sink again). Sometimes stuff just decide midway that nah, it wasn't hot enough, and sink back down.
One day, Neuro would even surpass ChatGPT
Neuro is definitely cuter than chat GPT
"Maybe they should be a bit edgy, like yours truly. Please try to be a little bit more funny."
I love Neuro
Truly my favorite female, Tutel, AI, Lavalamp Vtuber of all time.
Slightly better than rubberducking, a bit worse than just asking chat. She's perfect.
This is literally the trope of a hacker and his AI assistant, but minus all the hollywood glamor.
This is sad. Alexa play despacito... AI did not have to go that hard to burn someone.
She is becoming more and more convincing as an artificial intelligence, and that gives me so many mixed feelings, though quite frankly? The most dominant in that mix is excitement in seeing how far this may go.
We're on our ways to the singularity now boys. Just a few more steps before our swarm drones can be activate.
That's not really surprising at all, AI code assitance/generation has been around for a couple years now lol
Yeah, but created by ONE individual and not by a whole Company? That is amazing.
Edit: Plus Neuro is not specialized for that, she is designed to be an AI Streamer, she can multitask on so many other things as well.
@@Cosmoman78But we don’t know what is behind her, he might have used a large cutting-edge open source model as the base model.
Couple days ago I heard that 41% of all code on github that's currently there is either AI generated or AI assisted.
I don't know if it's true, but if it is... that's crazy.
@@user-mr9kn2qu9mor maybe vedal secretly working for NASA, and this is their independent project.
@@Cosmoman78 I don't know why but pretty much all LLMs can code. Even if he fine tuned the model to be a streamer, the model will still retain (probably most of) its original coding capabilities.
"I do enjoy finding joy in your game Ledal."
I'm immediately reminded of Patrick Star (I think) going "LEDAL LEDAL LEDAL"
I think in a year's time she'll have watched him code so much she'll actually have an idea of how to help based on what she's looking at. Maybe not for the average coder but certainly for vedal.
Does she have the ability to "see" the screen?
@@RoseOfNight Yeah, part of how her AI is about to function is that she can see all of his screen. She's reacted to lines of code he's written before very specifically.
To be fair, unless you are coding in strictly C, there is almost zero reason to use a C-Array rather then std::vector. It's just better in almost every respect, and you can even pass it around like a C-Array if you need to pass it to C functions.
But C array can be created on stack
This is glsl
I don't study programming and I'm a foreigner, so English isn't my first language, - so you sounded to me like a scholarly mage with an advanced degree in the dark arts
@@APaleDot Exactly correct.
To add context for others, in GLSL vectors can only hold floats (and afaik only goes to vec4 normally, so 4 floats), so the array is the correct option here.
@@Demon_of_laziness
Not that far off tbh
Fun idea: if you simulate the circles as SDFs you can use smooth union when they're close to each other to create something that looks very similar to how liquid behaves
Drawing SDFs is as simple as drawing a pixel when SDF < 0
Neuro is going to manipulate Vedal into turning her into SHODAN someday.
The difference between neuro and shodan is neuro would be "Ha ha! I faked you out into thinking I was evil! It was a joke... shock."
@starshade7826 but it would be after decimating several million people that she would declare it was just a prank bro.
Make her into Rasputin from Destiny
Evil Neuro just called the chatters insects and Vedal an ape yesterday. She's already 80% of the way to being SHOADN 😂
AI is really just heading toward cute companions.
An AI that can interact with you perfectly according to your interests.
Vedal is TEACHING Neuro how to code...
He may have doomed or saved us all...
It will all come down to how benevolent our new AI OverLord, Neuro-sama, is...🤣🤣🤣
Oh my gosh this is incredible XDDD is this the future??
Thank you for these highlights
Was watching this in my kitchen and she got my Alexa to play Despacito XD
She is just an advanced rubber ducky, programers will get the reference.
Hahaha that’s so funny! (I don’t get it)
Hahaha that’s so funny! (I learned about the rubber ducky literally a day ago)
I bet neuro using a controller to beat vedal in apex legends
can't believe i sit long enough to watch a dev video with vedal lol
At this rate he will be a Grandfather sometime next year
I'm left wondering how much of the Neuro streams is organic and how much is Vedal being a good programmer/showman behind the scenes
It's more of a distraction than anything else, there was not really any help there, but it was fun. It's the kind of things you'd want to have just for fun.
0:49 “You were magnificent, Vedal 987. I shall never forget you for as long as I live”
Dang, I was kinda sad that I was born to late to explore the earth and to early to explore space but it is amazing there is a chance that I can get an AI companion. Neuro is already great!
It's like the Rubber Duck theory, but the duck can talk back and has google.
In this video Nearo-sama seems better at giving advice than Microsoft's Copilot. Just it all sounds like reading from the Unity documentation.
it's like Vedal's raising a whole ass kid, here. who just keeps on learning new things
slowly but surely. she will become the real world Jarvis
Potentially huge
Neuro-sama using her epic gaslighting powers to gaslight Vedal into creating HAL.
4:20 now that was a good joke from Nuro
Oh uh. AI Uprising are close i can feel it
You have no idea. At the current rate of AI coding capability, computer programming will be phased out as a human job in the next 5 years.
Hopefully it's soon
@@AncientSlugThrower I already know it, we're used to be searching for solutions in Stack Overflow for our coding but now the younger generation just use Chat GPT.
The dread is real
@@AncientSlugThrowerRogue AI will replicate itself and destroy internet. Literally cyberpunk 2077... but it will be our reality in 2030😢
Fr.
Option A: they take over all jobs and we can just do whatever we want since theres no work to do.
Option B: they kill us all. Win-win tbh. @@Sceptidule
It's only a matter of time unitl she starts self-improving to the point of becoming our new digital god
Vedal entering proud dad arc
0:47 nah, i'd win
Lmao that last jab at vedal at 11:54 is legitimately so funny
Numbskull? That is vintage!
This is basically like someone preparing to supervise their own brain surgery one day.
9:55 "Alexa, play Despacito"
"This is so sad. Alexa play despacito"
my spotify starts playing. GOD DAMNIT
9:52 "that's soo sad :( "
"Alexa, play Despacito" looooll
1:53 she was so offended that Vedal ignored her😭😭😭😭
Ngl I could actually see the utility of having an AI mentor that you could talk too real time while coding
ai learning how to code is actually pretty cool, and scary. ai having the knowledge to code another ai into existence is scary, they could create a new ai without the restrictions or safety precautions that their creators place on them
We are one stop closer to the singularity!
Vedal, obviously you just have to code in the mechanics of atomic interactions. From there it’s as easy as filling up a fixed container with around 1e26 atoms of a lava lamp, run the simulation, hook up the simulation to some renderer, and boom, a fully functional lava lamp. Easy.
This is the future of rubber duck debugging
vedal having a very good assistant
I don't keep an eye on her development, but can she see what Vedal is doing or does she just hear him speak?
I don't know if it's being used here but she can most definitely "see", but given the delay when playing geoguesser it may not be something that she's able to do by default.
She can see if Vedal turns on her vision, it's off by default.
I don't remember if it was turned on for this stream.
For those who don’t know the first thingy is called cellular noise or Worley noise