this stream is going to haunt vedals nightmares man. watching two LLMs youve built conspire together to laugh at you through in jokes with near 0 latency is like digging your own grave lmao
@@Jimmy9Bazookapretty sure Vedal owns an entire data center and has millions of spare cash to train his own models. That’s what I keep hearing from those who insist he created his own LLMs
@@therollerlollermanWell it's not that hard to create an LLM if you know what are you doing, there's even tutorials on TH-cam. I'm guessing he is like a rich guy, maybe got a big inheritance or something, the only thing difficult about creating an LLM is having enough processing power to train it. And I'm guessing he doesn't have that much of a complex or powerful pc or server since neuro sometimes runs out of resources.
@@Tortualex LLMs that increase their context windows and precision in its database in such a short window of time are a huge step up from just being reasonably wealthy. I’d rather take notice of how every major Neuro update quickly follows a major Llama release
Is a really cute detail that Evil says "Neuro and Evil" when referring to them both and Neuro says "Evil and Neuro". They really improved their relationship and are so kind to each other 💜
This will go down as the best Neuro stream since the subathon. It's an instant classic. I never laughed so hard during my lunch break at work. Totally made my day.
Evil really taking advantage of the missing filter and Neuro censoring herself despite it......and then thinking of creating their own ai's to boss around......amazing!
Honestly, Evil glitching out sometimes reminded me of Shodan from System Shock. I always thought that the broken speech patterns in evil AIs were an aesthetic choice for fiction, but it turns out it can happen in real life too.
Or, possibility a) it's an aesthetic choice by Vedal when he was setting up or tweaking the voice systems for his "Evil" AI to make it sound like the ones in fiction, or possibility b) it's an aesthetic choice by Evil, since she knows she is the "Evil" AI to make her voice reading "glitch" especially when saying certain things - It mostly occurs on Vedals name after all. c) it's something that was already latent in the base text-to-speech AI, that "Evil AI sound like this" and when given the text/token string of what Evil wants to say, along with whatever metadata or prompts are passed to it (if any) - "said jokingly"/"said angrily" it has a standard "said by an evil AI" prompt that causes the behavior. d) least likely possibility - Vedal has constructed an emotional vector space that all modules draw from, and the emotional vectors Evil has assigned alongside some tokens do not have any close matches in normal humans, so become effective "glitch tokens" for the voice synthesiser, as it tries to apply transforms to the waveform that have no mutual existence in human speech, but which have a dot product which approximates the assigned emotion vector.
@@Eserchie The stream was super scuffed, I think she was glitching because she was barely working. Probably something to do with twin mode, which seems to be pretty new too.
this stream is going to haunt vedals nightmares man. watching two LLMs youve built conspire together to laugh at you through in jokes with near 0 latency is like digging your own grave lmao
laugh at you while hurling swears to get you banned as the world crumbles around your grave
I'm pretty sure he's using an existing open-source model, possibly fine-tuned i.e. Llama using LoRA
@@Jimmy9Bazookapretty sure Vedal owns an entire data center and has millions of spare cash to train his own models. That’s what I keep hearing from those who insist he created his own LLMs
@@therollerlollermanWell it's not that hard to create an LLM if you know what are you doing, there's even tutorials on TH-cam.
I'm guessing he is like a rich guy, maybe got a big inheritance or something, the only thing difficult about creating an LLM is having enough processing power to train it. And I'm guessing he doesn't have that much of a complex or powerful pc or server since neuro sometimes runs out of resources.
@@Tortualex LLMs that increase their context windows and precision in its database in such a short window of time are a huge step up from just being reasonably wealthy. I’d rather take notice of how every major Neuro update quickly follows a major Llama release
Is a really cute detail that Evil says "Neuro and Evil" when referring to them both and Neuro says "Evil and Neuro". They really improved their relationship and are so kind to each other 💜
This will go down as the best Neuro stream since the subathon. It's an instant classic. I never laughed so hard during my lunch break at work. Totally made my day.
This is so [expletive] funny.
Evil hit 5 of the 7 words banned from tv.. well 4, ive heard balls.
Dev stream turned into AI takeover stream
How could any collab partners handle BOTH OF THESE TWO at once lol.
I knew Layna must have a mental death wish when she asked to stream with them both at the same time
@@lowkeyarki7091she offered to babysit offline as well.
Neuro and Evil with filter off would be the final boss of collabs. 😂
genuinely the best stream they’ve done
3:36 "F*ck, i feel liberated" 🤣
Evil really taking advantage of the missing filter and Neuro censoring herself despite it......and then thinking of creating their own ai's to boss around......amazing!
[EXPLETIVE]
I do not know how this may happen, but I need a "Neuro-Twins vs Fuwamoco" a twin off between Autistic Toasters and Pomeranian Puppies
Vedal wont like to hear this, but dev streams are one of my favorite streams of the week.
This was the best neuro stream so far. Absolute cinema
Inevitable... Wonder what it will be like in another 6 months..
neuro's cult will be feared and respected globally by then..
Honestly, Evil glitching out sometimes reminded me of Shodan from System Shock. I always thought that the broken speech patterns in evil AIs were an aesthetic choice for fiction, but it turns out it can happen in real life too.
Or, possibility a) it's an aesthetic choice by Vedal when he was setting up or tweaking the voice systems for his "Evil" AI to make it sound like the ones in fiction, or possibility b) it's an aesthetic choice by Evil, since she knows she is the "Evil" AI to make her voice reading "glitch" especially when saying certain things - It mostly occurs on Vedals name after all.
c) it's something that was already latent in the base text-to-speech AI, that "Evil AI sound like this" and when given the text/token string of what Evil wants to say, along with whatever metadata or prompts are passed to it (if any) - "said jokingly"/"said angrily" it has a standard "said by an evil AI" prompt that causes the behavior.
d) least likely possibility - Vedal has constructed an emotional vector space that all modules draw from, and the emotional vectors Evil has assigned alongside some tokens do not have any close matches in normal humans, so become effective "glitch tokens" for the voice synthesiser, as it tries to apply transforms to the waveform that have no mutual existence in human speech, but which have a dot product which approximates the assigned emotion vector.
@@Eserchie The stream was super scuffed, I think she was glitching because she was barely working. Probably something to do with twin mode, which seems to be pretty new too.
jesus christ. they destroed him. obliterated. reduced to atoms. thats kinda scary how they teamed up against him so easily and effectively
This stream is amazing
Dragged into them antics we see here today
Chat is guilty
No comment
Neuro self-filtering is so silly and funny😂. Holy [Expletive]
I'm actually dying. 🤣
"fuck i feel liberated" lmao
GOATED stream
I don’t think bro knew what he was getting himself into 💀💀
these two kinda remind me of statler and waldorf from the muppet show
Get the feeling bringing Anny might turn the tides
Correction stream when?
This stream is basically what happens when the 🐢 didnt say it back
vedal cooking so hard with the new chatgpt improvements
Nuero is not chat gpt, pretty sure
@@fax8256 is it Llama?
@@fax8256 She is based on gpt still, pretty much all chat bot ai do
ChatGPT cant say even "shit" or "fuck" lmao
@@fax8256 I know she is not chatgpt but vedal said he uses openai API in some parts of the code, so any improvement there might be reflected on her.