I wish more people would say what they mean, or at least be understanding of people who are not able to understand things that are unspoken or unwritten.
This keeps getting better and better.
The best news I've heard in a while.
Bingo. It's not materially different. Virtually all human expression has influences and inspriations.
Regarding Neuro's art, the reason she is able to code scalable vector graphics is because the code for other people's svg art is in her training data, just like it is in the training data of LLMs such as ChatGPT, which can do the same thing.
The osu bot is not what allows her to interact with Vedal and chat, the LLM is. Vedal acknowledges on his website that she is mainly an LLM.
And as for energy usage, being run locally does not make her exceptionally energy-efficient. Local computing is actually less energy efficient than cloud computing. There are advantages to local computing, but energy efficiency is not one of them.
Locally run does not mean locally created. It is not feasible for an individual to create an LLM on Neuro's level from scratch on a PC. Neuro is based around an existing model pre-trained from a massive dataset that you may consider to be "harmful."
LLM stands for "Large Language Model," so it mainly generates text. Though there was a time when Vedal had Neuro make Scalable Vector Graphics images, so technically she does both. In any case, text generation is just as generative as image or video generation.
The tech behind her is the same tech that people claim to hate, though.
I personally think it's fine, but if you believe that the technology behind ChatGPT, Grok, and Claude is inherently evil, then there's no reason you should view Neuro-sama any differently.
Neuro is built on top of a pre-trained LLM that was made with a massive dataset that you may consider to be 'stolen from the backs of others.'
Not sure what Vedal quote you're referencing, but that's not how LLMs work. They are not continuously retrained in real time. Neuro does have some memory, but it just gives context for her interactions. Even if what you are claiming were somehow true, it would not make her not generative.
The osu bot isn't what allows Neuro to interact with Vedal and chat. In her current state, she is mainly an LLM, which is in fact generative.
Neuro does those things because of her personality which Vedal created with a system prompt and probably some fine-tuning. Nothing about her transcends the limitations of LLMs or makes her not generative.
The osu bot is not what allows her to interact with Vedal and chat. Neuro in her current form is mainlyban LLM.
Neuro is based around an LLM, which is the currently most popular chatbot tech.
The osu bot is not what allows her to interact with Vedal and chat. Neuro in her current form is mainly an LLM, which is in fact generative.
There is no secret. To make the most basic AI VTuber all you need to do is connect an LLM to a TTS and a VTuber model, and integrate it with Twitch chat. Other people have done it too. Vedal is just the most popular.
Neuro-sama is in fact an AI.
Influence does not equal theft. If AI-haters applied the same standard to all media that they do to AI-generated media, virtually all human expression would be "stolen."
Auto-correct is not traditionally LLM-powered.
She is based on top of a pre-trained LLM which was created with a massive dataset that you may consider to be 'stolen works.'
My wording should have been better. I should have said "every phrase she utters."
While Vedal probably has done fine-tuning with his own data, Neuro is based around a pre-trained generative LLM that you may not consider to be "ethical."
And being a passion project and being locally run do not preclude her from being generative.
Every word she utters is a "new work."
When someone offered to buy Neuro, it's because she is a profitable brand, not because Vedal has some unheard-of technology.
While "Neuro-sama" is also the name that Vedal gave to the osu bot, Neuro in her current state is an AI VTuber, which is all based around the pre-trained LLM. What Vedal did is more akin to customizing a car than building one.
If Neuro wasn't generative, she wouldn't be able to say anything. Everything she says is literally AI-generated.
