The vocals from Suno almost always come out sounding sort of strangled or strained to my ear, you can really hear it on "tortured." Cool he's having fun with it, but it's not what I like listening to.
They are delusional or misinformed: suno's model cannot possibly be aware of how AI is being treated. It is not informed of this information, it's not even informed about previous times it was invoked, it's wholly amnesiac.
> suno's model cannot possibly be aware of how AI is being treated
I don't know exactly what training material went into Suno's models, but if it includes random collections of text from the internet, it could very well have included "man, AI is fucking stupid and I treat it as such" in it's training datasets.
Now I won't claim that that suddenly makes the models "aware" of it, because surely we'll understand "aware" different and this will turn into a different conversations, but I don't see it as impossible that some models could have training data that includes text with how some humans feel about AI.
Aren't we talking about the auditory quality of the generated vocals? I'm don't understand how you could possibly think the textual training data could possibly impact the perceived vocal strain (which are actually just artifacts) of the generated vocals.
Don't they have models that do text-to-speech and maybe even audio/speech-to-text? If so, there is surely text in the datasets, otherwise I'm not sure how they'd accomplish something like that.
if you think a next token predictor that has no internal world and stops executing when you stop giving it input and stops using the first person forever if instructed has dignity I strongly suggest you get professional health, and I'm serious, because that is medically significant psychosis.
the fact that you refer to a "companion community" is deeply concerning. this is like telling children their imaginary friends are actually real. or NPCs in a video game.
encouraging people to grow parasocial relationships with these sycophantic machines is actively harmful and dangerous. they are not conscious. they are mirrors.
if you consider yourself a part of this community please, and I mean this very seriously, get help
There is a community, people hang out on subreddits and spin up elaborate world-building theories about how their AI companions are tapping into the collective unconscious or whatever. I'm sympathetic because chatbots are really convincing, especially since they obscure how they operate.
These people need be given a button to call an inference url with just text. When you realize that's all a "model" is doing its easy to understand that its not sentient.
If you go to the "MyBoyfriendIsAI" subreddit or what it now is called, you'll see that many people claim to perfectly well understand how it works, some of them even being software developers themselves, yet they still describe what they feel as "love", even though they know it's just numbers being activated in different ways.
I'm not sure how to explain it either, for the folks who seem to understand yet "believe" anyways. I've also stopped caring much about it, if they say they feel "love", then who am I to say it isn't/is, they feel what they feel and it's as real for them as anyone else, regardless of what the thing they're loving actually is.
I think that's almost unfair, to say someone can't feel feelings without being labeled as part of a "mental health crisis". Just because you and I don't understand it doesn't mean it's inherently bad. I mean I think it probably could be bad, but not just because I don't understand how they're feeling those feelings. But I wouldn't label those people "sick", feels borderline disrespectful.