3 ms·

Zero shot implies that it was given no direct examples [0] So, in this case, it wasn't given any examples of the exact voice in combination with text, it is jus

by pstorm 3y ago

Zero shot implies that it was given no direct examples [0] So, in this case, it wasn't given any examples of the exact voice in combination with text, it is just using the prompt voice + a prompt text to generate new audio. [0] https://en.wikipedia.org/wiki/Zero-shot_learning https://en.wikipedia.org/wiki/Zero-shot_learning

cwkoss 3y ago
But isn't nearly all AI generated content "zero shot" to some extent? Like even if it has training for "foo" and training data for "bar", the combination of "foo" and "bar" would be novel and 'zero shot'-y if the training set didn't have "foo bar" examples. To me, it seems that the only AI generated content that isn't zero shot would be the narrow subset of generations where it has multiple training examples for the exact requested prompt. i.e. anything that is composing multiple pieces of information together is "zero shot". How should I parse 'shot' in the context of this term? Shot makes me think 'attempts', seems like a weird word to use for ~= "training examples"
CyberDildonics 3y ago
If you have the audio you want to generate, that's just called playback. If you are synthesizing something of course you don't have a direct sample of it.
- InspiredIdiot 3y ago
  I think parent is saying that the model does not require any paired samples of the voice to be synthesized and corresponding text. So based on my understanding: one shot - given the text "run faster" along with Alan Greenspan's voice pronouncing that phrase, the model can produce Alan Greenspan's voice saying any other phrase zero shot - given only Alan Greenspan's voice pronouncing "run faster" but no text version of what was said, the model can produce Alan Greenspan's voice saying any other phrase
  - CyberDildonics 3y ago
    Does that mean a shot is text?

3 ms·

Zero shot implies that it was given no direct examples [0] So, in this case, it wasn't given any examples of the exact voice in combination with text, it is jus

by pstorm 3y ago

cwkoss 3y ago
But isn't nearly all AI generated content "zero shot" to some extent? Like even if it has training for "foo" and training data for "bar", the combination of "foo" and "bar" would be novel and 'zero shot'-y if the training set didn't have "foo bar" examples. To me, it seems that the only AI generated content that isn't zero shot would be the narrow subset of generations where it has multiple training examples for the exact requested prompt. i.e. anything that is composing multiple pieces of information together is "zero shot". How should I parse 'shot' in the context of this term? Shot makes me think 'attempts', seems like a weird word to use for ~= "training examples"
CyberDildonics 3y ago
If you have the audio you want to generate, that's just called playback. If you are synthesizing something of course you don't have a direct sample of it.
- InspiredIdiot 3y ago
  I think parent is saying that the model does not require any paired samples of the voice to be synthesized and corresponding text. So based on my understanding: one shot - given the text "run faster" along with Alan Greenspan's voice pronouncing that phrase, the model can produce Alan Greenspan's voice saying any other phrase zero shot - given only Alan Greenspan's voice pronouncing "run faster" but no text version of what was said, the model can produce Alan Greenspan's voice saying any other phrase
  - CyberDildonics 3y ago
    Does that mean a shot is text?