3 ms·
One thing I wonder is what happens after AI starts getting trained on its own content? Its content is already hitting the places the next round will be trained
by jalapenos 4y ago
One thing I wonder is what happens after AI starts getting trained on its own content?
Its content is already hitting the places the next round will be trained on.
Any AI experts here, how does that play out? Does it pollute, poison, or cacophonize?
(And yes I got "cacophonization" from GPT4)
- m463 4y ago42
- sinuhe69 4y agoAt worst, it’ll be kind of infinite recursive with psychopathic tendency. If more lucky, the law of dismissing return will strike unmercifully.
- afro88 4y agoMy guess is it'll be like feedbacking audio. If you route some audio through a processor and set it up in a feedback loop, the way the processor changes the audio gets accentuated more and more as you allow it to feedback more. Weights will be reinforced and accentuated and the model will become more extreme and probably less useful.
- Karrot_Kream 4y agoThat only happens in audio feedback because audio amplification is an increasing function. ML models do complicated things with their inputs.
- jamilton 4y agoI think training on output is one method that people use to create customized Stable Diffusion models. IIRC, it doesn't work well unless you manually label it, in which case it works fine. Letting existing AI label it doesn't do very well... but maybe GPT-4 will be good enough at that sort of task?