3 ms·
what if the models are used to generate a new training set ?
by seydor 4y ago
what if the models are used to generate a new training set ?
- visarga 4y agoAbout 150m images have been generated with SD so far. That's already a new large scale training set. Generate and curate before retraining to create a virtuous cycle.
- pmoriarty 4y agoThis looks more like polluting the training set to me.
- fredoliveira 4y ago> That somehow international legislation will converge on the strictest possible interpretation of intellectual property, and those models will become illegal by the mere fact they were trained on copyrighted material. Doesn't this ultimately result in local maxima? All the biases get reinforced and all the novelty (things the system hasn't seen/produced yet) goes away. A tiny example: Dall-E (and SD) both struggled with eye positioning, for example. Wouldn't training a model on their output then reinforce that particular bias of poorly positioning eyes? Now multiply this by every existing quirk in the models.