Comments

You must log in or register to comment.

londoncalling wrote (edited )

Several of the new deep learning tricks are done using a model where two AIs are created called a generative adversarial network; one which creates the deepfakes and one which guesses if its input is real or generated by the former; and the two are used to train each other. It would be very useful if the second type of AI was published, as it could probably detect deepfakes better than a human reading off a list of features to check.

4