529 karma · joined May 31, 2022
There is a reason those words are the ones chosen by racists in the first place. Not just because they're hurtful, but because there is an objective similarity that makes the rest of the comparison seem, however shallowly, to validate their views.
Removing bias with censorship at the training set level is silly, and will likely hamper the AI's performance. Better to train the AI to not produce problematic output at the higher level, and ensure class membership in the training data is representative.
If they received push back from dogfooding, that just means they need to knock those sharp corners off. I'd suggest what they want is a meeting application and making VR easy to put on and off, and combining the two worlds.
It's entirely realistic to dogfood here, people are just reacting to the obvious reactions from the first iteration. You don't stop there though.
On the other hand, he died so...
The first layer consists of 1 node per input pattern, with activation and weights such that only one node is active that matches the pattern. And a second layer of weights that provides the output per input pattern.
Especially since the weights are tied entirely to how well calibrated the individual predictive models are.
Social media is a punishment for those to stupid to avoid it.
Wait a minute... You're trying to tell me how to live my life!
It'll work regardless. All it needs is X artist uses this medium, does these sorts of pictures, looks like artists Y & Z, etc. etc.
SD does okay on some artists that don't have works in the image dataset. All their name does is pick out a position in the embedding space.