His position, from what I understand, is that the ML algorithm which reconstructed a photo of Obama as white was biased solely because the training data set consisted of mostly white people. If it had been a more diverse data set then the algorithm would have behaved more appropriately.
On the one hand this guy was pushed off twitter for a seemingly innocuous comment - one that some people just happened to take as insensitive. The type of comment that, in fact, increasingly large numbers of people seek out to take offense at - so it seems like an overreaction. On the other hand this is twitter, exactly where these types of narratives play out, he was unable to defend his position competently or even acknowledge the criticisms against him, gave up, and walked off.