[avinium]

I strongly prefer they keep the bias. It shows plainly that these models have zero intelligence, they're just pattern matching over specific datasets.

I don't want the appearance of fairness (introduced by human dataset curation) to be mistaken for "intelligence".

Keeping the bias would hopefully cause people to think more critically about why such bias exists in the model in the first place.

[jobigoud]

Disagree that it proves zero intelligence. Intelligent humans also have bias.

[avinium]

I'm not saying that intelligence and bias are orthogonal. That discussion requires a much deeper consideration of human cognition and psychology :)

I'm just saying that model bias is a very easy thing to explain (usually, data imbalance).

You can also fiddle with the label ratios to change the bias - which is also a good way of showing that the models aren't really intelligent.