4 ms·
>In my opinion, every supervised machine learning model is hopelessly biased by the intent of its creator(s). Namely, it inherits the bias of its training datas
by pseudostem 4y ago
>In my opinion, every supervised machine learning model is hopelessly biased by the intent of its creator(s). Namely, it inherits the bias of its training dataset (both geographic and semantic).
Profound. And True. Sometimes I wonder whether we can truly call them learning models at all.
- bmelton 4y agoBias is learned behavior, so it seems that "learning models" is precisely the right name for it despite whether we considered the ramifications of learning
- campchase 4y agoAuthor here - not an original insight, although it's clichéd enough that I can't point you to where I picked it up from. I also want to emphasize that I do not view bias as a bad thing in the context of supervised models. In some ways, I think it's the whole point of a supervised model (to inherit the judgment of its creators). If the bias helps filter predictions that are useful for your goals, it's a good thing.
- kk58 4y agoWe can walk around bias through self supervision with sampling techniques to select training pairs
- charcircuit 4y agoIt's why they are called models and not just "how it works" or absolute truth.
- Frost1x 4y agoI know several industries filled with questionable researchers who might disagree with you. For clarity, I agree with you for most all cases.
- NoPicklez 4y agoYou can call them learning models, the same way our kids "learn" in environments that are also biased.