3 ms·
Isn't the "why" the weights of the model themselves?
by ewuhic 4y ago
Isn't the "why" the weights of the model themselves?
- sdenton4 4y agoPeople want brief explanations of reasons for outputs, and a 10gb pile of weights isn't really what they mean. Human explanations meanwhile are often invented to fit evidence to personal biases and beliefs, and are thus typically deeply flawed. But we're more ok with humans making suspect decisions than ML, in many cases.