3 comments on “Neural Networks are rather Transparent

  1. I would say that matching “how humans implement concepts” is insufficiently clear for relying on. We’ve got a whole pile of stupid hacks trying to fix the biases that arise, many of them seemingly directly tied in to our conceptual framework. A slightly different thinker that has the same amount of clarity but distributed differently, such that its thought processes are alien, would be trusted much less than a human.

Comments are closed.