Wednesday, January 10, 2018

In a way, deep-learning systems could be seen as living in a Groundhog Day movie implemented: given (almost) unlimited chance to try out different choices and learn their consequences, as well as (almost) unlimited memory, design the most optimal policy for a given environment (which could be also seen as a very sophisticated set of reflexes).

Which makes one wonder:

1) whether conscious thinking, and the human civilization as a consequence of that, have developed as an evolutionary shortcut to mitigate the shortage of experience which could be obtained naturally by a single human being.

2) whether the "set of reflexes" approach can scale to emulate human conscience and not turn the Universe into paperclips in the process.