At the recent AWS re-invent conference 2019 Amazon demonstrated DeepComposer, a music program that can take a few hints from your melody (you basically play a tune on a keyboard, even just a few notes) and use that hint to compose a structured musical piece. The result is a brand new composition (that you can use as soundtrack for your slide show, with no issue of copyright infringement!) that someone may take as an example of musical creativity.
Now, for a few that would take this as an example of computers, AI, demonstrating some traits of human creativity there will be (many) others that would strongly say that is no creativity at all.
The opposers reasoning is usually based on the observation that a robot/a computer/a software is basically a program that some humans wrote (actually today more and more programs are written by programs… but in the end there is an initial program that was written by a human being). Hence, what is demonstrated is not creativity springing out of nowhere but the result of human creativity that a machine translated using a well defined algorithm.
This reasoning is rooted in the untold hypotheses that we, humans, are not “programmed”. This is a debate we have had in several discussion in the Autonomous Systems Initiative now renamed into Digital Reality. But is this hypotheses a valid one to assume (you cannot say it is true or false, it is an hypotheses you take as the starting point of your reasoning)? Aren’t we, humans, working through the expression of a code (A-C-G-T in the DNA) which is the result of billions of years of evolution and an unaccountable gazillion casual variations constrained through natural selection to what we have today?
AI is using a sort of similar approach. It starts from a relatively tiny set of instructions that are guiding it to observe the world and learn from it and from mistakes or beneficial outcome through a reinforcing learning process. You feed a machine with a chest radiography pointing to an area where a tumour is present. Then you submit hundreds, thousands, millions of other radiographies correcting the machine interpretation till you reach a point that the machine can be as good as you, actually better, in detecting a tumour presence. Aren’t we, humans, using a very similar approach to learning the meaning of the world around us? How many times a toddler throw something on the floor before finally “understanding” that if you through something that something will tend to move downward eventually resting on the floor? That even if you are not longer seeing it it should be there somewhere because it “exists” and it cannot vanish?
It is becoming more and more difficult to draw a line between our (pretended) smartness and creativity and the one shown by machines.
Besides, the very fact that we are starting to debate if AI, computers, robots, are showing creativity capabilities is a clear indication that we have reached a fuzzy zone where our defence of human capabilities is shifting to debating what is creativity and what is not. A clear sign that we feel somehow threatened and we are in a defensive mode. I personally feel that rather than debating on what creativity really is and trying to explaining why AI creativity is not really creativity we should work on finding way to embrace these growing AI capabilities to keep enhancing ours.
We have stopped long time ago, as a species, to rely on biological evolution to become better, our evolution is now strongly dependent on cultural and technological evolution. So in a way there is nothing new. It is just that culturally, and technologically, we are getting closer to a thresholds where we can improve our species beyond what used to be impassable thresholds.