The Creativity Code: Order out of Chaos [part 3]

Visual programming poses itself as a serious enigma to engineers. How could we teach an AI to identify visual patterns and decode images? The human brain has an amazing ability to process visual information. Rorschach inkblots testify to the level of creativity that can be achieved through visual projections. It is one of the most impressive evolutionary adaptations that ensured our survival in the past. “Any animal’s ability to survive depends in part on its ability to pick out structure in the visual mess that Nature confronts us with” (Du Sautoy, M., 2019).

The very same mechanism that was responsible for our ability to identify, respond and predict structural changes in sensory data within a hostile environment, allows us to engage with our higher mental faculties and develop an appreciation for the arts and the sciences.

Games like chess are too restrictive, their rule-systems are too rigid and so they fail to level with our more broad and complex ability to recognize patterns. It is the surveyability of a situation on an intuitive level that accounts for real creativity. A split-second apprehension. How could an AI develop such creativity? The answer has to do with our previous discussions on failure and something called Reinforcement Learning.

RL is a programming system that allows an AI to learn from experience through repeated failures. By evaluating and re-evaluating its system of rules based on successive achievements/failures, the AI creates an automatic feedback loop of reinforcements and rule-overrides. RL allows order to emerge out of chaos. The system starts out with random behavior until it achieves a level of organization and predictability. This is exactly how living systems operate. That’s pretty much the creativity code.

REF

Du Sautoy, M. (2019). The Creativity Code: How AI is learning to write, paint and think. HarperCollins UK