Discover more from Zhebrak et al.
The most common mathematical optimisation methods employ randomness (noise) to avoid local minima. Similarly, humans can randomise search and decision-making to improve the experience dynamics. By reading a random book, listening to a random person, or using a random app, we can obtain the momentum needed to escape the current minimum. Too much randomness might compromise the direction and the quality; too little would make no effect.
External machine intelligence tools could greatly help in these scenarios. For example, conversational agents have built-in noise levers we can tune to request the desired level of randomness in their suggestions. By outsourcing the search for directions to synthetic intelligence, we can avoid our internal biases and connect to more objective navigation instruments. However, the boundary between asking for advice and following through is arbitrary. Such a collaboration could indicate the shift towards perceiving and internalising externally suggested ideas rather than developing them ourselves. Soon, ideas presented by synthetic intelligence will be more precise and powerful than human ones, and we will have to decide which aspects of our human nature we want to preserve. Until then, we could schedule self-induced randomness to avoid local minima.