Workshop: We are more energy-efficient than AI

For things to work, their functional parts must first be assembled, and the relationships between those parts must then be maintained as a whole.

AI does this by making a prediction, comparing it with the expected result, measuring the error, and correcting itself. It adjusts billions of internal values, known as weights, through a method called backpropagation. This requires enormous amounts of energy.

The human brain works differently. It separates the functional from the relational. It stores functional information in a way that is similar to AI, but it stores relational information in an entirely different way. Relational connection is not measured or predicted. It is felt and anticipated. This is a far more energy-efficient way to preserve relationships than the approach used by AI. In response to AI’s growing role, we need to become more mindful of how relationships between functions relate as a whole.