Bojana Grujičić (Science of Intelligence), “Artificial possibilities”

Science often deals with issues pertaining to possibilities, contingencies and necessities, by engaging in thought experiments and modeling. This talk discusses how much deep learning can be helpful for navigating the possibility space for intelligence, adding to our scientific understanding of possibilities. One epistemically useful feature of neural networks is their runnability – they can be trained to perform a cognitive task and can run when given novel stimuli, demonstrating possibilities of cognitive phenomena based on sets of inductive biases. I focus on the problem of justification of neural network-based inferences about possibilities and outline a plausible justificatory strategy. I consider a number of reasons for taking neural network-demonstrated possibilities to be technological, rather than biological possibilities. Despite this, I argue that they add to our scientific understanding of possibilities related to intelligence.

Photo by Joakim Honkasalo on Unsplash.