What happens when you simulate an algorithm?

Abstract:

This talk explores the philosophical implications of simulating algorithms, particularly in the context of artificial intelligence and its relationship to biological cognition. I begin with a brief examination of the nature of simulating (implemented) algorithms, with emphasis on one particular difference that is relevant to distinguishing between simulating algorithms as opposed to simulating the outputs of algorithms. Then, I pull select pieces of evidence from two distinct fields (biological neuroscience and mechanistic interpretability literature within deep-learning) to show how recent studies have found implemented algorithms in both biological and artificial brains. Finally, I conclude with an invitation to consider how these results might inform conversations on the moral status of AI, the measurement of consciousness and sentience in AI, AI well-being, and long-term AI risk.