729: Universal Principles of Intelligence (Across Humans and Machines) — with Prof. Blake Richards

Intelligence Norms

The discussion on intelligence reveals its multifaceted nature, with Blake Richards emphasizing the diversity of norms that define different types of intelligence. He explains that intelligence, whether human or artificial, is the ability to adhere to these norms, which vary based on context and goals 1. Richards notes that measuring artificial general intelligence (AGI) is challenging due to its complexity, suggesting that a collection of metrics is necessary to evaluate progress 2. He also highlights the importance of functional mimicry over biomimicry in AI development, using episodic memory as an example of a capability that AI systems need to replicate human performance across tasks 3.

Intelligence is the ability to adhere to the kinds of norms that human beings can and to reach goals and plan out steps necessary to reach goals of the sort that human society demands of an intelligent agent.

--- Blake Richards

This perspective underscores the complexity of defining and measuring intelligence across different domains.

   

Neuroscience Role

Insights from neuroscience play a crucial role in understanding intelligence, as Blake Richards discusses the connection between the hippocampus and reinforcement learning representations 4. He suggests that the brain might use mechanisms akin to gradient descent, a hypothesis supported by physiological evidence, although experimental verification is still pending 5. Understanding the brain's loss functions is pivotal for interpreting neural circuits, as these functions quantify the norms that guide intelligent behavior 6.

Even if we can't identify what every individual neuron in the circuit is doing and how it's contributing to the computation, at least we can answer the question, what was this circuit collectively optimized for?

--- Blake Richards

This approach highlights the potential of neuroscience to inform AI development by identifying the objectives that shape both natural and artificial intelligence.