Predictive Processing in brains and machines

Brains@Bay Meetups are designed to bring together the fields of neuroscience and artificial intelligence. Speakers for this meetup have been selected from each discipline to provide unique views of the topic of Predictive Processing.

Modern neuroscience has gained an increasing appreciation that predicting aspects of the environment, like the sensory consequences of action, is a core component of intelligent computation. In recent decades many models, such as Hierarchical Temporal Memory, have surfaced to propose how the brain generates predictions. In this event, our speakers will talk about one family of such models: predictive processing.

Predictive processing states that the brain uses internal models of the world to continually make predictions about events in the environment. These predictions are compared with sensory input to generate prediction errors, which can be used to update the brain's internal models.

The speakers for this event are Avi Pfeffer & Georg Keller.

Avi Pfeffer received his BS in computer science from UC Berkeley and his PhD, also in computer science, from Stanford University. He developed the first general-purpose probabilistic programming language, IBAL, while at Harvard as an associate professor. He is currently chief scientist at Charles River Analytics, where he developed Figaro, a probabilistic programming language in Scala. He is a pioneer in probabilistic programming, statistical relational learning, and several other probabilistic machine learning frameworks. Find him on Twitter at https://twitter.com/avipfeffer

Georg Keller received his PhD at the Institute for Neuroinformatics at ETH Zurich, where he worked on feedback processing of zebra finch song. He finished his postdoc at the Max Planck Institute of Neurobiology in Munich, where he found sensorimotor mismatch signals in the visual cortex of behaving mice. He is a professor at the Friedrich Miescher Institute for Biomedical Research and the University of Basel, where his group has worked on establishing the functional and circuit mechanisms of sensorimotor predictions and prediction errors in the mouse cortex. Find him on Twitter at https://twitter.com/georg98keller

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • -

TALK DESCRIPTIONS:

Building Long-Lived AI Systems Using Predictive Processing (Avi Pfeffer)

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • -

In this talk, Avi will introduce Scruff, a new probabilistic programming language designed for long-lived AI systems that interact with their environment and improve over time. Scruff is based on the cognitive principle of predictive processing, according to which the brain perceives by predicting what it expects to sense and processes errors to produce its beliefs. Predictive processing models are organized hierarchically. In Scruff, each level of the hierarchy is a probabilistic program that generates the level below. Scruff reasons about the state of the environment using an asynchronous belief propagation process and provides three mechanisms for learning, including Bayesian update, parameter learning via gradient descent, and abductive generation of new hypotheses in novel situations.

Predictive processing in cortex (Georg Keller)

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • -

Georg will discuss the framework of predictive processing and a possible implementation in cortical circuits. The evidence he will present for this comes from physiological experiments performed in mouse visual cortex. Finally, he will outline what he believes the computational advantages are of a processing framework that is based on prediction errors.