"Efficient Learning in AI" by Dr. Rachel St.Clair

ABSTRACT

Various interpretations of the literature detailing the neural basis of learning have in part led to disagreements concerning how consciousness arises. Further, artificial learning model design has suffered in replicating intelligence as it occurs in the human brain. Here, we present a novel learning model, which we term the “Recommendation Architecture (RA) Model” from [Coward, 2004], using a dual-learning approach featuring both consequence feedback and non-consequence feedback. The RA model is tested on a categorical learning task where no two inputs are the same throughout training and/or testing. We compare this to three consequence feedback only models based on backpropagation, reinforcement learning, and a combination of the two. Results indicate that the RA model learns novelty more efficiently and can accurately return to prior learning after new learning with less computational resources expenditure. The final results of the study show that consequence feedback as interpretation, not creation, of cortical activity creates a learning style more similar to human learning in terms of resource efficiency. The work provided here attempts to link the neural basis of nonconscious and conscious learning while providing early results for a learning protocol more similar to human brains than is currently available.

BIO

Rachel St.Clair, PhD., is a scientist, entrepreneur and author. She is the founder and CEO of Simuli Inc. Her passion and goal is to help build beneficial AGI. Rachel envisions an AGI as a computer system which can outperform human abilities in nearly all tasks while helping humanity avoid existential risks. In her pursuit of developing AGI, she then founded the company Simuli to develop the necessary hardware to build her preferred system and build hardware for other AGI pursuits. Rachel relates the ability of processing large amounts of information over time without losing prior information as a key component of resource management, a driving force of practical AGI systems. Amongst her role in Simuli, Rachel continues to be an active scientific member of the AGI community by publishing and collaborating in the field. Her current work is focused on learning algorithms which emerge from resource constrained systems, drawing inspiration from neuroscience models such as The Recommendation Architecture by Andrew Coward and mathematical approaches in complex systems. When Rachel’s not working on AGI, which she reports as being hardly ever, she likes to go whale watching, write and read science fiction novels, and develops cartoons and video games.