Skip to main content

From machine learning to machine psychology: Artificial general intelligence from the perspective of non-axiomatic logic

Save to calendar

Feb 25

We are happy to present Robert Johansson, PhD, is an associate professor of psychology at Stockholm University, and a researcher in computer science at Linköping University. He is also a licensed psychologist with a special interest in emotion-focused psychotherapy.

Date and time: 25 February 2021, 12 pm – 1 pm
Speaker: Robert Johansson
Title: From machine learning to machine psychology: Artificial general intelligence from the perspective of non-axiomatic logic
Zoom: https://kth-se.zoom.us/j/67432682790?pwd=dVgzbjRSbUVFT2FOYTByYlZrTU9BUT09
Meeting ID: 674 3268 2790
Password: DF2020

Watch the recorded presentation:

 

Photo of Robert JohanssonAbstract: Recently there has been substantial theoretical progress in artificial intelligence research, regarding the field of AI research that aims to build machines with human-level capabilities. This field is commonly referred to as Artificial General Intelligence (AGI) and studies models of AI that could perform at human-level on various tasks across different domains.

One approach to AGI is Pei Wang’s Non-Axiomatic Logic, with its implementation NARS (Non-Axiomatic Reasoning System). This model is based on the assumption that intelligence is the ability of a system to work and adapt with insufficient knowledge and resources. The model has recently been used to approach real-world problems like enhanced autonomy in robots and applications in the smart-city domain.

A recent NARS implementation is called “OpenNARS for applications” (Hammer and Lofthouse, 2020) which allows sensory input from multiple modalities in integration with semantic and sensorimotor inference, and procedure learning. In fact, this NARS model allows any type of typical neural network-based classifier, like a Multi-Class Multi-Object Tracker, to be used as sensory input. This solution opens up for a nice way for a NARS system to learn from direct sensory experience and infer new knowledge using semantic reasoning.

Within the behavioral psychology tradition, there is a strong empirical basis for how learning can happen over repeated interactions between organisms and environment, and how learning in one domain can transfer to another. This theoretical account seems highly relevant to current problems faced in contemporary AI research.

In this talk, we will argue how behavioral psychology provides a roadmap for AGI research, especially in systems where sensory channels need to be integrated with semantic knowledge and procedure learning.

Bio: Robert Johansson, PhD, is an associate professor of psychology at Stockholm University, and a researcher in computer science at Linköping University. He is also a licensed psychologist with a special interest in emotion-focused psychotherapy. In his earlier work, he and his colleagues developed a model of affect-focused psychotherapy that enabled it to be delivered as guided self-help through the Internet. The effectiveness of the model has been proven in several clinical trials. Currently, he is using a behavioral psychology framework to guide his research in how NARS can model clinically relevant psychological processes in machines.