Modelling Humans
Kind.AI is a paradigm shift in AI and robotics.
That is a big statement, what justifies it? At present AI technologies can mimic a range of different cognitive and symbolic behaviours, but critically they don’t acquire these skills in the same way humans do. They are either directly programmed or emerge from connectionist methods. While the latter are good at learning, they are not good at symbolic reasoning, and while the former are good at symbolic reasoning, they are not good at learning. Some AI architectures have tried to combine symbolic and connectionist modules, but not in the way humans solve this challenge. As a consequence these technologies are either rigidly narrow or too general to produce meaningful symbolic processing.
Kind.AI takes a different approach. We model the specific processes that enable humans to learn and generalise the core relational skills that underpin cognitive functioning in state of the art AI models and architectures.
Central to our work is Relational Frame Theory (RFT), a Contextual Behavioural Science (CBS) account of language and cognition that has been studied for decades with human beings and other animals. Today we know the processes described by RFT underpin many key cognitive skills, such as general intelligence, executive functioning, perspective taking and theory of mind. By modelling these processes in AI systems, Kind.AI is opening up an exciting new domain within AI and robotics.