Two-Point Neurons-Inspired Economic and Ethical Neuromorphic Co-Design (TREND)

A new form of AI chip

As an R&D Creator in the ARIA's Nature Computes Better opportunity space, the TREND project at the University of Stirling aims to develop a radically new class of reliable, economical AI chips incorporating biologically inspired higher-order mental-state dynamics, enabling intelligent agents to make informed, context-sensitive judgments aligned with their users’ needs and values.

Converging evidence from cellular neurobiology, high-resolution computational modelling, and biologically plausible simulations suggests that pyramidal two-point neurons exhibit distinct functional modes associated with slow-wave sleep, wakeful cognition, and imaginative thought. These findings provide a scientific foundation for investigating whether the underlying cellular mechanisms can be translated into AI systems to enable more effective, reliable, and computationally efficient cognitive capabilities.

Going beyond the oversimplified 20th-century neuroscience

Going beyond the oversimplified 20th-century neuroscience concept of pyramidal Point Neurons (PNs)—upon which many current brain theories, AI models, and neuromorphic hardware architectures are based—we aim to unlock the revolutionary computational potential of the recently discovered Two-Point Neurons (TPNs) in the mammalian neocortex, identified as fundamental for learning and information processing in the brain. TPNs have been suggested to be a hallmark of conscious processing (J. Aru et al., Trends in Cognitive Sciences, 2020; J. F. Storm, Neuron, 2024), and their dysfunction has been associated with intellectual and learning disabilities (D. Nelson et al., Developmental Neuroscience, 2021; A. Granato et al., Neuroscience & Biobehavioural Reviews, 2024).

AI at less than 1/1000 of the cost

TREND algorithms aim to achieve AI capabilities at less than 1/1000th of the computational cost of conventional approaches by mimicking both the structure and function of biological two-point neurons (TPNs). Our technology is designed to enable out-of-distribution generalisation and reliable behaviour that current AI paradigms struggle to achieve or require orders of magnitude greater computational resources to attain.

Application Areas:

Robots; Speech; Language; Computer Vision

Funding