​​​​Knowledge Engineering Consulting is honoured to have been approached by UK Startup Conscium, the world's first commercial research organisation investigating the implications of building Machine Consciousness.

 

As part of the journey into Machine Consciousness, Conscium and Knoweldge Engineering are collaborating to develop a brain-inspired explainable, self-adaptive AI framework that enables conscious perception of stimuli.

​​​

​

Background

​

Contemporary AI systems rely on static, pre-learned models that lack dynamic adaptability and conscious interpretation of meaningful stimuli.

 

In contrast, the human brain continuously constructs internal models of reality, updating its understanding through experience and associative learning, building moral values.

 

The project develops a brain-inspired explainable AI framework, called eXCube, using novel Spiking Neural Networks (SNN) and Evolving Spatio-Temporal Associative Memories (ESTAM), that enable conscious perception of stimuli, self-adaptive and context-aware learning and decision-making.

​

The Integrated Information Theory (IIT) is considered a fundamental theoretical basis for consciousness, suggesting the presence of features such as intrinsic existence, compositionality, information, integration, and exclusion. But there is still something missing.

 

The hypothesis is that the missing part is the identification of an evolving spatio-temporal associative memory (ESTAM) that evolves in space and time, when conscious processes take place. The project further develops the concepts of ESTAM to define, for a certain modality of stimuli, generic, dynamic patterns in brains and machines regardless of the concrete content of the stimuli.

 

And this is demonstrated on the problem of defining generic patterns of meaningfulness using visual and

audio information to illustrate this approach.

​

​

Project Objectives

​

The project introduces the eXCube framework and showcases its use in two case studies, revealing brain patterns in fMRI data linked to the perception of meaningful and meaningless visual and auditory stimuli.

 

Experiments on benchmark data show that eXCube can detect and explain such patterns.

​

The challenges that the project addresses are divided into two phases, both making use of a brain-inspired SNN architecture NeuCube:

 

1. Revealing generic spatio-temporal patterns/signatures in the human brain associated with meaningfulness.

​

2. Making machines that identify the meaningfulness of stimuli, regardless of the concrete content of these stimuli.

​

​

​

Long-term Study Objectives

​

Long-term study objectives include:

​

1. Understanding the mechanisms of conscious perception in the human brain through brain data modelling.

​

2. Developing biologically inspired AI models that replicate this process.

​

3. Developing AI software agents capable of ethical, moral, and context-sensitive decision-making, with applications in robotics, healthcare, and assistive technologies.

 

This research could shift how AI machines interact with people and the world, making AI safer, more efficient, more ethical, and more aligned with human values.

​​​​​