Ground C - Artificial intelligence, maths and modelling

Cognition and learning involve processes operating all the way from the molecular level to large networks of interconnected neurons, spanning nine orders of magnitude and involving electrical, chemical, and mechanical signals. To begin to understand how these complex processes operate requires interdisciplinary teams bringing skills and techniques from fields such as neuroscience, psychology, psychiatry, artificial intelligence, statistics, imaging, genetics, engineering and mathematical modelling.

The physical sciences have their foundations firmly based in mathematical modelling. And modelling has facilitated many advances including enabling the development of complex systems that exhibit levels of artificial intelligence. For instance, in electrical engineering, multi-scale modelling techniques permit us to model and build complex systems of transistors, acting in unison, which can beat humans at games such as 'Jeopardy'. These multi-scale models include models of transistors, nanoscale devices of a few thousand atoms operating at the quantum mechanical level; to circuits, comprising a few hundred to a few thousand transistors, that aggregate and process signals to produce complex outputs; to systems of circuits that interact with their environment (including humans) and generate higher order outputs.

The group will discuss questions such as:

• Are multi-scale mathematical models, used effectively in other disciplines, applicable to neuroscience and can they be used to provide new insights into the neurosciences and complex processes such as cognition and learning?
• Are the mathematical models and software and hardware description languages sufficiently rich to enable us to model complex and multi-scale processes such as cognition and learning? For example, how can the knowledge from fields such as psychology, neurology and psychiatry be represented as part of mathematical models of neuronal and brain function?
• To produce adequate models, with predictive power, requires robust and reliable data from multiple modalities. Do we have the right types of data at the requisite quality? If not, what other data and instruments do we require?
• What questions would these new data and instruments help us answer?
• Gathering data is time consuming and expensive. What should be the strategies to bring together the available multi-modal data to make it available to the research community?
• The fields of artificial intelligence and neuroscience have made phenomenal progress. What can the fields of neuroscience and artificial intelligence learn from each other?

Chair

• Professor Stan Skafidas
• Professor Steve Furber

Rapporteurs

• Dr Andreas Fouras
• Dr Elizabeth Holliday

Participants

• Dr Andrew Barron
• Dr Girija Chetty
• Dr Alexandre Cristino
• Dr Mathew Dafilis
• Dr Vincent Daria
• Dr Yulin Ding
• Dr Jurgen Fripp
• Dr Jean Yee Hwa Yang
• Dr Emily O’Brien
• Dr Sivalogeswaran Ratnasingam
• Dr Karen Siu
• Professor Murat Yücel
• Dr Andrew Zalesky