UK Multidisciplinary Centre for Neuromorphic Computing

Context The increasing energy footprint of global ICT infrastructure, amplified by the explosive growth in AI technologies is unsustainable, and impacts negatively on the goal of achieving net-zero CO2 emissions. Moreover, there is clear realisation that current computing architectures are organised sub-optimally. In contrast, the human brain is highly efficient in carrying out complex task while consuming about 20W. This has led to increasing interest in "brain-inspired", energy-efficient neuromorphic computing, aiming to utilise brain-like processing mechanisms in new computational technologies. However, a fundamental problem is that it is not fully understood how the brain "computes" at the cellular or network level. This deficit is largely due to historical lack of data from controlled experiments on human neurons. The challenge the project addresses The fundamental challenge we address is to understand the mechanisms and principles of human brain function from neuron to network to develop neuromorphic computing that better emulates human brain function. Most neuroscience knowledge is from the study of non-human, primarily rodent brains and tissue. However, recent rapid development of human induced pluripotent stem-cell (hiPSC) technology, enables us, for the first time, to reproducibly study human neuronal function at cellular, synaptic and network scale. Our radical vision for the Centre for Neuromorphic Computing is to combine co-designing experiments on hiPSC derived networks to determine human neuronal properties and mechanisms, relevant for establishing the foundation of next generation neuromorphic computing, with developing and integrating advanced computational models, materials, and devices. This framework will coherently unite experts across disciplines: neuroscience, material science, photonics, semiconductors, computer science and mathematics. Our focus will be on creating an ecosystem facilitating the generation of new knowledge, advancing the field and applying the developed technologies to real-world problems. The challenges of neuromorphic computing will be tackled holistically, where experimental design, hardware, analysis and algorithmic expertise will go hand in hand, with the ultimate goal to unlock the complex processing capabilities of the human brain and apply them to transformative new computing technologies. Aims and objectives The project will create a new UK brand: UK Centre for Neuromorphic Computing. Its objectives during the initial 4-year period: 1) To implement a robust functional and structural framework for the Centre that ensures its sustainable long-term operation beyond the project lifetime through the support of stakeholders, industry, international projects and networks. The Centre will be inclusive to all appropriate UK groups across relevant disciplines. 2) Design and develop new foundational computational concepts and advanced neuromorphic algorithms adapted to the challenges of low power consumption, and high efficiency, including efficient training algorithms. 3) Develop novel hardware implementations ranging from human-neuronal-based cortical devices to material substrates; advance and enhance existing computational concepts linked to neuromorphic computing that could be implemented using alternative technologies. 4) Implement a programme of annual workshops, training events and summer schools aimed to create a UK ecosystem where academia, industry and policy makers work together and exchange ideas, from fundamental concepts to industry-oriented research and to finding solutions of global problems. 5) Identify existing and emerging industrial applications with emphasis on strategic features (energy efficiency, decentralized computing, speed, communication efficiency), to critically demonstrate and exploit the economic and societal relevance of neuromorphic computing, specifically by addressing emerging applications. This will facilitate the smooth transition of innovation to industry through direct involvement of stakeholders from academic groups to companies.