A fast spreading adoption of AI in wide-ranging industrial sectors reflects its potential to enhance productivity and economic efficiency. The main technological engine behind AI driven innovation has been a large family of deep neural networks (DNNs), more recently empowered by generative approaches. While generative AI opens up new opportunities for innovation far beyond the industrial domain, AI applications stimulate ever-increasing demands in computing and data among others, resulting in excessive need for resources with unreasonable carbon footprint for industrial deployments. This is particularly challenging for applications with real-time constraints yet requiring functionality, robust adaptivity and operational flexibility, especially in resource limited environments. Current DNNs do not offer a robust and sustainable solution for systems that require low-power consumption, dynamic scalability, adaptation to changing conditions, sustainable scalability, data efficient learning and deployment flexibility in the edge-cloud continuum (ECC). In the attempt to find a manageable compromise between these objectives and requirements, today’s DNNs are also vulnerable to various threats that undermine user trust and reliability. It is therefore strategically important to develop and deploy new core AI technologies that can be more sustainably scaled and utilised in resource constrained application domains, thereby driving the innovation potential of European industries. EXTRA-BRAIN intends to respond to this strategic demand by promoting a brain-like computing paradigm as a technological backbone of the next-generation AI systems. Our mission is to develop brain-like neural networks (BLNNs) building on principles of neural information processing in the human brain and exploiting low-power neuromorphic implementations with the aim of providing AI solutions that are resource efficient, still operationally robust, functionally flexible, trustworthy, and with potential for deployments in the ECC. To further align with the ambition for a sustainable and trustworthy AI the project also integrates application dependent data processing pipelines and explores how different explainability techniques promote trust in EXTRA-BRAIN’s solutions. Finally, to demonstrate the innovative potential, human centricity and real-world applicability of the proposed brain-like AI approach we deploy and systematically evaluate the EXTRA-BRAIN’s AI methods in a range of use case (UC) scenarios spanning digital finance, telecommunication and search-and-rescue (SAR) robotics.
