Professor Jordi Suñé, guest editor of a special issue of ‘Materials’ dedicated to memristors for applications in AI and neuromorphic circuits

May 6, 2018 | Author: neuromimeTICs

The applications of artificial intelligence and neuromorphic circuits built from memristors and CMOS circuits will be the focus of the next Special Issue of Materials, which will include articles about the state-of-the-art of deep learning and hardware-based neuromorphic systems without neglecting the ethical aspects that surround the unstoppable advance of the AI.

Cover artwork: Jaime Moroldo

Cover artwork: Jaime Moroldo


Machine learning is impacting our society in every corner. It has been introduced in areas as diverse as medical applications, the automotive industry, the global climate forecasting and prevention, the management of emergencies or the prevention of terrorist attacks, deep learning based AI has practically become an ubiquitous technology in everyday life. It is part of the so-called artificial intelligence (AI) revolution, unstoppable and at the same time somehow disturbing.

Given this panorama of present and future, Materials has programmed a Special Issue to deal with the hardware implementation of AI applications with memristor-based neuromorphic circuits.  Dr. Jordi Suñé, professor of Electronics Engineering at the Universitat Autònoma de Barcelona (UAB) and coordinator of neuromimeTICs, has been appointed guest editor of this Special Issue, which is entitled "Memristors for neuromorphic circuits and applications in artificial intelligence".

Software and hardware based technologies

The AI s a ubiquitous technology that, at present, is mainly implemented in software, although since 2008 it can also be implemented in hardware. This has been possible thanks to the development of a nanoelectric component, the memristor, from which neuromorphic electrical circuits can be built. The existence of the memristor was predicted by Professor Leon Chua in 1971 but it was not until 2008 that the HP group led by Dr. Stanley Williams managed to implement the memristor in solid state for the first time. The availability of this component, the memristor, has opened new frontiers for the development of AI applications, the so-called deep learning ICs.

These neuromorphic systems are not widely known by the general public, and they systems that mimic the neural networks of the brain, using memristors to implement synapses and CMOS circuits to play the role of neurons. "Being hardware-based, deep learning ICs will enable the distributed deployment of energy efficient IA in many areas that require real-time response, intelligent decision and quick action. In this special issue “we will try to offer an up-to-date overview of this new technology", explains Professor Jordi Suñé, as the guest editor of Materials.

Technical and ethical issues

"We will review the concepts of machine learning and deep learning, with a focus on their applications. We will cover the state-of-the-art of technological implementation of memristor electron devices with particular emphasis on resistive devices (both ReRAM and PCM). We will also present the current state of the art of deep learning prototypes based on memristor for different applications. Finally, we will dedicate some articles to ethical questions related to AI and also to neurosciences ", explains Professor Suñé.

Both technologies, deep learning ICs and hardware-based neuromorphic systems, have many potential applications that will help to address global social challenges (AI for Good). However, "there are also many serious ethical challenges to face, since these technologies also have many potentially deleterious applications (AI for the Devil) that could even put the future of humanity at serious risk," warns the neuromimeTICs coordinator, quoting Prof. Nick Bostrom, founder and director of the Future of Humanity Institute of Oxford University.