In recent years, artificial intelligence (AI) has become increasingly intertwined with our daily lives. However, AI such as that currently supported by most major players in the industry like GAFAM, is decentralized to servers. Since the electricity consumption of Internet infrastructures represents 7% of the world’s entire electricity production and because Internet traffic can be expected to triple every three years, we are in great need of alternative, energy-saving methods of calculation, so that the large-scale rise in AI does not lead to widespread disillusionment. In addition, embedded systems requiring AI are not necessarily permanently connected objects. The need to develop an energy-efficient material substrate for the implantation of AI in nomadic systems is becoming increasingly urgent.
The development of a new hardware substrate must be accompanied by a more ambitious technological solution based on event-based computing. In this promising new computational paradigm, information is created, processed or transmitted only when a change occurs either at the level of the sensor or the calculator. In other words, the system has extremely low consumption if the activity is null. Beyond reducing the amount of data, event-based computing requires fewer operations per second during the inference phase compared to classical artificial neural networks. Both characteristics make event-based computing a promising framework for designing and building energy-efficient hardware for AI.
The GrAI project is disruptive in that it addresses data acquisition (sensors) and data processing (neural network) simultaneously.
The GrAI project will tackle the issue of edge computing, i.e. the processing of data immediately it leaves the sensor. Moreover, spiking neural networks are naturally able to merge data. Therefore, the GrAI project will go further than now established event-based visual sensors thanks to the design and production of other kinds of sensors. This data fusion can be leveraged by industrialists (robotics, monitoring of drivers, autonomous vehicles, etc.), and it involves original technology that is highly suitable for embedded systems that need low-power consumption and low-data transmission. Moreover, GrAI’s AI will not be fed off-line by a large amount of data for learning, as in customary in AI, but rather by on-the-fly input. This ambitious project will provide helpful insights for unsupervised learning, which is one the biggest challenges in AI.
France has set itself the ambition of being a leader in the Green Planet revolution. Therefore, getting the next generation of Masters and PhD students in AI to think in terms of energy sustainability will reinforce the leadership of France in the quest to ‘Make our Planet Green Again’.
Monsieur Sylvain SAÏGHI (LABORATOIRE D'INTEGRATION DU MATERIAU AU SYSTEME)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
IMS LABORATOIRE D'INTEGRATION DU MATERIAU AU SYSTEME
Help of the ANR 401,544 euros
Beginning and duration of the scientific project: August 2020 - 48 Months