Next step in AI: Chip-based artificial synapse that can learn autonomously

Chip-Based Artificial Synapse That Can Learn Autonomously-1

The field of artificial intelligence (AI) has seen massive advancements in recent years, most notably pertaining to technologies that replicate the workings of the human brain. Referred to as artificial neutral network (ANN), this kind of AI system makes use of algorithms that can in turn be trained to process information, such as speech or images. Currently used by Facebook for automatic image recognition, such processes however are known for their very high energy consumption. Dr. Vincent Garcia of CNRS explains:

There are many breakthroughs from software companies that use algorithms based on artificial neural networks for pattern recognition. However, as these algorithms are simulated on standard processors they require a lot of power. Developing artificial neural networks directly on a chip would make this kind of tasks available to everyone and much more power efficient.

Recently a team from Thales-based National Center for Scientific Research (CNRS) joined hands with Norwegian IT company Evry, and researchers from the University of Bordeaux, to develop an incredibly futuristic artificial synapse on a chip. According to the scientists, the technology known as memristor could greatly increase the learning speed of AI systems, and even enable them to learn autonomously.

A synapse, in case you are wondering, is a special structure in the human nervous system that serves as a connection between two neurons, allowing electrical (or chemical) signals to pass from one to the other. With increased synaptic stimulation, our learning capacity also improves over time. The newly-created device, containing a ferroelectric layer held between two electrodes, attempts to do the same thing. The resistance of the two electrodes can be manipulated with the help of voltage pulses, which in turn improves the synaptic connection. The latter, the researchers state, is stronger at low resistance and, vice versa. Garcia goes on to say:

Here we use a specific kind of memristors based on purely electronic effects. In these devices, the active part is a ferroelectric film, which contains electric dipoles that can be switched with an electric field. Depending on the orientation of these dipoles, the resistance is on or off. In addition, we can control configurations in which domains with up or down dipoles coexist, with intermediate voltage pulses. This gives rise to an analog device with many resistance levels. In our paper, we were able to understand how the resistance of the memristor evolves with voltage pulses and make a model based on the dynamics of ferroelectric domains.

As explained by the team, the resistance also determines the memristor’s learning capacity. Recently published in the Nature Communications journal, the breakthrough could in fact pave the way for AI systems with autonomous learning capabilities. Garcia adds:

The final goal of this project would be to integrate this bio-inspired camera in a car to assist the driver when unexpected objects or persons are crossing the road.

Source: National Center for Scientific Research (CNRS)

Via: Digital Trends

  Subscribe to HEXAPOLIS

To join over 1,200 of our dedicated subscribers, simply provide your email address:


ps_menu_class_0
ps_menu_class_1
ps_menu_class_2
ps_menu_class_3
ps_menu_class_4
ps_menu_class_5
ps_menu_class_6

Next step in AI: Chip-based artificial synapse that can learn autonomously

Chip-Based Artificial Synapse That Can Learn Autonomously-1

The field of artificial intelligence (AI) has seen massive advancements in recent years, most notably pertaining to technologies that replicate the workings of the human brain. Referred to as artificial neutral network (ANN), this kind of AI system makes use of algorithms that can in turn be trained to process information, such as speech or images. Currently used by Facebook for automatic image recognition, such processes however are known for their very high energy consumption. Dr. Vincent Garcia of CNRS explains:

There are many breakthroughs from software companies that use algorithms based on artificial neural networks for pattern recognition. However, as these algorithms are simulated on standard processors they require a lot of power. Developing artificial neural networks directly on a chip would make this kind of tasks available to everyone and much more power efficient.

Recently a team from Thales-based National Center for Scientific Research (CNRS) joined hands with Norwegian IT company Evry, and researchers from the University of Bordeaux, to develop an incredibly futuristic artificial synapse on a chip. According to the scientists, the technology known as memristor could greatly increase the learning speed of AI systems, and even enable them to learn autonomously.

A synapse, in case you are wondering, is a special structure in the human nervous system that serves as a connection between two neurons, allowing electrical (or chemical) signals to pass from one to the other. With increased synaptic stimulation, our learning capacity also improves over time. The newly-created device, containing a ferroelectric layer held between two electrodes, attempts to do the same thing. The resistance of the two electrodes can be manipulated with the help of voltage pulses, which in turn improves the synaptic connection. The latter, the researchers state, is stronger at low resistance and, vice versa. Garcia goes on to say:

Here we use a specific kind of memristors based on purely electronic effects. In these devices, the active part is a ferroelectric film, which contains electric dipoles that can be switched with an electric field. Depending on the orientation of these dipoles, the resistance is on or off. In addition, we can control configurations in which domains with up or down dipoles coexist, with intermediate voltage pulses. This gives rise to an analog device with many resistance levels. In our paper, we were able to understand how the resistance of the memristor evolves with voltage pulses and make a model based on the dynamics of ferroelectric domains.

As explained by the team, the resistance also determines the memristor’s learning capacity. Recently published in the Nature Communications journal, the breakthrough could in fact pave the way for AI systems with autonomous learning capabilities. Garcia adds:

The final goal of this project would be to integrate this bio-inspired camera in a car to assist the driver when unexpected objects or persons are crossing the road.

Source: National Center for Scientific Research (CNRS)

Via: Digital Trends

  Subscribe to HEXAPOLIS

To join over 1,200 of our dedicated subscribers, simply provide your email address: