China develops energy-efficient AI microchips
Chinese scientists have developed what they claim are the world's most energy-efficient artificial intelligence (AI) microchips . Small enough to fit into smart devices, such as mobile phones, AI chips could open the door to innovative offline functions, such as voice control and mind control.
AI chips designed for demanding tasks often require significant power due to high computational demands, which limits their use in real-world scenarios, especially when it comes to mobile instruments, which therefore rely on energy provided by batteries. Professor Zhou Jun and his team at the University of Electronic Science and Technology of China (UESTC) have significantly reduced power consumption through optimization of algorithms and architecture.
The team presented two AI chips at the IEEE's 2024 I nternational Solid-State Circuits Conference (ISSCC) . The ISSCC, in case you don't know, is considered the “Olympics” of the integrated circuit (IC) industry.
The conference is an annual event that brings together top researchers, engineers and professionals from around the world to discuss the latest developments in solid-state circuits and the future of chip technology. This year it was held in San Francisco from February 18th to 22nd.
Ultra-low power consumption AI chip
The first of the two AI chips was designed to be incorporated into smart devices to allow offline voice control, therefore without an internet connection that allows external voice recognition. This chip, according to the researchers, excels at keyword detection and speaker verification by recognizing speech signals from a target speaker.
The main advantage of this new AI chip is its ability to overcome the constraints of standard speech recognition systems. Unlike others, this chip can accurately recognize a speaker's speech even in noisy situations with distractions such as television, music or conversations.
“The chip achieves recognition power consumption of less than two microjoules per instance of use, with an accuracy rate greater than 95 percent in quiet situations and 90 percent in noisy environments, setting new global benchmarks for both energy efficiency that to be precise,” the researchers explained.
In a demonstration of the system, a chip, measuring 1×2 cm, was integrated into a microcontroller unit inside a toy car to control its movements. The AI chip can presumably be used in low-power voice control scenarios, such as in smart homes, wearable devices and smart toys.
The team's second chip, also presented at the ISSCC conference, is designed to detect seizure signals in people with epilepsy. Designed for wearable devices, this technology uses electroencephalogram (EEG) recognition to detect seizures and alert the patient for medical assistance or treatment. This would be an essential chip in the treatment of people suffering from this important neurological disease.
It could be used for brain-computer interface
“Existing projects rely on extensive patient seizure data for training to achieve high accuracy, a process that is time-consuming and expensive due to the infrequency of seizures and the need for hospitalization,” they explained the researchers.
Thanks to improved feature extraction and on-chip learning engines, this chip has the most energy-efficient design internationally, with an average power consumption for recognition of just 0.07 microjoules.
During a demonstration at the ISSCC, as reported by the South China Morning Post (SCMP), EEG signals collected by a wearable brain-computer interface device were transmitted in real time to a test board using Bluetooth technology.
However, this tool can also have other applications beyond the strictly medical ones: the chip was reprogrammed to recognize imaginary motor commands, which allowed the user to control the movement of a robot. The robot could be commanded to move forward, to stop or to move in reverse, demonstrating that, without the need for subcutaneous interventions as in the Neuralink, it is still possible to control simple mechanisms with thought.
“This chip also has potential applications beyond seizure detection. the researchers stated, and indeed the potential of this new AI chip is still to be explored.
Thanks to our Telegram channel you can stay updated on the publication of new Economic Scenarios articles.
The article China develops energy-efficient AI microchips comes from Economic Scenarios .
This is a machine translation of a post published on Scenari Economici at the URL https://scenarieconomici.it/la-cina-sviluppa-microchip-per-ai-ad-alta-efficienza-energetica/ on Tue, 26 Mar 2024 10:30:21 +0000.