In Brief
- A team of researchers are working on developing micromote computers that can enhance sensor capabilities while limiting power usage in devices such as smartphones.
- This would eliminate the need to use the cloud for complex analysis, thus increasing the efficiency of the Internet of Things, which is expected to contain 1 trillion devices by 2035.
Small Things, Big Packages
Technology has a habit of shrinking in size while simultaneously growing in functionality year after year. This theme is consistent in the micromote computer products developed by computer scientists David Blaauw and Dennis Sylvester of the University of Michigan.
The duo’s micromote computers are designed to enhance sensor capabilities while limiting power usage on devices. Many of the devices we use today collect information from our surroundings but lack the ability to fully analyze that data on their stand-alone software. Instead, most of these devices transmit the data to the cloud, which Blaauw and Sylvester believe makes devices less energy efficient and less secure. This idea, coupled with the prediction that almost 1 trillion devices will be in circulation within the Internet of Things (IoT) by 2035, led to the creation of the tiny, energy-efficient computer sensors.
Image Credit: University of Michigan/TSMC
The pair also worked with Taiwan Semiconductor Manufacturing Company (TSMC) on making it possible to run artificial intelligence algorithms known as deep neural networks on these tiny computers. Older sensors utilize low-powered SRAM (static RAM), which makes it difficult to optimize the full potential of video and sound. With that said, Blaauw and Sylvester worked with TSMC to equip the micromote computers with flash memory that incorporates an energy-efficient charge pump. While the flash memory is less compact than the SRAM, the added energy savings and boosted memory is worth the trade off.
A Neural Network for Tomorrow
CLICK TO VIEW FULL INFOGRAPHIC
With the added features, these advanced computer sensors could be used to enhance cloud-connected devices with a comprehensive neural network. These neural network-capable devices could incorporate a deep-learning processor that performs facial and voice recognition directly within the device.
For example, a simple thermostat with AI capabilities could notice that guests are putting on their coats. Rather than sending that information to the cloud for analysis, the device would have the hardware and software needed to know it should turn the heat up. According to the creators, this ability to operate outside of the cloud is more secure and efficient, important developments as the number of devices connected to the IoT increases.
Blauuw and Slyvester have a start-up, CubeWorks, that plans on researching and prototyping such devices, and Intel Capital has announced that it has invested an undisclosed amount in the company. While we don’t know if that’s big money the company invested in these tiny computers, any backing from the tech giant is likely to help propel development on these powerful little devices.
