Home / Blog / Pervasive AI target the edges

Pervasive AI target the edges

This tiny chip has been designed to embed AI in edge devices, like cars. Image credit: Hailo

Artificial Intelligence started through massive number crunching and entangled rules (I still remember Prolog…) requiring big computers to support it. Over time the “rules” based approach shrunk a bit with number crunching taking the upper hand (as both more data and more processing power become available). This increased the use of powerful computers restricting the number of players. That generated a sort of paradox: as AI was becoming more and more useful, applicable to larger and larger sectors, the required support became more and more demanding, out of reach of most companies.

The solution was (and is) the Cloud. There we can tap on data crunching resources, accessing it from anywhere and returning results everywhere. At this point the bottleneck was no longer the processing power, rather the connectivity infrastructure. In most situations the connectivity infrastructure is available, reliable and sufficiently powerful to leverage on a cloud based AI.

The situation gets trickier if wee need AI in devices that cannot be connected by a fixed line, such as the case of several robots in industrial environment, of vehicles, of wearables … Here the wireless infrastructure comes to the fore (WiFi 6, private 5G in industrial environment, 5G, 4G, … satellites…). Again, in many situations the present wireless infrastructure meets the demand, although in others it lacks coverage and reliability. Whilst in industrial environment it is possible to design and deploy the required wireless infrastructure, leveraging on current technology within acceptable and affordable economic constraints, in open areas, including urban and rural environment this is not always possible (from an affordable economic standpoint).

Hence the current, growing, interest for a massively distributed AI, with local, embedded, capabilities in vehicles, devices, wearable, …

It is in this framework the activity of the Edge AI Alliance. Hailo is an Israeli start up that is operating in the Edge AI framework and is proposing a tiny chip, see the photo, that can be used to bring AI capability to serve the increasing automation in vehicles. Their chip, the Hailo-8 processor, has a 26 TOPS capability (26 thousand billion operations per second) with a power consumption of the order of 9W, and supports deep learning applications that were previously requiring the Cloud. This for vehicles is crucial in increasing safety, since it would be no longer required the connectivity between vehicle and cloud, something that cannot be guarantee at any time in any place.

The chip is already being integrated in system for image recognition to provide context awareness, like in the offer of Leopard Imaging where the chip processes data coming from two 4k Sony cameras (the Hailo-8 can process up to 20 video streams in real time!).

What is also amazing is the price of this chip: around 25$.

We are really going to see a shift from the Cloud to the Edge in terms of AI support, enabling devices, and consequently environment to become more and more context aware, a fundamental stepping stone towards increased intelligence and usefulness.

About Roberto Saracco

Roberto Saracco fell in love with technology and its implications long time ago. His background is in math and computer science. Until April 2017 he led the EIT Digital Italian Node and then was head of the Industrial Doctoral School of EIT Digital up to September 2018. Previously, up to December 2011 he was the Director of the Telecom Italia Future Centre in Venice, looking at the interplay of technology evolution, economics and society. At the turn of the century he led a World Bank-Infodev project to stimulate entrepreneurship in Latin America. He is a senior member of IEEE where he leads the Industry Advisory Board within the Future Directions Committee and co-chairs the Digital Reality Initiative. He teaches a Master course on Technology Forecasting and Market impact at the University of Trento. He has published over 100 papers in journals and magazines and 14 books.