Currently the CEO of AI chip company XMOS, Mark Lippett is an experienced business leader with over 25 years’ experience in technology.
getty
The last 12 months have undoubtedly been incredibly tough. The pandemic has wreaked havoc across the world, and everyone is now realizing that, unfortunately, Covid-19 is here to stay.
As we begin to accept this fact, we are now looking to adapt the way we live and interact with the world around us. Industries, businesses and governments are all adjusting the status quo in an attempt to keep people safe, productive and able to live their lives.
How — and where — we work is just one of the ways most people have had to change. Working from home has been the norm for so many over the past year, and even as we progress out of this pandemic, we may see businesses continue to allow employees to work from home. As a result, this way of working has placed a renewed focus on the importance of our homes, and discussions around the tech-enabled "smart" home have never been timelier.
Just as the pandemic was taking off, a new kind of technology ecosystem called the "artificial intelligence of things" (AIoT) was taking a foothold. The AIoT represents a convergence of connected things (the IoT) and artificial intelligence (the AI) deployed within those things. I've previously written about what the AIoT is and how it's set to transform all kinds of industries, from healthcare to transport, but there's no bigger opportunity for it than the smart home.
It's still a relatively nascent industry, and so last year, we conducted research into the barriers holding back the AIoT. Within that research, electronics engineers highlighted significant market-level and device-level concerns. We then conducted the same research a year later to see how things had changed. The headline? Necessity is the mother of all invention, and the pandemic has accelerated the development and adoption of the AIoT for the smart home and beyond.
Security, connectivity and scalability are all becoming easier to address.
In our original report in 2020, engineers cited security, connectivity and scalability as the biggest market issues facing the AIoT. Over the last year, however, opinions have softened, and many believe the barriers are more surmountable.
With security, AI raises privacy concerns because of its reliance on data. The "smarter" the device, the more information it requires. However, in the last 12 months, engineers have realized that processing data locally instead of in the cloud can solve the privacy issue. Homes can keep their data within their four walls without the need to send it to third parties in the cloud, reducing the risk of leaks.
By keeping data within a home, a remote cybercriminal would have to turn into a common burglar to steal that data. While that's unlikely to happen, it's still important that device manufacturers ensure the processing that happens on their devices is secure. A gamut of device-level security features — including secure key storage, accelerated encryption and true random number generators — can provide a foundation for significantly improved safety for both data and decision-making.
Security aside, engineers also felt connectivity posed a large barrier to AI deployment, with 38% voicing concerns about the technology's ability to overcome latency problems. In-home healthcare monitoring, for example, cannot afford to be burdened by unreliable connectivity issues when they need to make decisions based on potentially life-changing situations like heart attacks. Now, however, on-device processing lessens the need for networks, making network latency a moot point; only 27% of industry experts consider connectivity as a major barrier to the technology.
The industry should move to on-device processing if it wants to create applications that aren't held back by latency. Certain AIoT chips are now lightning-fast and predictable, with execution determinism measured in single-digit nanoseconds, which enables products to think and make decisions at speed.
Finally, last year, engineers highlighted the issue of scalability. Engineers are aware that the number of connected devices is rising, placing more and more strain on cloud infrastructure. In 2020, about a quarter of engineers believed scalability to be a major barrier to the success of edge technology. However, experts are now beginning to see the benefits of the AIoT's deep-rooted scalability. Processing at the edge removes the dependence on the cloud, negating any potential growth and scaling issues. Now, fewer than one-fifth of engineers believe that cloud infrastructure could hold back edge AI.
The good news is that the electronics industry doesn't need to do anything specifically to maintain the AIoT's scalability, as one of the major technical barriers to the AIoT's expansion — the need for the cloud to process billions more devices and petabytes of data in the future — has been extinguished.
Power capability up, power consumption down.
As the AIoT market has matured over the past year, it's also progressed on a technical front. On-device processing capabilities have increased while reducing the amount of power and expenditure it takes to enable AI. Now, chips are flexible enough to meet the diverse needs of the AIoT at a much more affordable price point than ever before.
As AIoT chips become a more realistic option for product manufacturers, how can engineers transition toward using them?
One of the key considerations is the development environment. Too often, new chip architectures mean new and immature proprietary programming platforms, which engineers need time to learn and get familiar with. Instead, engineers should look for flexible platforms that are accessible using industry-standard techniques that they are already familiar with — full programmability in C, runtime environments like FreeRTOS and AI tool flows like TensorFlow Lite. Working with familiar platforms means engineers can program chips quickly without having to learn new languages, tools or techniques.
A single programming environment for all of an AIoT system's compute needs is a fundamental enabler to the design speed that is crucial to ushering in the new era of fast and secure AI in the home.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
