Making the AIoT a reality – part 1

The artificial intelligence of things (AIoT) represents the convergence of two of today’s most significant technology trends.

The internet of things (IoT) connects “dumb” devices to the internet, with an ever-growing 10 billion devices surrounding our everyday lives. These devices are already generating huge volumes of data from the world around us – providing access to valuable insights and enabling enhanced efficiencies.

While that is incredibly valuable in its own right, the AIoT takes the IoT to the next level – adding artificial intelligence to these devices to give them a “brain.” 

This convergence opens the door to a truly transformative technological shift – with the potential to create complex, intelligent networks of decision-making and analytical systems. It is no understatement to say that this combination is capable of changing the world as we know it – transforming experience through technology and enhancing business- and mission-critical applications in end devices.

The possibilities are endless

The potential applications for the AIoT spread across multiple verticals, including the smart home, connected health, smart cities, smart mobility, and industrial IoT. 

Edge-based AI processing brings intelligence close to the end user for a premium experience. Instead of needing to wait for connectivity to the cloud, localised AI models can make instantaneous predictions or decisions on their own – extremely useful for latency-sensitive application. 

By enabling functionality such as face and image identification, voice identification and interface, presence detection, comms and control and actuation, the impact of the AIoT can be huge.

For example, in a connected healthcare scenario the AIoT will build on devices that are capable of monitoring things like heart rate or breathing patterns and support pre-emptive alerts for potentially serious incidents.

In the smart home, the AIoT opens up opportunities for greater safety, convenience and automation – offering total control of the home environment without the need to divert your attention.

For the smart city, the AIoT will enable new standards in convenience, efficiency and security. From directing road users straight to empty parking spaces in a busy city, to ‘on demand’ streetlights triggered by proximity sensors and using microphones to guide emergency services more quickly and accurately towards noises like gunshots or breaking glass.

In industrial settings the AIoT can support complex far-field voice interaction, enabling people to operate machines without the need for physical interaction where hygiene is paramount with no loss of control – a factor that is only becoming more important as the world looks for solutions to mitigate the risk of the COVID-19 pandemic. Indeed, it’s likely that voice interface capabilities developed for industrial purposes will find uses outside of the factory. Back in the factory, on-device data analysis will also enable rapid fault detection in machinery, supporting pre-emptive maintenance. 

Edge inference is also transformative for some mission-critical IoT devices that demand real-time decision making. Data privacy appears to be the standout benefit at the moment. Many consumers and enterprises are reluctant to share their data with other parties, fearing a lack of control, potential misuse and cyber-crime. Processing data on the end device rather than sending it to the cloud, reduces the perceived risk of data leakage.

A need for a new architecture

The capabilities that unite these seemingly disparate applications are simple – ultra-low latency, the highest possible security, and high compute demands. With all these combined, the existing IoT model is no longer sufficient. 

This diversity of demand, together with the need for speed, reliability and security, creates challenges for designers of connected products. Not only is the cost of the typical high-performance CPU required to deliver AI functionality prohibitive, the fundamental reliance on cloud networks is a severe limitation on the AIoT. 

The cloud’s compute capabilities simply cannot scale proportionately with the volume of connected devices that will be deployed in the coming years. Even more significantly, the networks that transport data back and forth between devices and the cloud are bandwidth limited. Even the most advanced communications networks won’t be able to support the explosion of data that devices will create. This will inevitably cause unacceptable delays in any decision made in the cloud.

Instead, for a natural, connected world that makes life simpler, safer and more satisfying, we’re going to need AI-enabled decisions taking place on-device, and not in the cloud. And that means we need a new approach to deliver the hardware that will provide the same level of performance as today’s high-end CPUs, but with greater economy and ease of use.

In part two of this series we will outline exactly what that new approach looks like.

Scroll to Top