The rise of AIoT brings both opportunities and challenges

Translated from EEtimes

In the 1980s, French sociologist Bruno Latour proposed ANT (Actor-Network Theory). In this theory, actors can be humans or objects, and the relationships between actors constitute the entire network.

This is a black box theory. From the outside, the connections between actors are not clear at a glance. But with the existence of actors, there are countless possibilities of connections between them. Everything is in this network.

Nowadays, the concept of the Internet of Everything is not new. 2017 is regarded as the first year of commercial use of the Internet of Things (IoT). With the official freezing of the narrowband Internet of Things standard, the focus of the three major domestic operators, and the strong promotion of equipment manufacturers, the large-scale commercial use of the Internet of Things is entering the fast lane. Artificial intelligence (AI) seems to have existed for decades. It is usually based on remote data centers, capable of collecting and examining large amounts of data and generating insights based on analytical algorithms. With varying degrees of autonomy, these capabilities are used to simplify the decision-making process.

Although AI is often thought of as a standalone product, it is increasingly being integrated into other industries or products. Chief among these is the Internet of Things (IoT), which enables previously isolated machines to "talk" to each other, while generating data that enables new modes of operation.

There is an obvious convergence point here, and it’s being expressed as Artificial Intelligence of the Internet of Things (AIoT). The vision of AIoT is to create a network of smart devices that can collect and analyze data remotely and turn that data into insights and actions locally, enabling a wide range of use cases that simply weren’t possible before.

Inversion of the cloud model

Perhaps the most revolutionary thing about this vision is our inverting of the cloud model. Moving services, on-premises solutions, and networks to cloud infrastructure has been a convenient and popular solution to digital problems to date - but with the AIoT model, data reasoning can be done locally on the device.

There is a basis for adopting this model. According to the data from the China Academy of Information and Communications Technology, as of mid-2018, the overall scale of my country's Internet of Things industry has reached 1.2 trillion yuan, completing 80% of the 1.5 trillion yuan scale of the Internet of Things industry proposed by the Ministry of Industry and Information Technology in 2016 during the 13th Five-Year Plan. In the process of technological development, we have seen that the Internet of Things has moved closer to the Intelligent Internet of Things (AIoT). AIoT refers to the integration of artificial intelligence technology and the Internet of Things in practical applications. As the countdown to 5G commercial use in 2020 begins, the Internet of Everything is within reach, and the "free, anytime, anywhere, human-machine integrated communication method" is coming.

In the future, the competitiveness of smart products will become weaker and weaker, while true interconnection will bring huge business opportunities. Chen Liaohan, chairman and president of Tuya Smart, believes that the Windows operating system was born in the Internet era. In the future era of the Internet of Things, it is very likely that every family will have a new operating system to realize the Internet of Things life of human-machine integration.

But cloud infrastructure and connectivity cannot scale in the same way, and more smart devices must be deployed at the edge.

AIoT's use of device processing power alleviates past network bandwidth, computing scalability, latency, and security issues. By reducing the load on the entire network while reducing expensive costs, AIoT can distribute workloads in a way that improves performance.

However, there is a caveat to this view. Simply putting today’s high-end CPUs into end devices is not feasible. These CPUs are too power hungry and expensive to be suitable as a commercially scalable model. So how can the necessary processing power be delivered to end devices in a cost-effective manner?

This is a problem that must be solved and a significant market opportunity if AIoT is to achieve explosive growth in the coming years.

Delivery requirements

Achieving the necessary processing power without the cost and energy demands of current high-end CPUs is an unenviable challenge.

Any new processor designed for AIoT must have a lower price point than existing solutions. But this also presents its fair share of challenges. Existing CPUs rely heavily on third-party hardware/software, with all the associated licensing costs. Therefore, new processors will require new architectures to eliminate the need for third-party IP. In addition, component requirements and costs throughout the system must be minimized.

At the same time, these cost-saving measures must not compromise performance. While some AIoT applications do not require the full processing power that a data center server can provide, the requirements are still very high for even basic AI and decision-making functions. This means that processing functions require new approaches, with complex algorithms doing most of the heavy lifting - allowing the hardware itself to be relatively lightweight.

Since the reality of the AIoT market is not monolithic—there are hundreds of markets and thousands of segments with widely varying standards and requirements—the biggest challenge may be creating a class of processors that can be applied to a wide range of applications.

We all know that the cost of developing a new CPU is not trivial. It is unrealistic to expect suppliers to produce thousands of different platforms to meet all these different needs. If you produce a chip that only serves one purpose or one application, then the problem has not been solved.

Instead, AIoT processors require ultra-high flexibility, with programmable trade-offs between compute classes (AI, DSP, control, and I/O) that can be defined by product designers rather than chip vendors.

Make dreams come true

If these requirements can be met, the impact will be enormous. A fully functional, commercially viable AIoT will revolutionize areas ranging from smart homes to connected healthcare, the automotive industry, Industry 4.0, and smart cities.

Making smart home possible

In 2017, the global shipment of wearable devices reached 115.4 million units, and it is expected to reach 122.6 million units this year, of which smart watches and bracelets account for the vast majority. Behind the numbers, it reflects the rapid growth of the smart market. Taking North America as an example, this year the growth of smart access control exceeded 115%, lighting equipment reached 40%, and home linkage products exceeded 30%.

At the same time, the consumer Internet of Things has developed a trend of "trendy branding". This is particularly common in the European and American markets with high penetration rates. In Germany, Osmart Zigbee Smart Plug, Philips Hue White Ambiance GU10 LED Spot smart bulb, Sandisk Micro SD memory card, etc. have become local Internet celebrity brands. On the other hand, the growth of the industrial Internet of Things is also very rapid. According to GSMA Intelligence, from 2017 to 2025, the number of industrial Internet of Things connections will increase by 4.7 times, while the number of consumer Internet of Things connections will increase by 2.5 times.

With AIoT capabilities, you can fully control every aspect of your surroundings without having to divert your attention, let alone navigate through our many apps.

While this may seem trivial, the benefits in other application areas may be more significant. An AIoT-based healthcare market could open up a world with a much higher content of preventive medicine. Devices in the home could track heart rate and breathing, detect problems early, provide alerts, and automatically provide real-time data to healthcare professionals who can provide the right treatment and care when needed.

Ultimately, this technology could radically improve our quality of life, bringing new conveniences and efficiencies to everything from the most mundane aspects of our daily lives (like finding a parking spot) to the most important things (like our safety and health). None of this would be possible without a new generation of electronics that can diffuse the power of AI to the devices around us.

It’s a huge challenge, but it could herald a true revolution in intelligence.