Despite being a relatively nascent technology, artificial intelligence feels like it has been around for years. Despite the buzz, the functional side of AI is decidedly unexciting. Predominantly housed in remote data centers, it’s capable of processing huge amounts of information using complex algorithms — with varying degrees of autonomy.
The exciting part is the benefits that AI provides to both business and consumer: the time it can save, the tasks that it can automate or eliminate, the safety enhancements, and the impression of sentience. AI is an invaluable tool to many businesses while simultaneously establishing itself as part of our non-working lives — whether it be in our phones, Google searches, or smart homes.
What is AIoT?
(Source: Shutterstock)
One of the most successful potential synergies for AI is the internet of things, a network of devices embedded with software, sensors, and technology that allow them to “talk” with one another. Enabling communication between these previously isolated units allows them to collate larger, more sophisticated datasets — which gives AI models a huge amount of material to analyze and react to.
The synergy here — a device that captures better, larger sets of data, matched with an AI to take full advantage of it — is a natural point of convergence that we refer to as the artificial intelligence of things (AIoT). In an ideal future, we can rely on a network of intelligent devices capable of capturing and interrogating data and using it to contextualize and trigger actions in new use cases that enhance our lives.
Why is AIoT needed?
Perhaps the most revolutionary element of the AIoT is that it frees us from the traditional cloud structure that we’re so used to. Rather than housing intelligence in the cloud, and relying on that to interpret data, the AIoT makes devices intelligent in and of themselves.
Processing and decision-making take place on the device, allowing for near-instantaneous decisions. Statista forecasts that there will be over 75 billion IoT devices in 2025, which cloud infrastructures simply cannot scale to. As devices begin to proliferate, cutting out the latency issues that come with constantly sending data to and from the cloud is imperative to the IoT network as a whole.
Beyond that is the question of user privacy: In an era when concerns around consumer privacy are continuing to grow, data security is more important than ever before. Research from Cisco in 2021 found that almost half (46%) of consumers didn’t feel that they could effectively protect their personal data: The feeling of being “watched” or “observed” is a growing concern.
Rather than capture and transmit data, AIoT devices can retain data in on-board devices, ensuring that it isn’t shared with tech giants with potentially nefarious advertorial intentions. In many cases, these devices don’t need to store data at all, minimizing data transfers and making it harder for hackers to access personal information.
That independence isn’t just digital but also geographical: AIoT devices have the potential to function almost anywhere. For example, our current smart devices constantly demand stable internet connections for their functionality. AIoT-ready chips enable devices like smart lights or thermostats to operate independently of such networks, offering full functionality even in the presence of unstable internet connections.
Given the benefits that the AIoT brings to our devices, it is no surprise that many engineers are excitedly watching its development with demand and opportunities growing year on year. In fact, it is estimated that the industry will be worth a staggering $78.3 billion by 2026, according to a Research and Markets report, promising to be one of the most exciting, fast-growing tech industries out there.
How will AIoT be implemented?
While the potential of the AIoT is clear, much of our current hardware limits its realization. There’s a fine art of balancing power, cost, and flexibility in designs, which often stems the flow of innovation.
Research conducted at XMOS demonstrates that processing power issues are a barrier ever present in the minds of engineers, with more than half (53%) citing a lack of power as a major issue. That’s not because power is unavailable but because it is expensive and not necessarily built for purpose — it can be nearly impossible to find the appropriate hardware to deliver an appropriate balance in the power/cost ratio.
As a result, in 2021, 48% of engineers cited costs as a major barrier to the adoption of the AIoT, with 73% claiming that hardware prices are already very high. The existing chips that pack enough of a punch to make use of the AIoT are simply too expensive, big, and rigid — a poor match for the many small devices that will make the best use of the technology.
However, while the barriers are clear, they are surmountable. We are starting to see the advent of processors built specifically for the AIoT that balance AI, DSP, control, and I/O in a manner that allows the product designer to define them in software, rather than the chip vendor fixing them in hardware, and to do so in a cost-effective package suitable for mass production.
The future: a new home for AI
Such hardware is paving the way for the AIoT to enter the mainstream, and it will revolutionize the technology industry as we know it.
The smart home is introducing new functionalities every year, working toward complete, seamless control over our surroundings. Facial-recognition–enabled cameras could easily differentiate between family and friends — being able to recognize unwanted visitors or identify a vulnerable relative in trouble. In time, these devices will speak to each other: An unwanted face at the back door could lock your windows.
You can take that logic beyond the smart home into, for example, the world of health care. Imagine devices that can monitor your breathing and heart rate, capable of alerting emergency services or sharing key findings or patterns with your doctor, without exposing any of that data to those who would use it for advertising.
Ultimately, once manufacturers have addressed the barriers to widespread adoption, we’ll start to see the wave of innovation that will herald the establishment of the AIoT in the mainstream. It is destined to become a multi-billion–dollar industry alongside the internet, computer, and smartphone as one of the biggest technological breakthroughs in modern history.
Many of the statistics shared throughout this article are taken from the XMOS 2021 Edge of Today report. The updated 2022 report will be published in Q2. If you would like to be among the first to receive the report, please enter your details here.
Advertisement
Learn more about XMOS
