Internet of Things
State of now: obstacles to AIoT market fall for third consecutive year
By Aneet Chopra, VP business development, marketing & product management, XMOS I
f there is a lesson for the consumer electronics market in the last five years, it is a simple one: AI sells. What else could possibly explain the explosion of smart kettles, smart doorbells, and smart speakers that have become staples of the tech-savvy home?
While it’s clear that artificial intelligence is an attractive proposition for consumers, the reality is that the actual intelligence is not in the devices themselves. Instead, it is in a cloud data centre potentially thousands of miles of away.
This poses several problems. Relying on cloud connectivity – and the lengthy communication chains it necessitates – not only poses a security risk for consumer data, but also makes it difficult to avoid performance constraints due to latency and bandwidth issues. As such, the ‘smart’ gadgets that are available today are only offering a superficial example of AI technology. This is why manufacturers have been so keen to explore artificial intelligence of things (AIoT) technologies that sit at the intersection of AI and the internet of things (IoT). Indeed, XMOS research from 2020 highlighted that 82 per cent of electronics engineers thought that incorporating AI capabilities into their endpoint devices would increase the competitive advantage of their designs.
The AIoT promises to embed intelligence directly into the machines in people’s homes and hands. Pushing the intelligence to the device allows them to interact directly with local environments, respond to commands instantaneously, and to make decisions both for themselves and in concert with others. This has the potential to unlock a new wave of innovation in ‘smart’ products, and to create compelling new user experiences that stretch from smart health to the factory floor. There is, however, a reason – two, actually – why we haven’t quite entered the age of the AIoT.
The barriers to the AIoT As might be obvious from the IoT in the
32 October 2022
name, the AIoT is a network of typically small, portable, and possibly battery-powered devices. In contrast, AI has been developed primarily to operate in the data centre. This creates a major disconnect for designers who want to create new concepts and ideas for the AIoT. AI-capable chips are generally
sophisticated, powerful processors that utilise the absolute bleeding edge process nodes – which come with commensurately high costs and power consumption requirements. When these chips end up in a server in a data centre this isn’t really a problem. But if we try and put the same chips in a smart speaker, the economics and the physics quickly fall apart. Despite the ambition and desire to tap into the AIoT, these two issues of cost and power have forced engineers to abandon their ideas at the design stage. Attempting to build the AIoT with server-grade AI chips is simply unfeasible.
A change ahead
That is not to say the AIoT is dead in the water. In fact, XMOS’ latest Edge of Now
Components in Electronics
report, the third in our ‘Edge of’ series, suggests that we are on the cusp of a sea change in the growth of the AIoT. For one, engineers have reported significant reductions in how much cost and power concerns are holding back their AIoT ambitions. Whereas two years ago 64 per cent of designers cited cost, and 65 per cent highlighted power requirements, as barriers to increasing the on-device processing capabilities of their designs. This year the figures are 24 per cent and 23 per cent respectively.
This is having a direct impact on production lines. In our 2022 research, almost two thirds (63 per cent) of engineers said they are currently working on projects embedding AI on an end point device that is due to launch in the next 12 months. Looking further ahead, two fifths (38 per cent) of engineers surveyed say that their entire product range will have the on-device processing power to support AIoT functionality by 2025. Another quarter (26 per cent) claim that 70 per cent of their product range will have on-device processing by this time.
Entering the AIoT age This optimism is hugely exciting for the future of AIoT – reflecting the momentum among engineers, the normalisation of existing gadgets like smart speakers, and the development of a new wave of versatile and cost-effective chips that speaks specifically to the needs of the AIoT.
Despite this surge in confidence, the challenges of designing for the AIoT have been mitigated rather than eliminated. Chip suppliers and the rest of the electronics supply chain must stay focused on the specific requirements of the AIoT. We need to empower engineers to keep smashing through barriers to adoption.
We need to continue to deliver low-power silicon at the right performance/cost point to make the mass-proliferation of AIoT-enabled designs a realistic ambition. For the vendors and manufacturers that can strike this balance there will be a significant reward – in claiming a slice of a potential multi-trillion-dollar opportunity. Because let’s not forget: AI sells.
https://www.xmos.ai/the-edge-of-now- report/
www.cieonline.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62
