emsleaderboard

Breaking down barriers: the increasingly positive attitudes towards AI adoption in electronic engineering

Aneet Chopra, VP Business Development, Marketing & Product Management, XMOS

 

Aneet Chopra VP business development XMOS

Aneet Chopra

The term “artificial intelligence” (AI) is far too appealing for its own good. It’s been employed as a fairly transparent sales tactic for some years now, with an ever-proliferating range of products gaining sentience at an unexpected rate. We’ve all heard about smart homes and smart cars, but we’ve also encountered marketers branding calculators and rice cookers with the same label.

The fact that so many technologies are being dubbed with the AI label is testament to how desirable the concept has become. The prospect of gadgets that can adapt and think for themselves is understandably compelling to the average consumer. Yet in reality, developing truly intelligent electronics requires powerful processing capabilities, and an array of sensors to capture various forms of unstructured data.

Currently most devices rely on the cloud, with data being transferred to server farms thousands of miles away. This can create a bigger threat surface with regards security, to say nothing of latency issues and a heavy reliance on internet connectivity.

A solution would be to on-board AI, rather than putting data through the lengthy and convoluted process mentioned above. Embedded intelligence will be possible to develop networks where applications make decisions not only for themselves, but collaboratively with others.

Thankfully this is becoming possible via an innovative new technology infrastructure – the artificial intelligence of things (AIoT). This is where traditional IoT devices converge with artificial intelligence, bearing implications for a range of manufacturers, whether they be concerned with cars, factories, or homes.

Until recently, there have been two market barriers to embedding AI during the design process: cost and power.

 

When the price isn’t right

The development of silicon to support AI historically focused on the cloud, which involved server farms expected to process billions of requests daily. Perhaps unsurprisingly, the price scaled up with these requirements.

However, using this hardware won’t work for the AIoT – not only due to issues around cost, but also the wastefulness related to design specifications. If you’re a smart homeowner, you don’t need your smart lighting system to analyse big data, you just need the lights to turn on.

The emphasis placed on the cloud has rendered many projects unsustainable in terms of cost. What’s more, scarcity issues arising from the pandemic affected cutting-edge semiconductors more acutely than chips featuring established nodes, increasing costs even further.

 

When devices go on a power trip

AI chips are often regarded as high-end, consume significant amounts of energy, and can be complicated by other design decisions. Technologies like ‘always-on’ sensors present significant difficulties for any manufacturer looking to keep designs within power requirements.

Concerns around cost and efficiency are obviously interrelated. Reaching the maximum power consumption provided by your blueprint leaves manufacturers with a difficult choice: either sacrifice functionalities which reduces the marketability of your product, or procure hardware for greater power, which increases the cost.

This balancing act has led to many an engineer’s design being scrapped, despite the creativity, and the potential of their designs to push boundaries within the relevant industry.

 

Breaking down barriers

Thankfully, our research suggests cost-related concerns are finally starting to subside. XMOS’ Edge of Now report – the third annual dive into engineer’s attitudes  to the artificial intelligence of things (AIoT) – suggests that the future outlook is more optimistic than ever.

Only 24% of the engineers surveyed cited cost as a barrier to boosting on-device processing abilities – that’s only half of the 48% mentioning the same in the 2021 research, and barely a third of the 64% from 2020. What’s more, just 16% regard cost as a barrier to AIoT adoption.

This decline in scepticism is matched by reduced concerns around energy efficiency, with only 23% of engineers seeing consumption as a barrier to increased on-board processing, dropping from 53% last year, and 65% the year before.

These figures highlight a sustained increase in optimism and confidence when it comes to implementing the AIoT, with a growing number of engineers primed to address market-related challenges.

 

Building momentum

This rise in optimism suggests we’ll bear witness to a new generation of innovative products. Almost two thirds of the engineers we surveyed are developing products due to hit shelves in the next 12 months, featuring the integrated processing needed to support AI functionality.

These devices are poised to become the norm, with two fifths (38%) of respondents saying their product ranges will have the processing needed to integrate AI into their designs, and 26% saying that will apply to at least 70% of their product range.

With this in mind, engineers and designers need to be empowered with cutting-edge, adaptable, and cost-effective components. There will also be a need for silicon that’s cheap enough for mass production, and applicable to evolving designs. With these in place, engineers will be able to address changing market requirements while minimising supply-related concerns.

Bringing legitimate on-board intelligence to market will provide opportunities worth billions of dollars, hosted on electronics that will changes our lives for the better. Whether it’s smart car parking preventing fines, healthcare gadgets saving lives, or anything in between – seeing the growing level of confidence in the AIoT is a genuinely exciting development.

x Brown

About The Author

EMS Now Mycronic Iris D vision technology x
panasonicmpu
Cleveland opening banner px
SUBSCRIBE Ad
emsmpublue
EMSNOW MPU Ads
SCJ x en