Steve Rogerson asks how AI will change the outlook for IoT.

A few years ago, when listening to a talk on how one day personal health devices would feed data constantly direct to doctors and hospitals, I asked how all that information was going to be processed. You were not going to have doctors spending their days looking at banks of monitors to see if somebody has a blip.

The question was shrugged off, as it was when I asked the same about industrial control and traffic monitoring and so on. There was a piece of the jigsaw missing and at that point nobody really knew what it was. We now know – the intelligent edge, where artificial intelligence (AI) is built into these edge devices so decisions can be made without going back to the cloud with the delays that involves.

And not just AI, but generative AI and large language models. Suddenly AI has leapt into the mainstream and it is the best news possible for IoT. Most of the problems of handling big data fade away when intelligent edge devices can suddenly make the big decisions.

Is there a catch? I asked that question to Miller Chang, president of Advantech’s (www.advantech.com) embedded IoT group, at the company’s World Partner Conference in Taiwan earlier this month.

“Ten years ago, we started talking about IoT for factory automation, energy management and so on,” he said. “Five years ago, IoT was still too slow. Our customers wanted it operating faster. And there was no killer IoT application. Then two years ago, many started adding AI to their IoT applications.”

IoT systems before that just used to gather basic information such as temperature. There was no intelligence.

“With AI added, they can make decisions based on the status,” said Miller. “AI plus IoT. We have started to push AI into our IoT products. We are pushing hard.”

However, he warned that people could not 100% rely on generative AI to make the decisions.

“The database might not be complete,” he said. “There are also security issues. AI needs more time to learn the different problems. Some are worried about putting all their data on AI. We need to see AI becoming more mature.”

What AI will do, and is doing, is moving information and decisions from the data centre to the edge.

“That will be the big change,” said Miller. “We will see AI in every edge device. We want to speed up AI development at the edge by working with the likes of Qualcomm, NXP, Intel and so on. Nvidia are focussed on the data centres and not the edge. That is a big opportunity for the other silicon companies like NXP and Intel.”

A big difficulty here, of course, is power consumption. One of the key criteria for an edge device is that it consumes little power as it may have to operate on a battery or energy harvesting for years. Higher processing capabilities draw more power.

“The silicon companies are getting round this with more integration,” said Miller. “Not Nvidia. They have high performance AI at the server side so are not worried about power consumption.”

At the conference, Qualcomm (www.qualcomm.com) partnership director Madav Murali Mohan stressed the importance of power consumption. “The edge is becoming more sophisticated and it is becoming low power,” he said. “Everything we do starts with low power, but you also need high-performance computing plus communications and AI.”

Miller said already there were parts of industry that just could not operate without automation and local intelligence, warehouse management being a classic example.

“The likes of Alibaba and Amazon are all using autonomous vehicles in their warehouses,” he said, but acknowledged that would make some jobs less safe. “In factories, AI will replace some operators,” he said. “Some of the jobs will be replaced by AI.”

So there are catches, but the IoT industry is committed to travelling the AI road, and the possibilities that will bring are only just beginning to be seen. The next few years, I predict, will be very exciting for those following technology trends, and for everyone else.

We won’t be able to get away from AI.