According to Gartner, by 2023 75% of data will be created and processed at the edge, and ABI Research suggests that 43% of AI applications will occur at the edge. When a company like Intel, which has been a world-leading chip manufacturer since the 80s, throws its weight behind edge AI computing, the industry sits up and takes notice.

Earlier this year, Intel launched the Intel Xeon D-2100 processor specifically to bring capabilities out to the edge so devices can gather data locally and become important sources of data.

While edge AI offers endless capabilities, like any emerging technology at an early point of maturity the industry needs a critical mass of successful use cases to spark its imagination and drive its adoption into the mainstream. With this in mind, Intel vice-president and chief operating officer, artificial intelligence product group Remi El-Ouazzane says he focussed his address to the AI summit on AI in real life.

“We gave an example of Phillip’s healthcare imaging diagnostics,” he says.

“They had two problems; an issue of transitioning to deep-learning implementation for their X-ray bone-edge prediction, and in CT they had a challenge to do segmentation for lung disease discovery.”

Edge AI in healthcare

Intel used its scalable Xeon processors with Phillips’ existing models and infrastructure, optimising the memory module and making the data fetch as efficient as possible to streamline the data flows.

“All those optimisations combined led us to improve by their bone-edge detection algorithm, which was at 1.5 to1.6 images per second, by 130 times. In the context of a CT scan, it was multiplied by 40 times.”

In another healthcare example, Intel has worked with Siemens Healthineer, a group working on advanced imaging diagnostics.

“Siemens is doing cardiovascular MRI, which they call CRI. At the point of care, it can detect whether or not you are subject to potential cardiovascular issues.”

“The number of deaths from cardiovascular issues in the US is rising, as I believe it is everywhere in the world” says El-Ouazzane.

“Siemens is doing cardiovascular MRI, which they call CRI. At the point of care, it can detect whether or not you are subject to potential cardiovascular issues.”

Using its Xeon processor and OpenVINO computer vision toolkit, Intel was able to improve the image processing time by five or six times, so at point of care, it can process more patients in the same timeframe or each patient faster. Intel processors also power the HOOBOX solution than translates facial expressions into wheelchair commands for paraplegic individuals who can only express themselves through facial movements.

Edge AI is also being used to analyse voice samples to provide an early diagnosis of Alzheimer’s disease.

AI beyond healthcare

More prosaically, AI will soon be finding its way into everyone’s workplace.

“We've been in at the very beginning of what Microsoft is going to do with Windows ML, which is a machine learning API extension to Windows that any developers can tap into,” says El-Ouazzane.

“There are really cool features where they deploy that into Word and Excel for automating tasks including a bunch of automation that they are thinking of for the Office suite. It could infer from an email that you want to create a slide out of the information in it, for example.”

AI won’t just feed our minds; in the near future, it could help feed the world too through farming applications.

“This is on the low end of tech, but it will evolve.”

“It's become more relevant with the number of satellites surrounding us that are providing a super-accurate map of the world where everybody could potentially be world farmers,” El-Ouazzane says.

“They could now potentially contract companies to give them recommendations on how to best farm their land without having to go through the steps of using genetically-modified organisms, just by knowing how to best exploit their land in an optimal way.”

Such is El-Ouazzane’s faith in AI that he even uses it to support personal projects.

“My wife and I are investing in a company in New York where they deploy machine learning in Oakland schools to personalise the maths curriculum for kids from less fortunate or less affluent neighbourhoods,” he says.

“It essentially extracts which atomic concept you're failing to understand and created a decision tree on how to evolve your personal maths curriculum. Education can also be positively impacted; this is on the low end of tech, but it will evolve.”

Share this article