ew 2024: The Future of Embedded Systems

What specific workloads and requirements are shaping the technology roadmap for Intel and its partners?

Bassel Haddad: The workloads that are driving the technology roadmap are the one that we see very common in the various verticals, industrial verticals like healthcare, retail, and industrial.

In healthcare, you can think about AI-enhanced medical imaging. In industrial it would be defect detection and preventive maintenance. In retail would be AI-enhanced or AI aided self-checkout and store operation.

If you take that to a workload viewpoint, that would be vision media analytics and data analytics for basically camera streams coming into retail store. It would be a time-serious data for industrial for things like productive maintenance. And it would be just imaging analysis for any anomalies in healthcare.

When you take that into the type of AI models, vision has started with CNNs, convolutional neural networks, and we start seeing some vision transformers coming into the market. We start seeing also generative AI starting to get into the edge market with things like chatbots and customer service through conversations. So these are the things from more of the technology side of things.

What can you tell us about the latest Intel^® technologies and the benefits they offer?

Bassel Haddad: So if you think about the approach we’re taking to Agile portfolio, it’s about scalable portfolio that addresses very different needs of the market. There’s no one size fits all in edge AI. So there’s different design points, different power consumption requirements, different performance needs. So the way we have built our roadmap is to enable infusing AI into existing Brownfield deployment. And of these applications which are infused by AI should be living alongside other existing workloads, basically, for customers to be able to continue to leverage their investments in software and hardware.

So if you think about it, basically you can take for example, Intel^® Core&^™ processors, and if you need more AI than what’s built in into that device, you can attach to it on Intel^® Arc^™ discrete GPU. The beauty of that is that you have the same foundational and GPU graphics IP between the integrated IP and the discrete. So from a software, there’s software consistency, and it’s really eases on the developer journey.

The other part is we have a common runtime that basically abstracts that hardware complexity across the heterogeneous AI engines, CPU integrated, GPU, and what we introduced in Intel^® Core^™ Ultra, which is NPU, stands for Neural Processing Unit. So from a developer, they don’t need to understand exactly the details of the hardware, and all of that is abstracted through runtime optimization that OpenVINO^™ drives.

How can users get the most value out of these emerging platforms?

Bassel Haddad: One of our key objectives is to streamline the developer journey and the developer workflow. And for us, OpenVINO is the runtime environment that not only covers the edge side through the heterogeneous hardware between CPU, GPU, integrated GPU and NPU, but also covers the inference that happens in the cloud in a way that enable what we call it, develop once and deploy everywhere, and enables what we refer to as hybrid AI where it’s a mechanism or it’s an inference approach where you have inference happening at the data center and happening at the edge to really dynamically set the inference based on regulation and on latency. So sometimes your latency or constraints are too strict to allow the data to travel to the cloud and come back. So you need to do that inference at the edge. And sometimes in a medical application, your patient data is private and cannot be sent out. So basically you can personalize stuff within the edge where the more deeper inference, more deeper context happening at the cloud.

What do you hope to take away from the event?

Bassel Haddad: Yeah, here at embedded word, we have the opportunity to meet a lot of our customers that I’ve been working with over more than a decade on deploying edge technologies and solutions and edge solutions. And it would be great to meet them and discuss with them innovation that happening based on our product. And around these products that we launched, this customer had been working with us for the last few months to really take those products and create real world solution that solve real problems around edge AI, and the approach we took with our products in terms of built-in AI acceleration and the scalability of the platform enable them to come up with solution that are more sustainable, more power efficient, and drive-optimized TCO, total cost of ownership. So, pretty excited to go around the booth, and some of them are in our Intel booth, and some of them are actually featuring those solutions on their own. So look forward to visit them and see all of these exciting solutions.