Edge AI means a world of smart devices
As mentioned in previous articles on System design for the AI eraAI is more than a change in hardware and software. It is a change in design methodology that will require subsequent changes across the value chain for electronic devices and services. While much of the focus is currently on re-architecting cloud data centers, AI-driven paradigm shifts will affect everything from cloud to embedded/Internet of Things (IoT) solutions. As a result, the proliferation of AI will drive continued demand for intelligence in all electronic devices. However, as many are predicting, this path to the spread of AI will likely come in waves or cycles. The first wave is cloud, the second wave is edge devices, and the third wave will be embedded systems/Internet of Things (IoT). The electronics industry is moving rapidly from the first wave to the second, and the third wave is just around the corner. However, we are just beginning to understand the need to design solutions that are AI-native.
Designing for the AI native
Instead of designing chips, systems and applications around a set of functions, AI shifts the focus of the entire platform to be designed around being intelligent. This means using AI at its core in a safe, responsive and reliable way. Unfortunately, this requires more than just integrating a neural processing unit (NPU) or using a neural network model. It requires starting a system-level design with the hardware, software, tools, and supporting ecosystem all included in the design.
In most current electronic systems designs, AI has been added as an afterthought by integrating a proprietary accelerator, which has led to adoption challenges. For example, each consumer platform has a different accelerator and proprietary development tools. This has resulted in a limited pool of knowledgeable developers, typically OEMs, silicon providers, and major software vendors, who are able to integrate AI into system functions and applications. It is an almost overwhelming task for the average developer due to the lack of cross-platform support and a common developer environment.
The need for cross-platform, open source solutions
Astra AI’s native platform for IoT applications
Before the industry moves into the third wave, Synaptics is pushing for change in the IoT segment. Earlier this year, the Synaptics Astra IoT computing platform was introduced, leading a family of SoCs for consumer, enterprise and industrial IoT applications. Astra combines scalable hardware (MCU to MPU) with a unified software experience across the portfolio, open source AI tools and wireless connectivity. The SL Series product family announced with Astra was explicitly designed to be AI native with integrated NPUs from the ground up, along with wing cores, GPUs and DSPs, providing a balance of compute resources and memory needed to handle targeted AI workloads for applications ranging from smart home devices and security systems to industrial control systems.
Most importantly, however, it was designed with the developer ecosystem in mind using open source software, including edge-optimized AI models and developer tools. The development of open tools that can be used across platforms is a step that has not yet been reached in the cloud and edge segments, but is especially critical for the IoT segment. In this segment, the number of platforms can reach hundreds or thousands, and the number of solutions is in the millions. Consequently, this inherent fragmentation makes it difficult to achieve economies of scale on any platform or solution without such tools. The Astra platform offers a solution that can be shared by other vendors and platforms, also pushing towards AI-powered IoT solutions.
Final thoughts
While providing an open software ecosystem does not ensure adoption, it will create a collaborative environment for developers. This, in turn, should build momentum throughout the ecosystem. Tirias Research believes that this type of open developer environment is needed to unite not only the major independent software vendors (ISVs), but to enable millions of application developers who will drive innovation in the use and integration of AI, just like the store of applications. did for the mobile segment. This move by Synaptics is a prime example of how the industry needs to change the way it thinks about developing solutions to be AI native.