Edge AI Thrives on Industry Collaboration

The Importance of Collaboration in Edge AI Development

According to Nebu Philips, Senior Director, Technical Product Marketing at Synaptics, collaboration is essential for the expansion of Edge AI development. Edge AI application developers are striving for optimal processing performance, but there's a scarcity of appropriate silicon within the desired performance range. The availability of suitable development kits is also limited, and as the market continues to evolve, customers may be hesitant about vendor lock-in. It is essential to promote the use of open-source technologies. If IoT vendors collaborated to overcome these obstacles, more customers could gain from cutting-edge Edge AI technology, leading to a more robust and scalable Edge AI ecosystem. This is where co-opetition, rather than competition, is beneficial. Co-opetition refers to a situation where rivals compete for customers while simultaneously cooperating to grow the overall market.

The Edge AI market is expected to grow, but this growth could be accelerated significantly if players in the IoT ecosystem worked together on several fronts. Recent technological advancements have made it possible to execute remarkably small AI models on low-power processors, frequently in conjunction with highly efficient AI accelerators, while using very little power and producing increasingly impressive results. Edge AI is now more adaptable, affordable, and widely applicable because of this fast-paced technological advancement. However, perception hasn't kept pace with progress. What is currently doable frequently surpasses what customers believe to be. Many people still think Edge AI is too difficult, expensive, or power-hungry for their needs, but it can be more simpler and approachable than they realize.

Overcoming Challenges in Edge AI

Educating the market about Edge AI's potential will happen naturally. The more difficult task is getting suppliers to agree. Silicon vendors are having trouble keeping up with developers' demands, and the industry as a whole lacks scalable silicon portfolios designed specifically for today's Edge AI applications. AI inferencing capability can be expressed in tera-operations per second (TOPS) to look at it from a processing standpoint. Some current solutions offer single-digit TOPS, while others offer high teens or greater. However, there is a sweet spot for AI acceleration for IoT devices in the gap between these extremes. These are frequently vision-based applications. For silicon vendors, this gap presents a chance to offer affordable solutions in this underserved area.

The lack of development kits for Edge IoT and Edge AI is one of the most disregarded obstacles in Edge AI. As a result, developers frequently choose a platform that is undoubtedly excessive. This is justifiable, since application developers experimenting with Edge AI will want assurance that their applications will function. It is therefore logical to select a sufficiently powerful platform, ideally one supported by a well-known development kit. The issue is that the prototype will be optimized for a platform that is overpowered for the application, making it too expensive. An AI solution like that will not be able to be scaled down to run on a less powerful but still capable silicon platform that is more economical for an Edge IoT application.

The Need for Accessible Tools

It is essential to utilize a development kit that enables the creation of a design that can be immediately put into production to create a business solution. The developer may simply give up there in the worst situation, failing to understand that a viable solution is definitely achievable. But imagine a customer who is driven enough to find a workable solution. Even in the best scenario, they will have wasted time and money creating an AI application that cannot be scaled down. The market clearly needs accessible, user-friendly tools that enable developers to move their solutions to silicon that is appropriate for their applications.

Finally, there is ambiguity in software development environments when incorporating AI functions into application workflows. Should an Edge AI developer use Linux or RTOS? C++ or Python? A developer may create an application and never update it, grow the portfolio with related products, or release a new version. In that instance, the decision has few repercussions. But how probable is that? Any ambitious developer will probably broaden their product line, release updates, and introduce new product generations over time. Developers may be concerned that they will be stuck in an environment that is unable to keep up with market changes given the extraordinary rate of AI and machine learning (ML) innovation. For this reason, Synaptics believes that an open-source strategy will always be more advantageous for developers of IoT and Edge AI applications. Unfortunately, the industry lacks a genuinely comprehensive, open-source development environment. This issue has to be resolved, and the sector as a whole needs to take responsibility for doing so.

The AI technology stack is already intricate. Each layer introduces a multitude of solution providers that can offer access to curated datasets or model creation. There are optimization firms, toolchain vendors, and businesses with diverse strategies for managing the device lifecycle of AI-enabled applications. It is a lot for a potential Edge AI application developer to handle. Perhaps it is time to offer resources to assist individuals entering this rapidly evolving sector.

According to Gartner, the market for AI semiconductors is expected to reach $159 billion by the end of 2028. Although this number includes all semiconductors for all AI applications, Gartner predicts that the market's compound annual growth rate will reach 24%, with a substantial boost coming from AI-based applications transitioning from data centers to PCs, smartphones, edge devices, and endpoint devices. AI use was already increasing, but now that AI has swiftly become essential to Internet search, demand across industries is predicted to increase even faster. Consumer electronics can clearly benefit from the addition of Edge AI. Edge AI is being used by other industries to improve operational effectiveness and cut latency by processing data closer to its source. Edge AI is being used in healthcare, telecommunications, and agriculture, as well as robotics, industrial automation, and other manufacturing systems.

The IoT market is prepared to incorporate AI/ML, but it appears that the sector is not in a position to take advantage of this interest. We have determined what is required to grow a market: suitable platforms, accessible tools, and strong support from an open source community. This is a call to action for the IoT and Edge AI ecosystem: let's collaborate to lower barriers, close gaps, and fully realize the potential of Edge AI. We are eager to hear from you!