Starting the premier trade show for embedded technologies, Salil Raje, the senior vice-president and general manager of AMD’s adaptive and embedded computing group, inaugurated Embedded World 2024 with a focus on the significance of artificial intelligence (AI) at the edge.
He emphasized that AI at the edge will play a critical role in ensuring the desired real-time performance, data security, and customization in key sectors such as autonomous driving, infotainment, and robotics. Raje added that generative AI (GenAI) systems like chat GPT have ignited unprecedented imagination in the technology industry, unlike anything seen in the past 50 years.
Raje observed that advancements in AI and robotics across healthcare, transportation, and homes were reshaping convenience and quality of life, fundamentally altering how people engage with technology.
Raje highlighted seven pivotal use cases where AI on the edge has been transformative in their development: healthcare and life sciences, smart retail, communications, smart city infrastructure, automotive technology, digital home solutions, and intelligent factories.
He provided several significant examples, expressing that AI was enabling industrial robots to learn and operate autonomously. In healthcare, AI-powered exoskeletons were making remarkable advancements. The fusion of robotics and AI was expected to restore lost limb functions for numerous individuals, vastly enhancing their quality of life.
Moving on with the automotive sector, AI assistants were on the brink of integration into vehicles. He added that such AI systems would enable users to customize their cars, from adjusting settings to making dinner reservations or accessing a user manual, all from within the vehicle.
Raje also emphasized that the growth of AI at the edge hinges on the technology industry’s ability to deliver high compute performance in limited settings. He identified three key drivers behind AI at the edge: real-time processing, data security/privacy, and personalization/customization. However, he cautioned that specific industries would encounter unique challenges.
As AI moves closer to the edge, it sparks new demands for decentralized intelligence, challenging the reliance on cloud and compute infrastructures. Raje underscored hurdles encompassing power usage, latency in data transfer, precision, environmental considerations, safety protocols, security measures, regulatory compliance, varying workloads, and frequency fluctuations.
He remarked that AI has been evolving rapidly, introducing numerous models and placing significant strain on computing resources. Managing the capabilities of these ever-evolving models presents considerable challenges, particularly due to the stringent constraints at the edge.
Raje notes that as each application evolves, the associated constraints and needs also evolve. Each use case and application imposes distinct demands on edge applications. For instance, safety may outweigh power consumption in industrial settings, while accuracy is paramount in healthcare.
The executive from AMD highlighted that healthcare challenges include safety, security, power, and data accuracy, while industrial uses face diverse workloads, regulation, safety, and accuracy. In automotive applications, paramount concerns are latency, accuracy, safety, and regulation. Designing edge AI applications necessitates flawless integration of multiple systems to ensure scalability and adaptability.
In the end, Raje revealed a potentially significant opportunity not only with the global car giant but the entire automotive industry. AMD collaborates with Subaru to develop a camera-based perception vision pipeline, integrating pre-processing, AI inference, post-processing, and functional safety into one device.
