Series 14 – Episode 10 – How AI is enabling safer and more affordable autonomous driving

Sulkes, with over twenty years of experience in high-tech industries, highlighted the rapid technological advancements within the automotive sector, particularly over the past decade.

Sulkes began by introducing Hailo, a company dedicated to creating specialised hardware for AI applications at the Edge. Founded over seven years ago, Hailo focuses on enabling AI in everyday devices like smartphones, laptops, cameras, and notably, cars. The company's mission centres on providing cost-effective and power-efficient AI solutions, crucial for devices with limited power resources.

Hailo's core technology, the Structure-Defined Dataflow Architecture, is a significant breakthrough. This semiconductor architecture allows efficient processing of neural networks, essential for AI, by integrating compute, memory, and control logic within a single chip. This innovation enables various AI tasks, such as object detection and large language models, to run efficiently on compact processors.

A notable development from Hailo is the Hailo-10 AI accelerator. Designed with large language models in mind, Hailo-10 caters to applications requiring substantial memory and processing power, such as those involving transformers. While not intended for massive models like ChatGPT, it supports relatively compact language models for Edge devices, making AI capabilities accessible in everyday scenarios, including in-car applications.

Sulkes emphasised AI's transformative role in the automotive industry, noting its integration into vehicles for enhanced safety and automation. AI already assists drivers by alerting them to lane deviations, pedestrians, and other potential hazards. The industry is now moving towards more autonomous functions, with significant progress seen in markets like China. AI's critical role lies in sensing and perceiving the vehicle's environment, providing a 360-degree view, and detecting objects beyond human capability.

However, the journey towards full autonomy presents challenges. Sulkes highlighted limited experience in planning and acting – key components of autonomous driving. Unlike sensing, which can be objectively evaluated, driving decisions involve complex, dynamic scenarios influenced by various factors. Additionally, public opinion, legislation, and insurance pose significant barriers to widespread adoption.

Hailo addresses these challenges by making AI applications cost-effective and accessible. By reducing costs, Hailo aims to bring advanced AI capabilities, previously available only in high-end vehicles, to mid-range and volume segments. This approach enables automakers to introduce desirable autonomous functions while maintaining affordability and reliability.

Recent funding of $120 million will fuel Hailo's product roadmap, supporting the development of next-generation AI products and expanding the company's global footprint. Sulkes highlighted the competitive nature of the AI industry, emphasising the need for continuous innovation to stay at the forefront.

To hear more about Hailo and much more, you can listen to Electronic Specifier’s interview with Yaniv Sulkes on Spotify or Apple podcasts.