Dave Selinger — AI and the Next Generation of Security Systems

Deployment Challenges

Deploying AI models from development to production presents unexpected challenges, particularly with specific chipsets and operands. Dave Selinger explains that while Qualcomm's stack of operands allows for acceleration, new architectures often use new operands, leading to performance issues 1. He notes, "If you're using an operand that isn't supported on your particular architecture, we just move everything from that point forward into your CPU."

It's not this kind of like nice smooth curve where things kind of get reconnected in the middle. It's just. And it blows up.

--- Dave Selinger

Additionally, the distinction between training and runtime operations is significant, requiring extensive OS tuning and firmware management 1.

   

Edge Processing

Edge processing in AI systems offers significant advantages, particularly in security applications. Dave Selinger discusses how Deep Sentinel's vertically integrated system runs AI in the customer's home, allowing for real-time decision-making without cloud latency 2. He emphasizes the importance of processing speed, stating, "The difference between real time and 500 millisecond latency in communication is everything."

We've really focused on moving everything to the edge.

--- Dave Selinger

The hardware, based on Qualcomm Snapdragon, is designed to handle complex operations locally, reducing the need for cloud-based processing and enhancing system efficiency 3.