The Latest in Artificial Intelligence Trends, Technologies, and Best Practices

Most neural networks are based on predetermined architectures. During optimization, only neuron parameters are optimized, while the architecture remains predetermined, increasing network size unnecessarily. This Neuton talk shows how a different approach to neural network architectures can result in models that are both 1000x smaller and faster.

Building a small, intelligent device that can recognize words, gestures, and even people or animals is now easier, thanks to new hardware, out-of-the-box libraries, and support from the tinyML community. In this talk, Arduino co-founder Massimo Banzi provides a series of examples showing how you can create tinyML projects with Arduino.

Spatial AI technologies give autonomous robots and drones both pose awareness and a geometric/semantic understanding of their environments. In this talk, SLAMcore explores the use of Spatial AI in robotic applications, examining VI-SLAM, SLAMcore Spatial AI SDK on Arm-based systems, and featuring a demo of Spatial AI on Arm processors.