The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Spiking Neural Nets and ML as a Systems Challenge with Jeff Gehlhaar - TWIML Talk #280
Jul 8, 2019
Jeff Gehlhaar, VP of Technology and Head of AI Software Platforms at Qualcomm, dives into the world of AI and machine learning. He shares his journey from engineering to leading AI innovations, focusing on spiking neural networks that mimic biological processes. The discussion highlights Qualcomm's advancements in AI software development, the role of federated learning in enhancing privacy, and the importance of interoperability among frameworks like ONNX. Gehlhaar emphasizes a holistic approach to AI, integrating hardware and algorithms for future innovations.
AI Snips
Chapters
Transcript
Episode notes
Early AI Work
- Jeff Gehlhaar's first AI experience involved a startup using Support Vector Machines.
- This startup work predates the deep learning era and involved government and commercial projects.
Spiking Neural Networks Research
- Qualcomm invested in Brain Corporation, collaborating on spiking neural networks research.
- This research, inspired by the DARPA Synapse Program, explored biologically accurate vision systems.
Spiking Neural Networks
- Spiking neural networks mimic biological neuron behavior, using time-coded inputs and thresholds.
- A key challenge with spiking neural networks was the lack of a stable training method like backpropagation.