VectorWave appears on Electronic Engineering Journal’s Fish Fry podcast

In one of VectorWave’s most candid and comprehensive chats to date, CTO and founder Dr Ron Davis fielded questions from EEJ on VectorWave’s neuromorphic analog compute platform and how it compares to traditional digital processing.

“Say you want to perform AI inference on a wireless signal using a traditional digital processor,” Davis explains. “You have your signal, receive it with your antenna. It goes into the wire, still analog. You clean it up with your RF front end, still analog. Then you digitize it into ones and zeros.”

“You store the data of the waveform into the memory and then move that data back and forth between memory and compute,” he continues. “The problem with this bottleneck is that digitizing the RF waveforms can produce up to gigabits of data every single second per antenna. That’s too much data. It clogs the digital pipes. It’s not physically possible to perform digital AI inference on all of that data, let alone do it in real time or with low power.”

Davis and Dalton also chat about what it really means to build AI-native connectivity rather than simply adding AI features onto existing wireless systems.

“How do we bring this AI intelligence to RF?” Davis asks. “The software exists, but the hardware doesn’t exist for this. So, for AI-added approaches, there’s this incompatibility between the software and the hardware. These RF neural networks and data you have to digitize are just too computationally expensive to actually deploy on today’s hardware.”

Listen to the entire podcast to learn more about the challenge of getting more bandwidth out of an increasingly crowded spectrum, as well as Davis’ love of Street Fighter and BBQ.