J
Jan Panteltje
Guest
IBM team builds low-power analog AI processor
Huge arrays of phase-change material perform in-memory processing.
https://arstechnica.com/science/2023/08/ibm-team-builds-low-power-analog-ai-processor/
paper:
An analog-AI chip for energy-efficient speech recognition and transcription
https://www.nature.com/articles/s41586-023-06337-5
So basically a neural network with programmable weights made of phase-change material that uses very little power.
Reconfiguring the weights (heating/cooling) takes a lot more power.
Voice to text chip? or voice commands to digital output signals chips coming?
Huge arrays of phase-change material perform in-memory processing.
https://arstechnica.com/science/2023/08/ibm-team-builds-low-power-analog-ai-processor/
paper:
An analog-AI chip for energy-efficient speech recognition and transcription
https://www.nature.com/articles/s41586-023-06337-5
So basically a neural network with programmable weights made of phase-change material that uses very little power.
Reconfiguring the weights (heating/cooling) takes a lot more power.
Voice to text chip? or voice commands to digital output signals chips coming?