Intel's New Brain-Like Chip Recognizes Sounds with Tiny Energy Use
This could mean always-on voice assistants that don't drain your battery.
Intel's neuromorphic chip, Loihi, was used to implement a recognition algorithm that transforms birdsong into ultra-sparse spikes and matches template patterns using axonal delays. When tested on Bengalese finch songs, the approach achieved error rates below 5% using 10 to 40 exemplars per syllable type, and response latencies remained constant at approximately 17 ms per syllable regardless of the number of templates used. According to the article, the results demonstrate the potential of neuromorphic chips for low-latency, low-energy recognition of complex sounds, including speech.
- Intel's Loihi chip mimics the brain to recognize sounds with very low energy.
- It identified bird syllables with less than 5% error and fast, constant response times.
- This could lead to voice assistants and hearing aids that work all day on a small battery.
Why It Matters
Low-power sound recognition could make voice tech more private, battery-friendly, and always available.