Deep In Tech #3 - Update on Brain-Like Microchips: Can Silicon Think Like Us?
9 snips
Sep 1, 2025 Explore the exciting world of brain-like microchips and their potential to revolutionize computing. Discover how the new Pulsar chip mimics human brain functions and its unique analog processing abilities. The discussion also sheds light on the trade-offs and limitations of this cutting-edge technology. Key applications across various fields are unveiled, making this a thrilling glimpse into the future of neuromorphic computing.
AI Snips
Chapters
Transcript
Episode notes
Compute And Memory Collocate With Resistive Cells
- Neuromorphic chips combine memory and computation at the device level using resistive memory.
- This lets the chip multiply inputs by weights in-place, mimicking neurons' compute+memory nature.
From Chip Design To Brain-Like Noise
- Anastasi recounts designing chips and supervising a startup team, contrasting deterministic silicon with noisy biology.
- She emphasizes that brains thrive on noise while engineered chips aim for perfect signals.
Tiny Spiking Core For Event Sensors
- Pulsar integrates an analog spiking core of ~1000 neurons for event-driven processing.
- That small SNN can efficiently handle sensor events like camera or microphone triggers with minimal power.