AI hardware design could filter out irrelevant visual data to reduce energy use
Electrical EngineeringComputer ScienceMaterials Science
THE AI ANGLE
Recognizing selected image regions through a spiking neural networkResearchers designed a reconfigurable transistor based on molybdenum disulfide and ferroelectric gating. The device switches between rule-based image filtering and spiking neural network functions. This architecture allows edge hardware to discard background image data before processing, reducing active hardware blocks and energy consumption.
THE TEACHING ANGLE
Filtering more image data saves energy, but aggressive cropping removes features needed for accurate recognition.Read the original at techxplore.com Generate teaching or study materials
More in Electrical Engineering
- Some Supabase customers are publicly exposing reams of people’s data to the webTechCrunch · September 26, 2026
- General Robotics is betting on modular intelligence, not one robot brainThe Robot Report · September 26, 2026
- Astra and Opus just passed Turing’s other testTechCrunch · September 26, 2026
- On Anthropic’s AI Misuse ReportSchneier on Security · September 25, 2026
- TechCrunch Disrupt 2026: Ricursive Intelligence’s Anna Goldie and Azalia Mirhoseini on when AI starts designing its own hardwareTechCrunch · September 25, 2026