AI Breakthrough: Scientists Mimic Human Brain's Adaptability in New Semiconductor
Scientists have made a significant breakthrough in ace hardware, mimicking the human brain's adaptability. The study, published in Advanced Materials, introduces an ultra-low-energy semiconductor that could revolutionize AI. The brain's intrinsic plasticity enables neurons to adjust sensitivity based on context, a feature absent in current AI semiconductors. A team of researchers, led by Professor Kyung Min Kim of KAIST, has developed a frequency switching neuristor that replicates this adaptability. The device combines a volatile Mott memristor and a non-volatile memristor to control neuron firing frequency. It automatically adjusts responses based on neuronal spike signals and memristor resistance changes, mimicking the brain's adaptive behavior. This technology, published with the DOI 10.1002/adma.202502255, marks a substantial advancement in AI hardware's energy efficiency and stability. Simulations revealed that this technology reduces energy consumption by 27.7% compared to conventional neural networks while maintaining performance. Moreover, it demonstrates resilience, allowing the network to reorganize and restore performance even if some neurons are damaged. This groundbreaking research, primarily conducted by institutions like Stanford University, MIT, and IBM Research, paves the way for more efficient and robust AI systems. The ultra-low-power semiconductor technology brings us closer to creating AI that truly learns and adapts like the human brain.
Read also:
- Tata Motors Establishes 25,000 Electric Vehicle Charging Stations Nationwide in India
- Tesla's Nevada workforce has escalated to a daily output of 1,000 Powerwall units.
- AI-Enhanced Battery-Swapping Station in Southeast Asia Officially Opens Its Doors
- HAW Hamburg's Pilot Plant Transforms Waste into Climate-Neutral Fuel
