MIT student develops brain-inspired chips to slash AI energy use

MIT PhD student Miranda Schwacke is developing neuromorphic computing devices that mimic the brain’s energy-efficient processing to tackle artificial intelligence’s massive power consumption problem. Her research on electrochemical ionic synapses could help create AI systems that process and store information in the same location, dramatically reducing the energy required for machine learning compared to traditional computing architectures.

Why this matters: Training large AI models consumes enormous amounts of energy, while the human brain operates far more efficiently when learning new information by processing and storing data in the same neural locations.

How it works: Schwacke’s devices replicate brain synapses using materials that can be “tuned” to adjust conductivity, similar to how neurons strengthen or weaken connections.

• Her current research focuses on understanding how magnesium ions inserted into tungsten oxide change the material’s electrical resistance.
• The tungsten oxide serves as a channel layer where resistance controls signal strength, much like synapses regulate signals in the brain.
• Unlike traditional computing that moves data back and forth between processing and storage, these devices handle both functions in one location.

The technical challenge: Schwacke is bridging two distinct scientific fields—electrochemistry and semiconductor physics—to create these brain-inspired devices.

• Her team was the first to use magnesium ions in this type of device, drawing inspiration from magnesium battery research.
• The main hurdle involves correctly interpreting complex experimental data to understand how the devices actually change conductance.
• She collaborates across disciplines including neuroscience and electrical engineering to overcome technical barriers.

What her advisor says: “This is electrochemistry for brain-inspired computing,” explains Bilge Yildiz, Schwacke’s advisor and the Breene M. Kerr Professor at MIT.

• “The energy consumption of computing is unsustainably increasing. We have to find new ways of doing computing with much lower energy, and this is one way that can help us move in that direction.”

Recognition and impact: Schwacke received MathWorks Fellowships from MIT’s School of Engineering in both 2023 and 2024 for her work using MATLAB in critical data analysis and visualization.

Beyond the lab: Schwacke actively engages in science communication through Kitchen Matters, a graduate student group that explains scientific concepts using food and cooking.

• The group creates educational videos and participates in outreach events like the Cambridge Science Fair.
• Past projects included explaining cookie texture science through gingerbread house construction and demonstrating pH concepts with cabbage juice indicators.
• She also served in leadership roles for MIT’s Graduate Materials Council and previously led science workshops for young women at Caltech.