Shining Light onto a Chip Could Give Robots a Faster AI Brain Update
Cornell Tech researchers have built a receiver that rewrites a chip's memory using flashes of light instead of electrical wires, and it could one day let warehouse robots refresh their AI models on the fly.
Key points
- Cornell Tech researchers presented a light-based memory receiver at the IEEE/JSAP Symposium on VLSI Technology and Circuits last month.
- The device rewrites a chip's built-in memory by shining QR-code-like patterns of red light directly onto photosensitive cells, cutting out the power-hungry circuits traditional optical systems need.
- An independent expert called the approach "a clever way" of tackling one of AI hardware's biggest bottlenecks.
- The team says the technology could reduce energy use in data centres, self-driving cars, and AI-powered robots.
- Larger-than-standard memory cells remain a key obstacle before the design could appear in commercial hardware.
Imagine your robot assistant needing a software update. Right now, getting new AI instructions into the robot's brain chip means shuffling data along copper wires, burning energy and slowing everything down. A team at Cornell Tech thinks light can do that job better.
Postdoctoral researcher Yifan He and associate professor Jae-sun Seo built a receiver that sits inside a chip and rewrites its own memory when a beam of light hits it. The beam carries a pattern of bright and dark squares that looks like a QR code. Each square switches a tiny memory cell on or off, encoding the numbers, called parameters, that tell an AI model how to behave.
Why does moving data with wires cost so much energy?
The short answer: distance and volume. AI models are enormous collections of numbers. A chip's built-in memory, called SRAM (static random-access memory, the fast but small storage baked into a processor), can't hold them all. So the extra data lives in a separate, larger store called DRAM (dynamic random-access memory), connected by metal wires. Shuttling numbers back and forth across those wires uses a surprising amount of electricity, and gets worse as AI systems grow.
Optical links, which carry data as light rather than electrical current, can move more data with less energy loss. The catch is that today's light-receiving circuits need power-hungry analogue components to convert light pulses into digital ones and zeroes. The Cornell design skips that conversion step entirely. Light hits a photodiode, a tiny component that turns light into electrical current, built directly into each memory cell. That current flips the cell's stored value straight from zero to one. No analogue middleman required.
"That's one of the major bottlenecks," Seo told IEEE Spectrum, which first reported the research. Independent expert Dennis Sylvester, who chairs the electrical and computer engineering department at the University of Michigan, called it "a really important problem" with "massive commercial implications".
What does this mean for robots and everyday gadgets?
For warehouse robots, it means a faster, cheaper way to push updated AI instructions to each machine without wiring every unit to a central server. For tiny microrobots, barely larger than a fingernail, it could eventually mean carrying useful AI smarts without the heavy memory hardware that currently makes that impractical.
Sylvester notes the technology sits far from shop shelves. The photosensitive memory cells are currently bigger than standard SRAM cells, so a chip using them holds less memory overall, which could cancel out the energy savings.
Seo's team is working on shrinking the cells by optimising transistor sizes. They are also partnering with optics groups to build a transmitter that can flash the light pattern millions of times per second, reaching the gigabits-per-second speeds real applications demand. The version He demonstrated in the lab emits a single, fixed 14-by-14-square pattern through a metal stencil.
How soon could this show up in real products?
Three to five years is the honest window for edge AI broadly, according to Sylvester. The memory-cell size problem means this specific technology needs more development before it can match what conventional chips offer. Think of the current prototype as proof that the idea works, not a product ready to ship.
