Scientists have taken a big step toward making faster and more energy-efficient computers by combining light and electricity in a new kind of computer chip.
These advanced “photonic” chips, described in two new studies published in Nature, could help meet the growing demands of artificial intelligence (AI) while cutting down on energy use.
Traditional computer chips use electricity and tiny electrons to process data.
But with AI models becoming bigger and more complex, these chips are struggling to keep up—and they use a lot of power.
Photonic computing offers a promising alternative by using light, or photons, instead of electrons to perform calculations.
Light can move faster and generate less heat, making it ideal for high-speed, low-energy computing.
One team, led by Bo Peng, introduced a new system called PACE—a large photonic accelerator made up of over 16,000 tiny components arranged in a 64 by 64 grid.
This system can perform calculations extremely fast, reaching speeds up to 1 gigahertz (GHz).
PACE showed an impressive 500 times improvement in reducing delay (known as latency) compared to earlier, smaller designs. It also solved complex problems known as Ising problems, which often show up in scientific and real-world tasks.
In a separate study, Nicholas Harris and his team developed another type of photonic chip designed to handle AI models with high accuracy.
Their chip includes four large matrices that can run well-known AI programs like BERT (used for understanding language) and ResNet (used for recognizing images). The results were just as accurate as traditional electronic processors.
Even more exciting, the chip was able to do creative and fun tasks like writing Shakespeare-style text, analyzing movie reviews, and even playing classic video games like Pac-Man.
These examples show that photonic chips could have many real-world uses, from helping computers understand human language to powering smart devices.
Both teams believe their photonic systems can be scaled up for even bigger and more powerful uses in the future. More fine-tuning is still needed, but the potential is clear.
According to an expert who wrote about the findings, these breakthroughs suggest that after decades of research, photonic computing is finally becoming a real option.
Using the power of light, future computers could be not only faster but also much more energy efficient—just what we need as AI technology continues to grow.
