According to @godofprompt on Twitter, researchers at Aalto University have published a breakthrough in Nature Photonics on November 14th, demonstrating a method called single-shot tensor computing that enables AI to run entirely on photons instead of electrons or GPUs. This approach encodes data directly into the amplitude and phase of light waves, allowing all tensor operations to occur in parallel at the speed of light, rather than sequentially as with traditional GPUs. The method eliminates the need for electronic switching and active control, drastically reducing energy consumption and heat generation. Professor Zhipei Sun confirms that this technology is already compatible with current photonic chip platforms, and Dr. Yufeng Zhang estimates that real-world integration could occur within 3-5 years, potentially making trillion-parameter AI models sustainable and highly efficient by 2028. This breakthrough opens significant opportunities for developing energy-efficient, high-performance AI hardware, transforming data center operations and enabling new scalable AI applications. (Source: @godofprompt, Nature Photonics)