|
In a significant step toward building machine learning infrastructure beyond Earth, Google has successfully tested its Trillium-generation Tensor Processing Units (TPUs) in a particle accelerator to simulate radiation conditions found in low-Earth orbit.
The development marks a significant milestone in Project Suncatcher, Google's ambitious initiative to explore scalable AI computing systems in space. In a social media post Google CEO Sundar Pichai stated, "Early research shows our Trillium-generation TPUs (our tensor processing units, purpose-built for AI) survived without damage when tested in a particle accelerator to simulate low-earth orbit levels of radiation" According to the Google CEO, the TPUs, specialised chips designed specifically to accelerate artificial intelligence workloads, showed no signs of damage during the radiation exposure tests. This suggests that Google's advanced hardware may be capable of surviving the harsh environment of outer space, where electronic systems are exposed to intense radiation and extreme temperature fluctuations. Project Suncatcher aims to harness the immense, untapped power of the sun by eventually deploying large-scale AI compute systems in orbit. The idea draws inspiration from Google's history of moonshot projects, efforts that push technological boundaries in pursuit of transformative innovation. The company noted that the sun emits more power than 100 trillion times humanity's total electricity production. This potential energy source could one day make space-based computing more sustainable and efficient. However, Google acknowledged that several significant engineering challenges remain before this vision can become a reality. Among them are thermal management, controlling heat in an environment with no natural convection, and ensuring the long-term reliability of systems operating in orbit. Further testing and technical breakthroughs will be necessary as the company moves forward. Google plans to launch two prototype satellites in partnership with Planet Labs by early 2027, marking a significant milestone for Project Suncatcher. As the countdown to launch begins, Google's experiment represents not just progress in AI hardware but a bold step toward redefining where and how machine learning could operate, literally reaching for the stars. (ANI)
|
