Putting artificial intelligence hardware into orbit may sound futuristic, but one of the biggest problems facing Google’s Project Suncatcher is surprisingly familiar: heat.
AI processors generate substantial amounts of heat while performing intensive calculations. On Earth, data centers use sophisticated cooling infrastructure to prevent chips from overheating.
Space works differently. There is no atmosphere in a vacuum, meaning conventional air-based cooling does not work in the same way.
Google says its Suncatcher project is experimenting with heat-transfer systems and radiators to move heat away from the Tensor Processing Units and release it into space.
The first orbital prototype will therefore serve as a critical experiment. Reports have said the prototype will operate its AI computing in short periods before cooling becomes necessary, highlighting the gap between a technology demonstration and a commercial data center.
A commercial AI facility would need to run continuously and process enormous workloads. A system that has to stop regularly for cooling would not yet meet those requirements.
Nevertheless, the experiment could provide valuable information. If engineers can develop more efficient thermal systems, future satellites could potentially carry much larger amounts of computing hardware.
Google’s October mission is therefore also a test of a new approach to space thermal engineering.
Source: Google Research
आपके लिए सुझाव
