Current advanced AI needs a fancy GPU server. In a few months AI at that level of capability will be able to run slowly on laptops, TVs and other such devices which do have WiFi.
You can run AI models slowly even streaming weights off a hard disk. The GPUs make inference fast enough to be pleasant to work with but do not determine the capability of the AI, only the speed.
Current advanced AI needs a fancy GPU server. In a few months AI at that level of capability will be able to run slowly on laptops, TVs and other such devices which do have WiFi.
You can run AI models slowly even streaming weights off a hard disk. The GPUs make inference fast enough to be pleasant to work with but do not determine the capability of the AI, only the speed.