Part 5/12:
AI expert Phil Beasel analyzed the demo, explaining that Optimus likely runs two main AI models: one for motion control and another for communication. The motion AI manages physical movements via sensor input, functioning similarly to Tesla's full self-driving system. This model probably operates on an onboard inference chip, such as Tesla's FSD hardware or an advanced AI chip, to handle real-time actions.
Conversely, the communication AI, responsible for voice recognition and response, seems to operate in the cloud, utilizing Tesla's Grock system. While offloading tasks to the cloud reduces onboard compute load, it introduces latency, which might explain some delays in Optimus' responses during the demo.