NewsNTech
The constraint in automotive AI sits at the edge of the stack.
Running useful AI inference inside a vehicle means working within a fixed thermal and compute envelope: the system-on-chip in an infotainment unit is not a cloud server, and any model that leans heavily on cellular connectivity degrades the moment a vehicle goes underground or into a rural dead zone.
General Motors plans to navigate that constraint with a proprietary in-vehicle AI system it expects to ship later this year, one the company describes as better tailored for its customers.
Why "proprietary" changes the economics Third-party AI integrations typically route queries through a shared model that the automaker neither trains nor fully controls.
Keep reading