This analysis assumes that there hasn’t already been mass deployment of generalist robots before an intelligence explosion, right? But such deployment might happen.
As a real-world example, consider the state of autonomous driving. If human-level AI were available today, Tesla’s fleet would be fully autonomous—they are limited by AI, not volume of cars. Even for purely-autonomy-focused Waymo, their scale-up seems more limited by AI than by car production.
Drones are another example to consider. There are a ton of drones out there of various types and purposes. If human-level AI existed, it could immediately be put to use controlling drones.
So in both those cases, the hardware deployment is well ahead of the AI you’d ideally like to have to control it. The same might turn out to be true of the sort of generalist robot that could, if operated by human-level AI, build and operate a factory.
This analysis assumes that there hasn’t already been mass deployment of generalist robots before an intelligence explosion, right? But such deployment might happen.
As a real-world example, consider the state of autonomous driving. If human-level AI were available today, Tesla’s fleet would be fully autonomous—they are limited by AI, not volume of cars. Even for purely-autonomy-focused Waymo, their scale-up seems more limited by AI than by car production.
Drones are another example to consider. There are a ton of drones out there of various types and purposes. If human-level AI existed, it could immediately be put to use controlling drones.
So in both those cases, the hardware deployment is well ahead of the AI you’d ideally like to have to control it. The same might turn out to be true of the sort of generalist robot that could, if operated by human-level AI, build and operate a factory.