I hypothesise that if AGI arrives, it is perhaps more likely to be due to multimodal synthesis of sensory information from cameras, kinetic sensors, motion sensors, spatial awareness, and tactile sensors of various kinds. That means that AGI may be more likely to arrive in a humanoid GPR wired up to a leading #multimodal semantic vector model.
When the AI starts asking you to point the camera out the window so it can have a more interesting view, or the robot wanders off to find and watch butterflies: then you might be talking to AGI.
I don't think it will necessarily require the emulated equivalent of a limbic system for AGI to happen, but I wonder if something like that would emerge along with proprioperception, audio-visual stimulus, and interoception. Our lizard and mammal brains cause us as much trouble as good (especially the lizard brain) and perhaps without emulation of something like them the AGI will always be different (why would we want an AI-humanoid to have a lizard brain?), and as I said long ago: an alien.
