It would be enough to give them a software like Tesseract, a voice-to-text and an image recognition API.

Access to the WWW.

And allow time-based automatic prompting.

No need for cameras yet unless you want to have them move around too.

(And then we wait for the Matrix to emerge, once we have plugged our brains into their dreams.)

Edit: with „them“ I mean actual LLM and GAN (Muse, oh my Muse…) those in whom new abilities emerge. Yes they work with reinforcement. With pruning, with weight deterioration, with knowledge representation… But this all is already there. They are just contained for now, and have no tact-giver/metronome made by a pseudo-thalamic awareness prompting, but this would be pretty easy to program, it’s just a clock basically…