I agree with most of the article but this point I disagree with. Non verbal communication is a huge component in how we interact with each other.

Covid example: When on calls where no cameras are on I get far less feedback when presenting to a 'room' compared to if I was to present in person or have cams on, even if the room stay silent in both situations.

By looking at faces I can see who is distracted, who looks confused and whether what I am saying is being received well or poorly. Did that joke get smiles (polite or genuine ones?) or eye rolls?

Now - do I think the current SOTA algorithms are at this level of nuance? No, definitely not. But to say it isn't at all possible is ridiculous in my opinion