> Humans communicate in much more complex ways than computers. Facial expressions, intonation, language, body language, etc. All things we do not quite understand.
That's the medium - but how many bits of information, after compression, are exchanged in a conversation between humans?
It's true that facial expressions and body language are difficult to measure with a camera, that intonation and speech are difficult to measure with a microphone, and that duplicating them with an artificial face on a monitor, a robot, or a speaker are all very difficult programs for humans to write.
But this isn't what Musk was trying to communicate. He's saying that the amount of information contained in the combination of these human communication mechanisms can be compressed to about 10 bits per second.
Imagine us talking for a minute, compressed to a movie script. Your wetware could easily use my expressions to come up with an estimate of my happiness, sincerity, confidence, truthfulness, sarcasm (and so on) on a scale of 1-10. These values might change from sentence to sentence. If you're very observant, in 60 seconds you might get 10 estimates for each of a dozen emotions with 3-bit fidelity from these observations - only a few bits per second of actual information transferred. And I might be creating sound waves that could be recorded at very high bit rates digitally, but aside from a one-time characterization of my voice, I would perhaps speak 200 English words in that minute. 200 words, averaging 4.5 characters each, with a 20% compression ratio, is only on the order of 4 bytes per second. And most of that is overhead for grammar.
Or imagine meeting two aliens in a wormhole/portal that's only going to be open for 60 seconds. You attempt to convey as much about our society to them as possible vocally, while I'll attempt to use a computer. Afterwards, you might be able to have politely (to a human) said "Hello, we come in peace" and smiled, be able to describe the sound of their unintelligible speech, and be able to make a sketch of their appearance. I'd have recorded the entire exchange in 4k/60 fps video, from which we could probably reconstruct the anatomy under their equivalent of skin, and my only communication lag would have been gesturing that I want to give them computer equipment containing Wikipedia and Archive.org and whatever else I can download, transferring it at a few pounds of multi-terabyte hard drives a second, and get computer equipment back at the same rate. Or if you want to remove the sneakernet exploit, fine, we can hook up a Thunderbolt cable and send it across, or even just play individual frames of text or images on high-resolution monitors and record that video. From that, we could learn an incalculable amount about each others' species. From a minute of talking, we would learn next to nothing.
The medium is irrelevant. The question is how much information can be exchanged. And computers win that with no contest.