How we Communicate

        What you are about to read contains about 53,704 bits of information (encoded as UTF-8). Your computer probably downloaded it in less than a second, but it will take much longer for you to receive it. Even if you have a 10Mbps connection to my server, the real bottleneck here is the connection between your brain and mine. What is the bandwidth of that true last hop? Communications pioneer Claude Shannon calculated that the average amount of raw information (entropy) in an English word is about 11.82 bits. If you're reading 250 words per minute, you’re really only receiving 49.25bps. Seems wasteful.

Humans and Computers Communicate in Similar Ways

Communication is the transmission of information, and there are many ways to implement it. You may be surprised that digital and personal communications protocols are remarkably alike. But given that communications theory is a mathematical theory, as opposed to say a physical one, it makes perfect sense.

The OSI Model is a standard way of characterizing a communications system using layers of abstraction. Each layer should not care about the implementation of the lower ones, and there must be agreement on the configuration used at every level.


At the physical layer, the tongue and esophagus act as a sort of adjustable horn, which sends pressure waves of varying pitch by changing its shape. The receiver in the ear has sensitive hairs which resonate with the waves of pressure. Similarly, a wireless router sends high frequency bursts of specially encoded electromagnetic waves, which are received by a sensitive antenna in a laptop. The antenna is also constantly receiving thermal noise and spurious signals, but filters them out. Your ear does the same and can even quickly tune in to faint signals. If there’s a lot of ‘noise’, you either speak louder, or wait for your turn to talk. When two people try to talk at the same time, they’ll both stop and wait a random amount of time before trying again. In networking this is known as CSMA/CD. I often wonder how the brain decides on this random amount of time.

Data Link

Often the first thing to ask someone you don’t know is: “What’s your name?” Your name is your local address. If there are several people around, you can easily direct your message by mentioning the person by name. Your name was given to you by your parents and it generally doesn’t change. Similarly, your computer has a MAC address determined by the network adapter’s manufacturer, which allows packets from the router to find it, even in the presence of similar adapters. Even if everyone hears a message meant for you, only you act on it. Just as spoken statements are often predicated with the target person’s name, all MAC frames start with a header that contains their source and destination addresses.


The network layer defines how a packet flows through the network of switches and routers that make up the internet. This is where computers are beating us. We can do this to an extent: “If you see Sally tell her I’m sorry I ran over her cat.” (2 hops) but it quickly breaks down, as every hop loses information. Word of mouth is notoriously unreliable, since our brain does not store exact copies of information.

Meanwhile your computer can effortlessly send and receive through dozens of hops across the globe. Before the internet the closest incarnation of this we had was postal mail, with our home address analogous to IP address, and the path determination taken care of by the postal service. Routers take care of this with complex least-cost routing algorithms with the best paths constantly being updated and propagating throughout the network, updated based on congestion and failures. This would be very hard for a person to manage.

Can you keep a secret? Gossip was the original BGP. Information deemed valuable would only travel along select links to preserve its content and scarcity, though this was more for the sake of interest than efficiency. If news was valuable enough, it would soon spread naturally to everyone.


The transport layer allows for endpoint connection handling and reliable delivery.

“Hello, Hey!” -> TCP SYN, open a connection
Nodding, eye contact, “yeah” -> ACK, “got it”
“What? Hm? confused look” -> NACK, resend
“Ummmm” -> Buffering…
“Bye” -> FIN, close connection

Note that the protocols are optimized to reduce overhead. The bits of communication that are used simply to facilitate communication are kept to a minimum. “Hey, what, bye, thanks, yeah” are all very short, easy to say words. Also, it’s annoying to talk to someone that’s not looking at you since you don’t know if your information is getting through or not.

You’re probably also noticing how much more diverse the human counterparts are. Redundancy provides robustness at the cost of efficiency and simplicity. This is a key difference between biological and synthetic systems.

Session / Presentation / Application

I’m going to lump these together since they are usually all taken care of by the application layer protocol anyways. Up until this point you could have been communicating with someone who doesn’t even speak the same language as you. Only now do you extract the payload, the original meaning of the information that was encoded. There are many application layer protocols, and there many types of conversation. When talking to a teacher, a lover, or a friend, the purpose of the conversation is usually different, though the underlying protocols are the same, just as FTP, HTTP, and SSH all accomplish different tasks over a similar connection.

I’d say this layer also involves body language and emotion, which can annotate your words. A significant percentage of communication is non-verbal. Lots of bits are encoded in tone, gestures, and stature. Computers haven’t quite cracked this yet. Until Siri recognizes your tone, its usefulness will be limited.


Channel capacity is a function of bandwidth, power, and noise.

A disagreement is information asymmetry. If you’re not getting through to someone a common response is to increase your power level by shouting. Conversely, as you get to know someone better, channel capacity generally increases, and you can exchange more information and talk about complex topics. It feels pleasant when your data is getting through, and the receiver indicates a strong link by smiling. People that have similar brains and experiences are easy to talk to, like radios conforming to the same specification.

So long as a signal is sampled at twice its frequency, it can always be reconstructed exactly. However, there is no lossless method of inter-brain communication. Our conversations have come a long way from pointing and grunting, but they still have a long way to go. It’s easy to transmit a word, but how does one encode an idea, an emotion, or an experience? We have great respect for authors, poets, musicians, and artists who are unusually gifted at encoding that which is hard to communicate. It can be hard to “get” art because there’s often a lot to unpack. The more complex an idea the harder it is to encode, like compressing a large file. All conversions heretofore have been lossy, always holding us back from a true connection with another person.

        By its very nature every embodied spirit is doomed to suffer and enjoy in solitude. Sensations, feelings, insights, fancies – all these are private and, except through symbols and at second hand, incommunicable. We can pool information about experiences, but never the experiences themselves[…]The mind is its own place, and the places inhabited by the insane and the exceptionally gifted are so different from the places where ordinary men and women live, that there is little or no common ground of memory to serve as a basis for understanding or fellow feeling. Words are uttered, but fail to enlighten. The things and events to which the symbols refer belong to mutually exclusive realms of experience.
         –Aldous Huxley, from The Doors of Perception

I don’t think we will always exist in this realm of solitude. Imagine a world where we could directly network minds with the bandwidth that our electronics enjoy. I eagerly look forward to the day when we can finally transcend words, and writings such as this will look as primitive as the barking of a dog.

  • Derryl


    Great article, I love examining parallels like this.


    I would argue that we don't exactly wait "random" amounts of time before attempting to speak again. Instead we continuously ping the other active nodes to determine whether they're still busy or if they can receive more packets (we do this using our eyes and ears). However there's certainly some sort of circadian rhythm going on… when the would-be-speaker is suddenly reminded about the thing they were going to say, a few minutes later and with no conscious reason for it.


    Also, we can only increase our power level so much… Goku on the other hand –>

  • Jeff

    I'm a bit hesitant of increasing the efficiency of our language. Having that low information density lets me think while I'm talking. I feel like if we could "transcend language" I'd be making a lot more faux pas than if I had that buffer time. :)

  • Peter


    I agree with you a lot. I've noticed that when I'm debating it seems like I understand speech a lot better, oddly enough. But where I disagree with you is that if we made our language more efficient there wouldn't be any communication issues between anyone.

    For example, often I'm given vague instructions where the speaker assumes I hear their thoughts. This can lead some (sometimes) embarrassing performances on my part. ("Can you get that?" "You mean this?" "No, dummy, the broom not the box!")

    As simple as this break in communication is, I think we could be more efficient. Most of these problems exist in the way we speak, though, not in the limits of our communication.

  • Vlad

    Our sences pickup mutch more data that we can process ourselfs, but even the fraction that we do process IS already much larger than what modern computer can analyzer on the same level as we do in real-tme.

  • Kyle

    I'm inspired by your thought processes. Soylent aside, it's really a great collection of thoughts you have here. Your writing style is also unusually clear. It's a rare gift. Coupled with fascinating concepts, it's a pleasure to read.