Based off this question, since it also involves brain uploading.

My Magnificent Bastard villain is a cognitive neuroscientist with his own lab who has interest in knowing how brains would work in a computerized environment. For that, he has built the following cycle:

  • Murder an individual and take their central nervous system out;
  • Connect the CNS to a high-end mainframe that will fool the spinal cord into thinking it's connected to real body nerves. The mainframe preserves the CNS on a long-term basis;
  • He will not bother studying the "whys" of brain functioning; he puts together a hardware device, which is part of the mainframe, that will simply convert chemical and electrical brain activity into binary data and vice-versa. This process is black-boxed;
  • He puts together a robotic construct capable of communicating with the mainframe through a mobile network. The construct contains artificial nerves throughout the body, which will respond to incoming and outgoing data from and to the mainframe. This is because the villain doesn't want to risk having the CNS inside the construct, since he personally considers human brains the "epitome of general-purpose problem-solving devices";
  • He constructs an artificial neural network powered by a genetic algorithm, which evolves the neural network, based off the binary data input, into a human being. The algorithm designs the structure of the neural network into being able to learn through supervised learning; the villain himself will manually tell the neural network whether its behaviour is human, by watching how the robotic construct behaves. This process takes a very long time, obviously;
  • After the genetic algorithm reaches local optima, a new brain is needed and the cycle loops over.

We can assume the following points:

  • The mainframe's processor is capable of processing data in humongous proportions; the villain might even plan to connect several CNSs to the same mainframe.
  • He will do anything to make a mobile network capable of transmitting the necessary data between the robotic construct and the mainframe. If he can't use a current mobile network service, he will come up with his own;
  • The mainframe generates additional binary data based off the transmitted data between mainframe and robotic construct. Data acquired by each brain is kept for the villain's research purposes and is kept isolated; the data must not be mixed;
  • The villain's reputation in the scientific community is extremely high; he is regarded as a lead figure in the cognitive psychology field, gives numerous speeches and has written several highly-acclaimed papers. He's middle-aged, very charismatic and has a well-toned body.
  • His true objective is to build a new human species by giving them the "immortality" specified in the previous points and, hence, enough time for them to find a way to "transcend" themselves into a superior human race. The villain is willing to die for this cause and will gracefully accept them as superior beings.

Of course, nothing in this is possible without a reliable mobile network. How can the villain find or come up with a mobile network fast and efficient enough for his objectives?

  • $\begingroup$ There's lots of bandwidth guesses about the spinal column. They're all guesses because science doesn't really know for sure the best way to convert the signaling of neurons in to a binary bandwidth measurement. However, for everyone's reference, the best answers I have found for this question is that the spine can manage 10GB/s and a minimum of about a 35ms latency. $\endgroup$
    – Cort Ammon
    Commented Sep 15, 2016 at 14:34
  • $\begingroup$ de-centralizing your system might be an interesting approach: let the various subsystem of your robot be semi-autonomous: you don't need to send every bit of motion command to your leg. just tell it where you want it, and have the semi-autonomous system handle the details. This will greatly reduce the bandwidth requirement, and kill most of the latency issues. $\endgroup$
    – Burki
    Commented Sep 16, 2016 at 11:54

3 Answers 3


Bit rate

The paper The Limiting Information Capacity Of A Neural Link says:

the mean frequency would be about 670 impulses per second and the average information content about 4.3 bits per impulse.

That would be 4.3*670 bits / second = 2881 bits/second.

We know that there are 24 cranial nerves and 30 spinal nerves. So we have to connect 30 + 24 = 54 streams of data of 2881 bits/second = 2881 * 54 = 155574 bits/second = approx. 19.45 kB/second.

Edit: later in the paper the figure of 6 bits per impulse in introduced, givin us: 6*670 bits / second = 4020 bits/second. So, 54 streams of 4020 bits/second = 54*4020 bits/second = 217080 bits/second = approx. 217.1kB/second.

It is not that much. It is over GSM bit rate, but any 3G, WiMAX or LTE connection will do.


According to the question How long does a signal from the brain take to reach the limbs? the latency of the nerves is between 100 ms to 300 ms.

According the research Measuring mobile broadband performance in the UK by ofcom.org.uk latency in 3G connections is between 63.5 ms and 85.1 ms.

So, again, any 3G connection will do just fine.


Bandwidth isn't the killer problem, latency is.

Bandwidth Requirements

There are 640 muscles in the human body, each with a range of values from totally relaxed to totally constricted, zero to one. Floating point values are stored in 32 bits. 640 * 32 = 20480 bits/second. Assuming that the same amount of data needs to be uploaded as download, then we're talking about 5 kilobits per second. Add required overhead such as packet headers and you're to 10Kbps. But....muscle state isn't enough. To work the construct, we will also need continuous streaming video/audio as well as sensory data to the mainframe. Netflix recommends a minimum of 3.0 megabit/second for SD quality streaming video. They recommend 25 megabit/second for Ultra HD quality streaming. Verizon claims that LTE can carry between 5 and 12 megabits/second. Under ideal circumstances of full bars in a dense urban environment, the construct could stream SD quality video back to the mainframe. During periods of high usage on the LTE network (outside of the construct's control), streaming video may stutter or stop, effectively rendering the construct blind and/or deaf. Kind of a big problem if the construct is driving a car or operating heavy machinery.

Or, if not the public cellular networks, our villain could make his own wireless network and build it specific to his needs.

Latency kills

The problem with remote administration of a construct without any kind of central nervous system is the time it takes to get state info from the construct back to the mainframe, process it and send commands back. This is true, even if the villain builds his own wireless network.

There are several sources of latency:

  • Video processing: It will take some amount of time to translate the binary data coming from the construct into electrical impulses on the optic nerve.
  • Audio processing: Same as visual processing. It will take some time to translate.
  • Sensor synthesis: If any amount of processing needs to be done on the sensory data coming, this will add to the delay.
  • Network latency (the speed of light gets you every time)

Note that the fastest recorded reaction time for a human being was 0.101 seconds and a median reaction time of 0.266 seconds according to HumanBenchmark.com. If the combination of the construct, cellular network, mainframe and CNS cannot respond in less than a quarter a second, the construct will appear clumsy, uncoordinated and slow.


I guess the obvious solution here would be to have multiple mobile connections between host and client. This increases the bandwidth and provides for some redundancy in case one or more of the connections drop for some reason.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .