Beyond Boundaries: The New Neuroscience of Connecting Brains with Machines—and How It Will Change Our Lives by Miguel Nicolelis

Score: 3/5

This book talks about a lot of strange phenomena with our brain. E.g. the “phantom limb” and which region it is tied to in our brain, whether we can perceive threat from the harm done to a mannequin, and whether our perceived sense of self changes over time and when we use tools.

A lot of them resemble what I’ve learned when doing ML. It is an interesting read, although I have to admit I cannot remember most of the interesting facts mentioned in the book.

The main takeaways are: our brains are biased towards things that look human. The internal image of a person matters a lot for phantom limbs. And any kind of cognitive information is distributed across populations of neurons.


  • In fact, if just one-sixtieth of the primary visual cortex remained, the animal would retain a visual-motor habit it had learned. Faced with simple tasks, the brain was amazingly resilient in handling sensory information…Yet, Lashley also had found that the brain was less able to recover from damage when faced with more complex behavioral tasks.
    • This is a drop out rate of 94%…
  • Coding a complex neuronal message or task into a large number of small, individual fragments or actions is similar to the work of an orchestra
    • The opposite of one-vector embedding we have
  • Few areas were left uncharted, but how a whole brain worked remained a deep and obscure mystery. After dividing and subdividing the brain into its minute units, neuroscientists still lacked a way to explain how those units came together to produce the seamless perceptual experiences that define human life.
  • Curiously, male transvestites who have undergone sex reassignment surgery do not experience a phantom penis, which suggests that, to their brains, these men already live in a woman’s body.
  • But when they cut the sensory nerves leading to the spinal cord, severed the nerves in the spinal cord itself, or even removed the parts of the brain that received the sensory neuronal tracts, the phantoms persisted. A patient’s pain might vanish temporarily, but it always returned—with a vengeance.
  • a significant increase in evoked skin conductance response, suggesting that the “threat” to the mannequin’s body spawned a great deal of anxiety in them. Similar effects were provoked using other parts of the mannequin’s body, such as the hands. However, if objects that did not resemble a human body were utilized, the out-of-body experience did not take place.
  • THE UNCERTAINTY PRINCIPLE OF NEUROPHYSIOLOGY One cannot define the spatial domain of a particular neuronal receptive field without specifying a particular moment in time. In other words, the spatial and temporal domains of neuronal firing are tightly coupled, defining a neuronal space-time continuum.
  • THE DISTRIBUTED CODING PRINCIPLE Any type of information processed by the brain involves the recruitment of widely distributed populations of neurons.
  • THE NEURONAL MULTITASKING PRINCIPLE Individual cortical neurons and their probabilistic firing can simultaneously participate in multiple functional neural ensembles. That means that the spikes produced by a single cortical neuron can be utilized by distinct neural ensembles to encode multiple functional and behavioral parameters.
  • THE RELATIVISTIC BRAIN HYPOTHESIS When faced with new ways to obtain information about the statistics of the surrounding world, a subject’s brain will readily assimilate those statistics, as well as the sensors or tools utilized to gather them. As a result, the brain will generate a new model of the world, a new simulation of the subject’s body, and a new set of boundaries or constraints that define the individual’s perception of reality and sense of self. This new brain model will then continue to be tested and reshaped throughout the subject’s life.
  • THE NEURONAL MASS EFFECT PRINCIPLE As the size of cortical neural ensembles grows beyond a certain large number, the amount of information embedded in the neural ensemble tends to asymptote, slowly converging to its maximum information capacity.