Reverse-Engineering the Brain

The brain is poorly understood, but adequately complex.

In the brain, sharp voltage spikes are propagated from neuron-to-neuron to communicate information (Hodgkin and Huxley [1952]). These voltage spikes are called action potentials. Action potentials can be elicited by a biological stimulus (such as eyes detecting light), or from the random fluctuation of ions crossing a neuron’s biological membrane (Diba et al. [2004]; Kole et al. [2006]; Faisal et al. [2008]). The noisiness of action potentials has led movement-based brain-machine interfaces to make use of only a small fraction of neurons that are well-correlated with a particular movement (Chapin and Nicolelis [1999]; Laubach et al. [1999]). The remaining neurons are a source of entropy. How the brain makes use of such a high level of noise is poorly understood (Dorval and White [2005]; Faisal et al. [2008]). A complete understanding would likely require detailed modeling on a molecular scale.

Today’s best recording implants can wirelessly record from hundreds to thousands of neurons (Schwarz et al. [2014]). For a brain-to-brain network to function optimally, stimulation and recording of neurons across multiple cortical layers must be engineered. Hundreds to millions of implantable, free-floating sensor nodes show promise for high-density, biocompatible brain recording (Seo et al. [2013]). Figure 1 depicts one recent pioneering innovation, aptly termed ‘neural dust’.

neuraldust

Figure 1: Ultrasound waves are used to interrogate implanted neural dust. Changes in dust composition are correlated with changes in electrical potential. Neural dust offers an effective means of extracting a large amount of information from multiple cortical layers, but may have insufficient resolution for full-scale brain recording. Theoretically, neural dust particles of a different piezoelectric composition could cause neural stimulation following electromagnetic interrogation, allowing for parallel recording and stimulation of the brain. Taken from Seo et al. [2013].

Action potentials can also be elicited through foreign electrical stimulation ([Hodgkin & Huxley 1952]). Clinically, artificial neural stimulation has proven to be an effective treatment for depression, bipolar disorder, schizophrenia, (McNamara et al. [2001]) and more (STX-Med [2014]).

In a binary model of a neuron, a neuron in the midst of an action potential is considered a ‘1’, while a neuron at rest is equivalent to a ‘0’. With greater than tens of billions of neurons in the brain, there are at least 210,000,000,000 possible states of the brain at any given moment. Given the brain is not truly binary, but non-Turing, modeling the human brain on a metabolic, molecular, and electrical scale remains a challenging computational problem (Yoosef et al. [2014]). Based on Intel’s BlueGreen experiments (Yoosef et al. [2014]), Moore’s Law indicates that it will take approximately 60 years until a computer may be capable of fully modeling a simplified, generic rat brain. Considering the significant variations from person-to-person in the upper cortex (Kelly et al. [2012]), it may never be possible to successfully model the brain of a living being.

In 2013, the world’s first brain-to-brain interface was constructed between two rats in the laboratories of Miguel Nicolelis (Pais-Vieira [2013]; Nicolelis [2015]). Sensory information was translated to motor action between one rat located in Durham, North Carolina, to another rat in Natal, Brazil. Critics of this experiment claim that information transfer rates between the two rats were not close to the information transfer rates of computers, or even modern brain-machine interfaces. This should not discourage research into brain-to-brain interfaces, but rather, signify a dangerous lack of inquiry into brain-to-brain interfaces relative to computer circuitry.

In the same year that the first brain-to-brain interface was constructed, U.S. President Barack Obama founded the BRAIN Initiative, a funding effort intended to boost the U.S. to the forefront of brain innovation (NIH [2014]). Unfortunately, our lack of understanding in neuroscience is correlated with a stark lack of funding compared to computational research. Even with the BRAIN Initiative in mind, when considering investment from both governments and private institutions, yearly U.S. funding into brain research is less than 10% of the hundreds of billions poured into computer science research (SFN [2011], Kennedy [2012], NIH [2014]). In order to reverse-engineer the brain at a rate similar to advances in computer science, these numbers must be flipped and exceeded. Cooperation is required from industry leaders, governments, and philanthropists to fund neurobiological and brain-machine interface research, particularly because the additional regulations and experimental time necessary for biological research will always exceed that of hardware and software research. So long as we continue to innovate computer circuitry and neglect biological integration, the computational abilities of machines will ultimately surpass humanity’s collective intelligence.

 

Reference: proof-of-cognition-implants , published May 2015. Disclaimer: Project Oblio’s mechanism does not rely on brain implants, but the mechanisms of action are the same. An early version of the paper provably exists in bitcoin address 13eeMVU5fXNfZdoBk5z4fEAbgSH9MawQ6H.

One thought on “Reverse-Engineering the Brain”

Leave a Reply

Your email address will not be published. Required fields are marked *