How much would it cost us to simulate a human brain? Much more than you think

How much would it cost us to simulate a human brain? Much more than you think

It is not difficult to find a science fiction novel where a human brain has been simulated, for example by scanning it to simulate it in a virtual environment, like the Matrix , and thus transform our neurons into bits (and achieve a kind of immortality). Permutation City , by Greg Egan , is one of the best explained examples from science that I can remember.

However, if we talk about real science, is that possibility very remote? Doing scientific prospecting is always very risky, but given the current figures, it is much more difficult than it might be supposed.

The three processes

Abstract 1392 404 960 720

To simulate the human brain, in principle, we do not need to understand the functioning of human cognition, it would be enough for us to understand the basic functional characteristics of the computational elements of the brain. The problem is that, for this, we need three technologies that already exist but that require much more advanced development .

These three technologies are:

  • Scanning : microscopes with sufficient resolution and ability to detect relevant properties of the brain.
  • Translation : automatic image analysis that transforms raw data into three-dimensionally interpreted models of relevant neural components.
  • Simulation – Hardware powerful enough to create a complete simulated model.

Quantum leap

These technologies already exist, that is a big step, but they must be developed so much that it does not seem that all this will happen in the near future, as Nick Bostrom explains in his book Superintelligence :

Computational models of many types of neurons and neural processes already exist. Image recognition software has been developed that can trace axons and dendrites through a stack of two-dimensional images (although its reliability needs to be improved). And there are visualization tools that provide the necessary resolution (with a tunneling microscope it is possible to "see" the individual atoms, in a much higher resolution than required).

Despite these advances, according to Bostrom, we are still far from achieving what it takes to recreate a brain. A recent study in this regard indicated that by the middle of the year 2050 we may already have the previous capacities for the emulation of a brain, but with a large interval of uncertainty .

We cannot underestimate the amount of work that we still have to do, as we have not even been able to emulate a simple brain yet, like that of C. elegans , a worm about 1 millimeter in length that only has 302 neurons .

1024 2000

The complete connective matrix of these neurons has been known since the mid-1980s, when it was painstakingly traced by electron microscope slicing and hand-labeling of specimens. But simply knowing which neurons connect to which others is not enough. To create an emulation of the brain, we would also have to know which synapses are excitatory and which are inhibitory; the strength of connections; and various dynamic properties of axons, synapses, and dendritic trees.

We don’t even know that about a worm’s tiny nervous system, although we might soon get it thanks to budding research . Once accomplished, however, jumping up to a human brain is not trivial: it will require a surprising advance of the aforementioned capabilities .

The exponential progress of technology always gives us surprises, so perhaps in a decade we will have to rewrite this entire text completely . But optimistically assuming that we will soon simulate a human brain is still more part of optimism than science.