Simulating the Human Brain: The Blue Brain Project

Blue Brain

In the previous article, we’ve looked at some of the logical barriers to accurate simulation of the human brain. We have divided the possibilities in to two tracks: software simulation and neuromorphic hardware simulation. This article is going to look in detail at the best shot we have of accurately simulating the human mind through software – the Swiss government-funded Blue Brain Project.


The Blue Brain Project has been running since 2005. Its aim? To accurately simulate a mammalian brain at the molecular level. In other words, to build a simulated brain from the individual neurons up.

But hold on, I thought we suggested that this was impossible? The previous article asserted that software layers suffered from an inability to deploy parallelism – a fundamental feature of human-type cognition. Well, the Blue Brain Project researchers may offer a solution.

The fundamental idea

The area under focus for simulation is the neocortex – the region of the brain responsible for conscious cognition. By focussing solely on neocortical modelling – specifically, rat neocortices – the project rapidly met its aims. But three years after starting, the Blue Brain Project had successfully simulated the cognitive action of 10,000 neurons – or 100,000,000 synapses (100 million). By 2011, 100 of these ‘neocortical columns’ were in parallel operation – not in real-time, but through the efforts of distributed supercomputing.

The future

The Blue Brain Project aims to simulate a whole human brain using the tenets established by rat neocortex simulation. Just as wireless and all-in-one printers are themselves an evolution from the original, wired variety, so incremental improvements in the simulation parameters may yield staggeringly progressive results.

The first aim of the continued project is to establish a functional model with one million cortical columns – the equivalent of a thousand rat brains. The researchers assume that such a scaling effect will yield a rise in cognitive capability akin to nascent human consciousness.

The second aim of the project is to move from individual neuron simulation to simulation of the molecular structures that go in to forming these neurons. This is a considerable number of degrees more complex, but accurate molecular simulation would allow for proper mapping of the evolution and learning behaviours of individual neurons in a network – a key feature of human learning.

When can we expect all this?

Henry Markram, director of the Blue Brain Project spoke at a TED conference in 2009, claiming that such accurate human brain-modelling was achievable on a ten-year timescale. Specifically, the researchers point towards a hundred billion cell simulation level by 2023. That’s in-line with Rat Kurzweil’s predictions for the molecularly accurate mapping of human brain functions by 2020.

Of course, there are objections – the assumption that consciousness is linearly scalable as an emergent effect due to neuronal density in the human brain is questionable. Perhaps there is some ‘special element’ to the structure of the human brain that makes our kind of cognition possible. That said, Markram was keen to suggest that a brain simulated in such a manner would “speak and have an intelligence and behave very much as a human does”.

Personally, while I am swayed by Markram’s argument I suggest that other critical parameters – such as the role of non-cognitive emotions in framing problems – will play an extremely subtle role in the development of human-esque cognition. While we may be able to accurately stimulate the traditional ‘five’ senses, it’s a little less clear to me as to how we might encode an emotional memory to make sense of this incoming data. That’s a problem for the future – but I believe it may prove more fundamental than expected.

In the next article, we’ll take a look at the best attempts by neuromorphic simulation research to accurately simulate the human brain.


About the Author - This article has been submitted to us by Dell