How to Build a Brain
Advances in brain research will soon allow us to build an artificial brain.
Posted May 31, 2012
Of course, there’s no single correct answer. Which is why all of these approaches and many more are currently being pursued. All in the interest of uncovering the mysteries of the brain and, just maybe one day, building something like it.
For instance, the IBM Blue Brain project, headed by Henry Markram, is working to model a rat’s brain on a supercomputer. Integrating data from thousands of experiments performed by researchers the world over, the project initially created a single neuron model. This incorporated information about neuron morphology, electrical properties, synaptic communication and relevant gene expression data. After this, about 10,000 of these models were linked into a simulation of a cortical column on a supercomputer. This was then used to model a simulation of 100 interconnected cortical columns. Blue Brain’s long-term goal is to eventually model an entire rat’s brain and place it in a small robot, allowing it to learn from sensory input and experience.
In contrast to this software modeling approach, some researchers are opting for the hardware route. SyNAPSE, Systems of Neuromorphic Adaptive Plastic Scalable Electronics is attempting to do just that, in part by developing neuromorphic chips. SyNAPSE is a program led by DARPA, the Defense Advanced Research Projects Agency. They’re currently working to fabricate analog computer chips based on the neural system with a million neurons and ten billion synapses per centimeter. Such an approach would benefit from highly parallel processing, much as our own brains do. The long range goal is to develop biological scale neuromorphic electronic systems for autonomous, unmanned, military robotic systems. Human-scale brain emulation under SyNAPSE appears to be on track for 2019, with Europe and China initiating similar projects. While a hardware-based approach such as this is less likely to reproduce the human brain with great fidelity, it could still lead to a highly intelligent synthetic mind.
Of course, in order for either of these approaches to succeed it would be helpful to understand the brain’s neural code in intricate detail. The new field of optogenetics is working to do just this. Using a combination of genetic and optical techniques, individual neurons can be activated and silenced in vivo with millisecond-scale temporal precision, using lasers or LEDs. This technique is proving instrumental in assessing how specific neurons contribute to brain function.
Unlike optogenetics which can study a functioning brain in real-time, the Allen Human Brain Atlas has created gene expression maps of the brain. Though the project’s primary goal is the understanding of neurobiological diseases, it will fuel discoveries in many brain-related fields. Using brain scans, medical data, genetic information and physiological data, maps of the brain’s gene expression were produced. Such a map allows researchers to correlate forms and functions.
There are a huge number of projects taking place worldwide dedicated to replicating different aspects of the brain and improving our understanding of its incredibly complex function. As computing power and realtime imaging techniques continue to improve exponentially, these will contribute greatly to our knowledge and capabilities. While target time frames may vary, it seems likely we will be building functioning brains within a decade or two.
Addendum: An important milestone was reached today (6/1/2012), when neuroscientists released the first 500 terabytes so far collected from the Mouse Brain Architecture (MBA) Project, an effort to construct the first whole-brain wiring diagram of a mouse brain.