Neuron-mimicking AMD chips help test role of cannabinoids in the brain

MIT simulation tracks flow of ions, long-term increase, decrease in potential for thought

It hasn't happened yet, so don't go asking your shady friends whether they've got a connection for some of the really good chip.

But researchers at MIT have figured out how to make a microprocessor simulate the operation of the human brain by switching from digital to analog and mimicking the flow of ions among imaginary synapses.

The result is a test bed "researchers" can use to play around with that behaves just like a human brain laid out on a lab table for live tests, but without all the screaming and sirens and prison.

Among the most important simulations? Not just how a human brain functions under the influence of Cannabis, but which opiates it makes itself, how to encourage it to make more and how to get it to stop using the PC speaker to call researchers "dude" while demanding munchies.

Cannabinoids affect depression? Far out, chip.

The chip built by MIT neuroscientists describe in a paper published this week in Proceedings of the National Academy of Sciences, has only 400 transistors and simulates the flow of ions between just two actual neurons. A "typical" human neuron has between 1,000 and 10,000 synapses, each of which creates, enables or enhances a different function, each of which creates a different "calculation" when connecting with any of the thousands of synapses of nearby neurons. Figure an average of 5,000 synapses and 100 billion neurons and the number of potential unique combinations is impressively large.

Even simulating just two neuron, this simulation provides a better picture of the adaptability of real brain cells better than models because it mimics the flow of ions that underlies the firing of synapses as they make specific connections, according to Chi-Sang Poon, a principal research scientist at the Harvard-MIT Division of Health Sciences and Technology, one of three researchers on the paper.

The flow of ions is the key to understanding changes in the long-term potential for energy transfer between synapses – whether the whole neuron or those synapses contain enough energy to carry out specific tasks like putting a piece of data in long-term memory or retrieving it.

According to a story in TheRegister, "a whole class of researchers" believe neurochemicals called endo-cannabinoids, which are very similar to the THC active in marijuana, are affiliated with the long-term downscaling of energy from synapses, though it's not clear what set of neurochemicals might be responsible for the opposite effect.

However useful for mapping the effect of pot on memory, the brain-chips are pure lab candy.

Trying to build an actual brain out of them, ignoring the difficulty of making the right number of connections among the chips, would take 133 million of the multi-synaptic chips, which are made from 16-core Opteron 6200s from AMD.

Just running the chips, according to TheReg, would take 15.3 megawatts of power.

Think how fast that would drain your iPhone.


Read more of Kevin Fogarty's CoreIT blog and follow the latest IT news at ITworld. Follow Kevin on Twitter at @KevinFogarty. For the latest IT news, analysis and how-tos, follow ITworld on Twitter and Facebook.

ITWorld DealPost: The best in tech deals and discounts.
Shop Tech Products at Amazon