A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation

"A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation" is the title of a paper from MIT and Harvard researcher. The paper can be downloaded from The PLoS Computational Biology. It is about brain modelling. They want to model the how our brain process visual information. The hardware used is GPU (graphical processing unit).

They publish the video which can be seen here. In the video you can see their comments about IBM cat brain. They say implicitly that IBM one do have the power of cat brain but it is not successful (yet) to model how the cat brain works. The same news from Smart Planet can be read here.

***

Finding a better way for computers to "see" from Cox Lab @ Rowland Institute on Vimeo.

***

Other interesting news recently is about Intel processor with 48 cores. Actually it is 24 dual core connected in mesh network. Beside GPU, this 48-core processor can be used also for brain modeling.

Are we closed to Singularity?

About IBM Cat Brain

My previous post is about IBM Brain Simulator. It turns out that there is still hot discussion about whether it can simulate brain activity or not, as written on an article in Spectrum IEEE.

The IBM Brain Simulator can model the neurons, the synapses and the connectivity of them. It is true that the numbers of neurons and so on are equal to cat's brain or human visual cortex. But one researcher of EPFL Blue Brain, Henry Makram disagreed with the method that IBM Brain Simulator from Almaden's lab used to model the synapses. Henry Makram questioned why ion channeling in the synapses was not modeled in the Almaden's cat brain simulator.

In my opinion, It is already a break-through to model neuronal connectivity. To model the synapses functionality for each connectivity is really hard job. It will take more processors and consume more energy than just 1.4 MW.

My interess is to combine brain simulator and brain-computer interface to create a quasi-telepathic between human and computer. I don't really care which brain model they use: the IBM Almaden's or the EPFL Blue Brain's.

IBM Biggest Brain Simulator in 2009

Three days ago, I read an article from IEEE Spectrum about IBM Brain Simulator. The following day, I read two articles about the brain simulator from the Smart Planet: the first focuses on the technology and the other concerns about what the future will be. Popular Mechanic also has an article about the brain modeling from IBM's Almaden research center compared to others from Stanford University and Neuroscience Institute in San Diego. IEEE Spectrum says that Europe also have similar project called Blue Brain, at EPFL in Lausanne, Switzerland.

The IBM Brain Simulator is featured in Supercomputing 2009 event in Portland, Oregon. This event is about high performance computing. So it is not about brain-computer interface (BCI) or brain robot. There is a possibility to combine BCI and this kind of brain simulator. Maybe I can participate in that kind of research.

There is a concern about the future about the relation between computers and humans. The second article from Smart Planet discussed this. A friend of mine, Mova Al'Afghani, put slides from Karl Fisch in his blog about future prediction, which says that in 2013, super computer will exceed human brain capability and in 2049, a $1000 computer will exceed the capability of entire human species. In 1999, Ray Kurzweil wrote a book: The Age of Spiritual Machines: When Computers Exceed Human Intelligence. In the book, it is predicted that in 2020, a $1000 computer will exceed human intelligence. So you know who makes the prediction. Ray Kurzweil also wrote another book: The Singularity is Near: When Human Trancends Biology. IBM said that in 2019, they can mimic human brain which has 20 billions neurons and 200 trillions synapses.

Singularity (in this context) means that the computer has reached human intelligence and capabilities. TV series Terminator SCC mentioned this singularities. Some people are afraid of this singularity and create Anti Skynet group. (Skynet is fictional "machine" in Terminator). The optimist people build Singularity University to prepare humanity for accelerating technological change.

This is the feature of IBM Biggest Brain Simulator:

• uses Dawn, BlueGene/P supercomputer
• uses C2 cortical simulator
• funded by DARPA (U.S. Defense Advanced Research Projects Agency)
  
• spended 40 million US dollars
  
• contains 147,456 processors
  
• uses 147 TB of RAM
  
• consumes 1.4 MW
• uses 10 rows of racks (and miles of cables)
  
• uses 6,675 tons of air-conditioning equipment spouting 2.7 million cubic feet of chilled air
• uses a universal neural circuit called a microcolumn to mimic a single neuron.
  
• exceed cat's brain capability
• can simulate only human visual cortex capability
  
• takes 500 seconds to simulate 5 seconds of real mammal's brain activity (in average).
  

One more interesting thing. With the same technology, to exceed human brain, the supercomputer will need between 100 MW and 1 GW. It takes a nuclear power plant for simulating only a single human brain. The real human brain takes only 20 watts.