The CCortex Project

[bascule]

http://www.ad.com/ccortex.asp?id=1

 

Artificial Development is building CCortex, a massive spiking neural network simulation of the human cortex and peripheral systems. Upon completion, CCortex will represent up to 20 billion neurons and 20 trillion connections, achieving a level of complexity that rivals the mammalian brain, and making it the largest, most biologically realistic neural network ever built. The system is up to 10,000 times larger than any previous attempt to replicate primary characteristics of human intelligence.

 

CCortex employs a layered distribution of neural nets and detailed synaptic interconnections, closely emulating specialized regions of the human cortex, basal ganglia, thalamus and hippocampus. CCortex also extends the classical neural network paradigm of Hebbian connectivity to include temporal, analog representations of neuron spiking, enabling dynamic, real-time communication between neuron populations.

 

CCortex runs on a high-performance, parallel supercomputer cluster composed of 500 nodes and 1,000 processors, 1 terabyte of RAM, and 200 terabytes of storage. CCortex is expected to reach a theoretical peak performance of 4,800 Gflops, placing it in the top 20 fastest computers in the world. The neurological data that drives the CCortex simulation is a unique synthesis of multiple Artificial Development research projects, including the Cortical DB and the NanoAtlas.

 

CCortex applications

 

The development of CCortex will enable a wide range of commercial products that will transform and enhance global business relationships, with advanced capabilities in pattern recognition, verbal and visual communication, knowledge acquisition, and decision-making capabilities. These products will have widespread applications in the fields of artificial intelligence, communications, medical modeling, and database and search technologies. CCortex will perform many tasks previously unavailable to traditional computing by emulating the complexity and functionality of cortical regions involved in high-level attributes of human intelligence.

 

AD recently announced its first CCortex-based Autonomous Cognitive Model (ACM), codenamed 'Kjell', a realistic representation of the workflow of the human cortex. 'Kjell' is the first such image to run on a computer system. The ACM may have immediate applications for data mining, network security, search engine technologies and natural language processing. It is encouraged to respond to simple text commands by associating previous input with rewarded responses. In the simulation, these responses are based upon competing neuron populations in the associative cortex and other brain structures.