Understanding the Brain

Artificial Intelligence System

From Cognopedia
Jump to: navigation, search

Artificial Intelligence System (AIS) was a distributed computing project undertaken by Intelligence Realm, Inc. with the long-term goal of simulating the human brain in real time, complete with artificial consciousness and artificial general intelligence. They claimed to have found, in research, the “mechanisms of knowledge representation in the brain which is equivalent to finding artificial intelligence”,[1] before moving into the developmental phase.

Science

The project's initial goal was recreating the largest brain simulation to date, performed by neuroscientist Eugene M. Izhikevich of The Neurosciences Institute in San Diego, California. Izhikevich simulated 1 second of activity of 100 billion neurons (the estimated number of neurons in the human brain) in 50 days using a cluster of 27 3-gigahertz processors.[2] He extrapolated that a real-time simulation of the brain could not be achieved before 2016.[3] The project aimed to disprove this prediction.

On July 12, 2008, AIS announced that the first phase of the project had been completed by reaching the 100 billion neuron mark. [4] The project then continued to simulate neurons while they completed the development of the other applications.

AIS simulated the brain via an artificial neural network, and used Hodgkin-Huxley models. The project utilized the BOINC distributed computing platform. In version 1.08 of the software each work unit received by a volunteer simulated 500,000 neurons for 100 milliseconds at 5 millisecond time steps (the estimated firing rate of a human neuron).

The application had four primary modules—for creating neurons, simulating neurons, visualizing neurons, and finally, knowledge acquisition. Intention was that the neuronal generator would eventually use genetic algorithms to generate neurons for simulation. The neuron simulator used mathematical models to simulate those neurons. Initially Hodgkin-Huxley models were used, but more models (perhaps hundreds) were intended to be utilized in the future. The visualization software was to allow the administrators to monitor and control the neuronal simulators. The knowledge acquisition module involved feeding information to the system and training it to build its knowledge base.

The AIS project had successfully simulated over 700 billion neurons by April 2009.[5]

The project was closed in November 2010 as the BOINC program of the project did not work.

See also

References