Design of a brain computer using the novel principles of output-driven operation and memory-based architecture

Abstract The brain is an organ especially differentiated to acquire algorithms in a self-organized fashion according to the genetic algorithm. These acquired algorithms allow the brain to respond to the ever-changing environment that surrounds it. We hypothesize that two general principles give the brain its auto-designing capacity and, consequently, the potential to acquire algorithms: (a) Output-driven operation; the brain receives numerous types of input signals to which it makes a single output at an instant in time. The output-driven operation principle implies that input signals change information within the system only when making an output. According to this principle, learning should be output-driven. Through this process, input signals are utilized by the system to select an output most suitable to its surrounding, resulting in an enhanced learning efficacy. (b) Memory-based architecture; the brain algorithm, once acquired according to the output-driven operation principle, is memorized in a non-erasable fashion for extended periods of time. Any newly acquired algorithms are added to those already acquired. Input signals leave their traces in the system according to the memory-based principle, even if they are not a strong enough stimulation to drive an output. The combination of output-driven operation and memory-based architecture endows the brain's learning algorithm with the property of predictability. Here we confirm these two principles of brain design by successfully implementing them in an autonomous mechanical device, namely a helicopter, that has a will to fly and ultimately acquires the motor control required to do so in a manner similar to a fledgling bird. We anticipate that these principles will eventually lead to the development of a brain computer as an engineering counterpart of a real brain.