Zenas C. Chao, 
"Toward the Neurocomputer: Goal-directed Learning in Embodied Cultured Networks" 
2007 (Georgia Tech PhD dissertation)

SUMMARY
Brains display very high-level parallel computation, fault-tolerance, and
adaptability, all of which are properties that we struggle to recreate in engineered systems.
The neurocomputer (an organic computer built from living neurons), where a
complicated language is naturally programmed and organized in the system, seems
possible and may lead to a new generation of computing device that can operate in a
brain-like manner. Cultured neuronal networks on multi-electrode arrays (MEAs) provide
a complex network connectivity pattern and greater freedom to manipulate and to access
the dynamics of groups of neurons, and become one of the best candidates for the next
neurocomputer.
I explored the possibility of the neurocomputer by studying whether we can show
goal-directed learning, one of the most fascinating behavior of brains, in cultured
networks. Inspired by the brain, which needs to be embodied in some way and interact
with its surroundings in order to give a purpose to its activities, we have developed tools
for closing the sensory-motor loop between a cultured network and a robot or an artificial
animal (an animat). This embodied hybrid neural-robotic system is termed a “hybrot”.
Unlike in the brain, sensory-motor mappings in a hybrot are defined by the experimenters.
In order to efficiently find an effective closed-loop design among infinite potential
mappings, I constructed a biologically-inspired simulated network, which exhibits similar
activity dynamics found in the cultured networks. By using this simulated network, I
designed a statistic that can effectively and efficiently decode network functional
plasticity better than some other existing statistics. Furthermore, in order to encode sensory 
information about the interactions between a hybrot and its environment, I
designed several stimulation protocols and an adaptive training algorithm that worked
cooperatively to direct network plasticity, and thus the hybrot’s behavior toward a userdefined
goal. By closing the sensory-motor loop with these decoding and encoding
designs, we successfully demonstrated a simple adaptive goal-directed behavior: learning
to move in a user-defined direction, and further showed that multiple tasks could be
learned simultaneously. These results suggest that even though a cultured network lacks
the 3-D structure of the brain, it still can be functionally shaped and show meaningful
behavior. Moreover, this demonstrates the possibility of utilizing living neurons for an
engineering purpose (e.g., control a robot to achieve a goal).
To our knowledge, this is the first demonstration of promising goal-directed
learning in a hybrot controlled by cultured neurons. Extending from these findings, I
further proposed a research plan to search for mappings that could help verify the
maximal learning capacity (or even true intelligence) of the cultured network, which can
help elucidate the possibility of the neurocomputer as the agent to future intelligent
machines. Knowledge gained from effective closed-loop designs also provides insights
about learning and memory in the nervous system, which could influence the design of
future artificial neural networks, more effective neuroprosthetics, and even the use of the
networks themselves as a biologically-based control system.