The Maelzel mission is to:
develop a technology (a machine) that demonstrates
apparent artificial intelligence. That mission will be
realized when an embedded and embodied hardware and
software technology, situated in a constrained
environment emerges demonstrating cybernetic, autonomous,
and adaptive agency.
Select a Machine Intelligence (MI).
Created by Alan Turing, this MI is a quasi
non-deterministic Finite State Machine. It is
unusual in that its transition table is undefined
until the machine starts to operate. Specifically,
the input, output, and potentially next-state
characters associated with the transition table
entries can be left semi-undefined until run-time
experiences shape them according to the Turing
P-Type algorithm. In this experiment, the MI was
instantiated as the control-agency of an
artificial rodent host. The rodent was tasked
with either finding food and remembering where it
found it or "die." The Turing algorithm worked so
well occasionally the artificial rodents had to
be sacrificed in order to have their (acquired)
memory traces examined. An essay describing this
work is on our
Demonstrate an Extensible Machine Intelligence (MI).
BASIC [Basic All-purpose Symbolic Instruction Code]
Unorganized Group Simulation (BUGS) demonstrated
an instantiation of a Learning Automata (LA) derived
from the later theories of Alan Turing. The experiment
featured predator and prey agents situated in a 2D
maze. The LA is a dynamic, artifactual model of
natural intelligence that self-organizes using
cybernetic theory. Ganged as a set of four parallel
LA, each machine fired across, and was prioritized
by, needs drawn from a layer of the Maslow Hierachy
of Needs pyramid. This work explored the extensibility
of the MI for use in social science modeling applications.
There is a short conference proceeding describing this
experiment and its results on our
Select a Virtual World.
This simulation was created to characterize the
native application program interface (API) between
The question was: Could the LibOpenMetaverse
API be used to provide embodied agents with
sufficient sensorimotor data for social simulation
purposes? Here, the avatars execute the same highly
choreographed fighting movements. However, because
Opensimulator uses an authoritative server model,
all avatars share the same (virtual) physical space
and time even though each was free to execute its own
asychronous behavioral thread. This was vital.
Also, social simulation often requires special
environments. Here a highly-detailed building was
constructed that architecturally resembled the
D-Street Port of Entry in Naco, Arizona. During
some of the testing the avatar physics capsules
Emulate Hominid Biomimesis.
"Sexually differentiated philopatry and dispersal: A
demonstration of the Baldwin effect and genetic drift".
This research used an agent-based model. The agents simulated
sexually dimorphic hominids on a simulated terrain. Agent
offspring inherited their parents artificial genetics and
demonstrated emergent community fission-fusion, genetic
drift (Sewall Wright Effect), and ontogenetic evolution
(Baldwin Effect). Additionally, the work considered the
cross-relationships of sexually differentiated philopatry
and dispersal, autonomous decision making, attitude-biased
foraging, and resultant cognitive dissonance.
A conference proceeding describing the experiment and
results are on our
Enable Cluster-based Machine Society.
"Midwife: CPU cluster load distribution of Virtual
Agent AIs" is work that developed an algorithm
to enable the instantiation and hosting of large
numbers of virtual agents within a CPU cluster. The
Java load distribution algorithm balanced avatar objects
as client-side agent controllers across networked CPUs.
In turn, the Java component can be integrated with a C#
driver that instantiates new avatar clients in a
Virtual World. The two-piece code set operates on
Windows-based CPU clusters and is designed for the
Multi-User Virtual Environment (MUVE) of OpenSimulator
and LibOpenMetaverse operating together. The graphic
above can be enlarged and viewed by
Extending Emergent Biomimetic Sociality.
This agent-based model extended the
Integrate biomimetics with Machine Intelligence (MI).
This research is being developed in parallel with our
Early North American Migrant Camp.
"Late Pleistocene Human Migrations: An Agent-based
Modeling Approach" was a simulation of a hypothetical
human migration some experts suggested may have
occured during the Last Glacial Maximum. Extending such
work to other migrations or similar refugee experiences
but within a Virtual World might allow us to examine
subtle social and environmental interactions in
detail and realtime visual accuracy. Although the
original 2D ABM was biologically detailed, we are now
capable of adding biomimesis and adaptive Machine
Intelligence to better demonstrate social and migratory
theories in relationship to environmental and possibly
social circumscription. There is a conference proceeding
available on the
Pedestrian and small-group social behavior
at a large venue. This was research using an
agent-based model built on a spatially accurate 2D grid
of a public venue and parking lot. On the grid were
placed gradient "heatmaps" acting as heuristic guides
for simulated pedestrian traffic. The "heatmaps" came
from empirical observation of customary pedestrian
behavior in the United States. No path-directives were
given the agents. Rather, agent behavior emerged from
"visually" detecting landmark "goals," moving towards
them, and by emulating observation-inspired small-group
locomotion behaviors taken by individuals and families
as they moved from car (parking lot) to building (doors).
Simulation identified anomolous crowd behavior of
interest to our sponsor. Hendrey, M. Crooks, A. and
Rouly, O. C. collaborators.