Glossary of Terms
Action Potential
The response by a neuron when its postsynaptic membrane potential
reaches a threshold value, causing it to depolarize. In natural
neurons, depolarization causes the vessicles in the axon to release
chemicals (neurotransmitters) into their synaptic clefts, which
are detected by the efferent neurons' receptors. In synthetic
neurons, depolarization causes the efferent neurons to be evaluated
and if they depolarize, their action potentials are scheduled for
the following epoch.
Architecture
The set of
Salient Attributes of a system that consistently
supports its requirements, specification, design, implementation, test and maintenance (or
in nature, expression). Architecture includes not only quantitative attributes, but also
qualitative ones as well; i.e., the level of fidelity to which a standard is adhered to, etc.
By way of example, a suspension bridge might have the requirement that it allow passage of
boat traffic underneath. A floating bridge might not need to allow boat traffic to pass,
but would need to be unsinkable. A salient attribute of the suspension bridge is that it
can be made of suspendable materials. A salient attribute of the floating bridge is that
it can float. The design of the suspension bridge might include being porous, to save weight,
supporting its salient attribute, while the design of the floating bridge would not include
being porous, because that would cause it to sink.
Behavior
We use the term
behavior to describe organized movements, or sequences of organized
movements, associated with an organism controlled by a brain. In synthetic organisms,
such as virtual ones controlled by a synthetic brain, behavior may also include
virtual
movements, such as the navigation through the virtual world, or taking actions that
have meaning in the virtual world, such as the production or editing of data, or the
interaction with a computer system so as to control its operation or provide input.
The term behavior does not apply to movements that are not organized or in any way
coordinated; i.e., the random kicks and jerking of the limbs in a newborn human are
not coordinated so as to collectively create locomotion, although over time, the kicks
begin working together to ultimately achieve aggregate motion of the organism, which
becomes behavior, as it becomes organized.
Brain
A set of interconnected neurons and their connections, whether natural or synthetic.
Cerebellum
The cerebellum is a major feature of the brain of all vertebrates, playing an important
role in motor control and motor learning. In humans, the cerebellum does not actually
initiate movement, but provides fine-tuned coordination and timing, offloading these
more real-time aspects of movement from the cerebral cortex.
Connectome
An abstract representation of all of the components of a natural or synthetic brain,
including its neurons and interconnections. Commonly represented as a directed cyclic graph.
Design
The definition of a system which describes
how it will meet its specification.
Suppose an hypothetical system is required to interact with people in many languages
(a
Requirement). Its
Specification
might dictate that the system display its text in those languages in the native
character sets. The
Design would define a method for displaying
text in those character sets; i.e., with a font library that includes those fonts.
Deployment
In synthetic brain simulations, the provisioning of a simulation's server set with an
instantiation of a model, including classes, nodes and their links, and I/O channels.
The same model may be deployed to many server sets, creating multiple instances of the
same sentience, each with its own history of interaction with its environment, and behavior.
Epoch
A discrete period of time during which all of the neurons in a synthetic brain
simulation are evaluated and their state updated from activities since the previous
epoch. As natural brains have action potentials that occur no more frequently than
1ms, this is a common lower bound timebase for the duration of an epoch. Synthetic
brains interacting with slower environments may not need the same performance, and
might use epochs of 10ms, 100ms, or even 1sec, conserving energy or increasing capacity.
Homeostatic Regulation
A process by which a system's overall functioning is up-leveled (increased) or
down-leveled (decreased) through various mechanisms, each of which may be
modulated independently. In the case of brains, overall alertness is an effect
of a regulatory process, which may be emulated in a brain simulation.
Input Channel
A simulation's connection to its environment, physical or virtual, that allows
external agents, such as robotic sensory devices, to stimulate the running simulation
with data from the environment.
MISI
Model Input Signal Interface, the Sentience Engine's interface used to
process input channel input.
Model
An abstract representation, often expressed in a modeling language, used as input to a simulator.
Modeling Language
A formal programming language used to describe a simulation model. NeuroSynthetica's modeling
language SOMA™ includes class definitions for nodes, receptors, signals, as well as objects
which contain nodes, arrays of nodes, and arrays of objects. Programmatic statements provide for
using FOR, WHILE, and IF statements to describe algorithms by which the elements are wired
together with efferent links.
MOSI
Model Output Signal Interface, the Sentience Engine's interface used to
process output channel input.
Natural Brain
A brain built by nature; a multicellular organ consisting of a network of neurons and
other supporting tissue.
Node
The simulation's representation of a neuron in the simulation space. A node has a
fixed (x,y,z) coordinate in 3D node space, and is an instance of a node class, which
defines the simulation behavior of nodes of that class. Nodes may have
efferent
links, which are connections to other nodes in the simulation, which are stimulated
whenever their afferent node generates an action potential.
Output Channel
A simulation's connection to its environment, physical or virtual, that allows
external agents, such as robotic sensory devices, to receive output from the running
simulation, so that it may exhibit behavior in the environmental context.
Plasticity
One or more mechanisms through which a brain's functioning is adjusted in order
to achieve a change in future processing. One example in nature is Hebbian learning,
which causes the links of connected nodes that fire together, to increase their
strength over time.
In simulo, the algorithm which is used to implement
this is commonly
Spike Timing Dependent Plasticity, or STDP.
Requirement
The definition of a system which describes
what the system needs to accomplish.
Suppose an hypothetical system is required to interact with people in many languages
(a
Requirement). Its
Specification
might dictate that the system display its text in those languages in the native
character sets. The
Design would define a method for displaying
text in those character sets; i.e., with a font library that includes those fonts.
Salient Attribute
A quality of a system that supports the
Requirements,
Specification,
Design, and implementation (or in nature, expression) of the
design.
Scientific Method
A commonly-used emperical method for acquiring knowledge based on the
hypothesize and
test paradigm. An inquiry begins with a question, such as "How does sentience arise?"
An hypothesis or conjecture is made, such as "Perhaps sentience occurs when an animal
eats oranges." An experiment is designed to
test the hypothesis, where newborn
animals are not fed oranges, and their sentience evaluated in an orange-less environment.
The experiment is run and reveals a finding; namely, that some animals are sentient
without having eaten oranges. The conclusion from the study is that the hypothesis is
not true. Many similar experiments could rule out other foods and environments, gaining
knowledge about sentience.
Sentience
Also,
Sentience Effect. The effect that arises from a natural or synthetic brain
when it senses the environment and elicits meaningful and relevant behavior responsive to
the environment. For our purposes, sentience
arises from general purpose structures
not specifically designed to exhibit a certain behavior given a certain input stimulus.
Sentience Engine™ Simulator
NeuroSynthetica's real-time neural network simulator used to test, develop and deploy
sentience
in simulo. Sentience Engine™ is a trademark of NeuroSynthetica.
Server Set
A set of servers, each running a copy of the Sentience Engine™ simulator, that
forms a logical computing cluster for hosting a simulation. Each server may have
membership in more than one server set, and server sets may contain from one to 255
servers.
Simulation
A continuously-evaluated model of a brain running on a computer system or network
of computer systems, emulating the physical processes performed by natural systems
at a functional level.
SOMA™ Modeling Language
A formal programming language invented by NeuroSynthetica for use with its Workbench product.
SOMA™ includes class definitions for nodes, receptors, signals, as well as objects
which contain nodes, arrays of nodes, and arrays of objects. Programmatic statements provide for
using FOR, WHILE, and IF statements to describe algorithms by which the elements are wired
together with efferent links. SOMA™ is a trademark of NeuroSynthetica.
Specification
The definition of a system which describes
what will be done to meet the requirements.
Suppose an hypothetical system is required to interact with people in many languages
(a
Requirement). Its
Specification
might dictate that the system display its text in those languages in the native
character sets. The
Design would define a method for displaying
text in those character sets; i.e., with a font library that includes those fonts.
Spiking Neural Network (SNN)
A neural network model consisting of nodes together with their interconnections, where
the communication between nodes is not represented by static levels but by the temporal
pattern of pulses.
Synthetic Brain
A brain not built by nature, but by a designer using CAD tools to build a model
which is then run on a real-time simulator, normally connected to sensory apparatus
and output devices capable of exhibiting behavior within the environment.
Synthetic Intelligence (SI)
Artificial intelligence through sentience, arising from a simulated synthetic
model of the most minimal brain architecture necessary and sufficient for behavior
to emerge. Synthetic Intelligence is the next evolution of Artificial Intelligence (AI),
which includes a spectrum of classes of systems, including Expert Systems, Blackboard
Systems, and Machine Learning.
Synthetic Sentience (SS)
The effect arising from a real-time simulation of a network of simulated spiking neurons,
on a NeuroSynthetica Sentience Engine™ server. Like the sentience effect arising
from the interaction of biological network of neurons and their real-time environment,
Synthetic Sentience has no awareness that it is implemented by a continuous neural network
simulation; it merely continuously self-trains its neural network through live interaction
with its environment as the biological version does.