CADIA is the first artificial intelligence (A.I.)
laboratory in Iceland. We conduct research in various areas of intelligent
agents, with a strong emphasis on interaction and real-time performance.
Our past and present projects include topics such as planning, games,
large-scale A.I. systems, robots, humanoids and agent-based modeling.
Our research facilities give
students plenty of space, software and hardware to do advanced research
in the core CADIA areas, which we frequently demonstrate publicly (photos
in top row - picts 1-5: Vísindavakan,
pict 6: AI Festival).
- Agent Orientation. Our conceptualization of
intelligent agent follows closely the one proposed by
the first president of the American Association for Artificial
Intelligence Allen Newell who, (1980) in his initial presidential
address, described an intelligent agent as consisting of cognitive
faculties such as a perceptual system, memory system, processing
system, motor system, and so on. One way to understand the implications
of this definition is by contrasting it with an omnipresent
intelligence. The key difference
lies in the kind of information that the agent
has access to at any given point in time: Agents have a sensory
apparatus that limits what they can directly observe and a memory
with restrictions such as access time and capacity. Having an
agent-based focus means that mental models have limitations related
to embodiment. Another important focus under this theme is situatedness:
The nature of the world around the agent, how that world affects
what the agent can know, how the agent understands the relationship
between itself and the environment and how the agent changes with
each interaction with this environment.
We are interested in realtime so we basically want everything
to run faster, right? Alas, if things were that simple
we would probably not make this a special focus of our
work. Realtime is not about speed alone, it is about the
speed of one system relative to another system, in our
case the speed of a thinking mind embedded in the real
world. So for us realtime has to do with building systems
that deal intelligently with time, meets deadlines and
uses time wisely for planning, decision making and execution.
Being realtime means taking time into considerations when
making decisions and behaving in the world; it's a bout
understanding your own limitations for solving problems
and doing whatever it is that you do. At CADIA we look
at the various issues tying together knowledge, time and
situated behavior and how this dictates the mental architectures
that we must build.
- Virtual & Augmented Environments.
Virtual worlds can serve at least two very important roles
in A.I. One is that of a testbed: The real world is complicated;
a virtual world can be used as a simplified version of the
real world. Some virtual worlds are self-contained in themselves,
as for example many board games. Other virtual worlds are
fully-fledged social environments, such as massively multiplayer
on-line games. Another aspect of virtual worlds is the one
where all our information is presently moving: Cyberspace.
Virtual environments provide the A.I. researcher with a lot
of interesting possibilities for both innovative experiments
and innovative applications.
Newell, A. (1980). The Knowledge Level. AI
Magazine, Summer 1981.