HUMANOBS is a CADIA-led €2M, 3-year research project to develop the next-generation AI architectures that grow by self-programming. We will develop an artificial television-show host that can learn how to conduct interviews by watching people. The AI will observe what people say and do and then imitate not just their behavior but also their intentions. Based on the concept of model-driven cognition, we aim to lay a foundation for machines with a more general-purpose intelligence.
CADIAPlayer
has been CADIA's entry into the 2007 and 2008 International Game Playing Competition
held at the AAAI conference. The competition draws participants
from a large number of AI labs around the world. CADIA
has walked away twice with the highest score, a considerable achievement as
these have been CADIAPlayer's first two years participating in such a competition.
We are
inventing new methods to create believable human behavior in animated
characters for massively multiplayer games. We focus on communicative
behavior between characters in the game world, emphasizing natural
motion and social interaction. Characters are either fully controlled
by the game AI or under the direction of human players. For the
latter, we are also exploring new high-level user interfaces. This
research is done in collaboration with CCP
Games. Related projects: BML Realizer.
Vélaldin
is a new development platform for exploring emergent phenomena and
complex systems. Based on cellular automata, Vélaldin has
been designed with an emphasis on flexibility of the cells, enabling
the user to define and implement a number of their features. Vélaldin
provides a complete environment in which emergent, complex systems
can be implemented and experimented with in a diverse manner.
Superhumanoid
One (S1) is the first phase of many in the Superhumanoids project,
a multi-year research project which aims to build the next generation
of social robots. S1 is currently being brought to life as a fully
autonomous radio show host, SuperRadioHost, fully equipped to create
a radio program and interview people in a live broadcast. The next
Superhumanoid, S2, due to be started in 2007, will will marry the
brain of S1 with a graphical SuperBody, as well as animation and
multimodal planning and sensing technologies.
We are developing
new domain-independent adaptive search techniques with a primary
focus on: enhancing ongoing research into adversary search techniques
for learning search control and generalizing these methods such
that they can also be applied to non-adversary domains such as single-agent
search and automated planning. The new techniques will be integrated
into existing heuristic-search solvers and planners for measuring
their effectiveness.