There is a new four-year, 6.25 million European project to
develop a cognitive
robot capable of learning, internal representations, planning,
understaning language meaning, and social interaction. According to a
new ElectronicsWeekly.com
article, researchers plan to put the robot together using the best
existing AI components that have been developed to date for things like
natural language, voice recognition, machine vision and other cognitive
and sensory processes. Researchers at the Univeristy of Birmingham
describe the process of combining all those bits and pieces into
something that works as ambitious and hard. For more details on the
project, see the EU Cognitive
Systems Integrated Project website or the 9 page project summary (PDF
format).
Sounds interesting and ambitious. I hope they make some progress in
four years. However, I think it would be a mistake only to rely upon
computer science experts. To make any real progress I think you would
need to get people from neuroscience and psychology involved, who have
at least a vague understanding of how real brains work.
The usual traps which such high profile projects fall into are:
- they become a gravy train for expensive consultants
- the temptation to invent yet another computer language, which never
really solves anything (it just shifts the problems around a bit)
- adopting a top down divide and conquer approach to the problem, where
each scientist narrowly concentrates upon solving some isolated problem
without considering the overall architecture and functioning of the system
- general lack of imagination