AI33 - Biological and Bioinspired Computing
The website will now be adapted to the 2008 winter term.
Course work will be announced soon.
The module is now taught by Marc de Kamps.
The only indisputably intelligent systems on this planet are
biological. Biological intelligences share several characteristics:
they were all designed by natural evolutionary processes, they are
typically controlled by nervous systems, and they often work together
in groups, swarms, or flocks. In contrast to human beings (and to
many of the artificial intelligences designed by human beings), the
vast majority of these biological intelligences are simple creatures:
they don't appear to use logic, mathematics, complex planning,
complicated modelling of their environment, or even memory in some
cases. Nevertheless, even the simplest of these natural intelligences
can achieve complex information processing and computational tasks that
current artificial intelligences find very challenging indeed.
Can these simple biological systems and the processes that created
them teach us lessons that will help us design better artificial
intelligences? Is it possible to abstract cartoon models of biological
systems that nonetheless capture some of the amazing feats that these
systems achieve? Where might such bio-inspired solutions fit in the
scheme of computer science today?
In this module we will explore the extent to which biological systems
such as neurons, brains, insects, and insect colonies have influenced
the design of artificial intelligences and our understanding of what
intelligence actually is. In addition we will learn how biological
processes such as evolution and coevolution can be harnessed to
automatically produce solutions to challenging problems in AI as well as
more generally in computer science, and how biological phenomena such
as flocking can inspire algorithms for robot control and the computer
generation of life-like animation.
A formal specification of the module, including objectives and syllabus, can be found
here. (The prerequisites listed here do not apply for 2008 students.)
Submission of Coursework:
Coursework 1: deadline to be announced
Detailed description 1
Coursework 2: deadline to be announced
Detailed description 2
For some course work you will need the BEAST tar-file. this contins the beast executable and
several project files. You need to compile the project files yourself, which is not as difficult as it may seem.
Detailed instructions are available to help you. The project files can then be modified.
Reading, suggested and otherwise.
The three demo on the Perceptron Theorem can be found here:
Website, demos, applets, etc.