Autonomy and Artificiality Margaret A. Boden Hojin Youn.
-
Upload
clinton-potter -
Category
Documents
-
view
222 -
download
7
Transcript of Autonomy and Artificiality Margaret A. Boden Hojin Youn.
Autonomy and Artificiality
Margaret A. Boden
Hojin Youn
1. The Problem - and Why It Matters
H. Simon : “The Science of the Artificial”– AI, Cybernetics
– A-Life : • uses informational concepts and computer-modelling to study
the functional principles of life in general (C. Langton, 1989)
A-Life vs. AI– abstract study of life : abstract study of mind
– the concept of autonomy, applies to A-Life.
1. The Problem - and Why It Matters
Human autonomy / freedom– Rollo May(1961)
• dehumanizing dangers in modern science
• refer to the mechanic implications of natural sciences(behaviourists psychology)
– Skinner(1971)• Freedom, is an illusion.
• Environmental pressures determine our behavior.
– What of artificial sciences?
1. The Problem - and Why It Matters
Contents– How A-Life addresses the phenomenon of
autonomy– The concept of autonomy– Artificial sciences doesn’t deny, downgrade,
our freedom
2. AI, A-Life, and Ants
Simon(1969)– much the same view with Skinner
– rational thought and skilled behaviour are triggered by specific environmental cues
– but allows internal, mentalistic cue GPS
– paid no attention to environmental factors
– human thought purely in terms of internal mental/computational process
2. AI, A-Life, and Ants
Robotics driven by internalist view– guided top-down by internal planning and
representation
– not real-world, real-time creatures : their env. were simple, highly predictable ‘toy-worlds’
noubelle AI– behaviour controlled by an interaction between
• low-level mechanisms in the system
• constansly changing details of the environment
2. AI, A-Life, and Ants Robotics, situated
– no need for the symbolic representations / detailed anticipatory planning
– Traditional robotics:• brittleness caused by unexpected input
• no way the problem environment can help
– “Best source of information about the real world is the real world itself”
– usually in hardware, but• Behavior apparently guided by goals and hierachical planning
can occur(Maes 1991)
2. AI, A-Life, and Ants Studying ‘emergent’ behaviors - GA and A-Life GA
– self-modifying programs, continually come up with new rules(structures)
– use rule-changing algorithms modelled on genetic processes
• Mutation : makes an change in a single rule
• Crossover
• Algorithms for identifying & selecting the successful rules
– e.g.) Karl Sims(1991)• use GA to generate new images
2. AI, A-Life, and Ants A-Life
– use computer modelling to study processes that start with relatively simple, locally interacting units, and generate complex individual/group behaviors
• Self-organization / Reproduction / Adaptation / Purposiveness / Evolution
– Self-organization• flocking : Boids(a collection of very simple units) modelling
• Possible for group-behavior to depend on very simple, local rules
Situated robotics / GA / A-Life share:– bottom-up, self-adaptive, parallel processing
2. AI, A-Life, and Ants Evolutionary Robotics
– simulation of insect-like robots
– adapts to its specific task-environment
Links with biology
noubell AI : autonomous agents A-life : autonomous systems
3. Autonomous Agency Artificial insects:
– specifically constructed to adapt to the particular environment
Autonomy1. The extent to which response to the environment is
direct or indirect– involves behavior mediated by inner mechanisms shaped by
experience
3. Autonomous Agency
2. The extent to which the controlling mechanisms were self-generated rather than externally imposed
– behavior which ‘emerges’ as a result of self-organizing process, not prefigured in the design of the creature
– emergence-hierachies, evolve as a result of new forms of perception
– intelligible vs. unintelligible emergence • (flocking : Sims’s program)
– e.g.) different thoughts in consciousness
3. Autonomous Agency3. The extent to which inner directing mechanisms can be reflected
upon, and/or selectively modified in the light of general interests or the particularities of the current problem in its environmental contexts
– conscious deliberation : the crux of human autonomy
– conscious thought requires a sequential ‘machine’ more like a von Neumann computer
– Creativity: an aspect of human autonomy
Autonomy and Unpredictablity– AI systems, not necessarily deterministic
– Determinism Predictability
4. Conclusion The science of artificial can model autonomy of various
kinds– highlights autonomy - as a characteristic of living things
– A-Life can teach us how increasing complexity arises from self-organization on successive levels, and how a creature can negotiate its environment by constant interaction with it.
– But, the kind of autonomy, free choice, is better illuminated by the classical AI.
AI does not reduce our respect for human minds.– Helps us to understand how it is possible