When does an AI system understand? What representations must it have? What computations must it perform? How must those computations fit together? What relationship must it have to its environment? This light seminar bridges readings on intentionality in the philosophy of mind with readings on agent architecture in AI to explore both the richness of our intuitions about intelligence on the one hand, and the subtle relationships between cognitive capacities and end-to-end intelligent behavior on the other. We'll work together to create some compelling demonstrations of the depth of these questions and an agenda for future research.
Rutgers sakai worksite: 16:198:500:06 F08
Coordinator: Matthew Stone
Class meeting: W7 (6:40-8:00), Hill 482 conference room
Office hours: Tues 3-5, Core 328
Introduction: history of AI architectures and the challenge of representation
Introduction continued:
Skim these papers to get a sense of how system design has changed over time, and a sense of how designers have taken different approaches to the problem of flexibility, robustness and effectiveness of behavior.
Robot videos:
The philosophy of mental representation. Read:
This is a short, dense reading; it's already a summary of some complex technical material so it makes sense to try to work through it carefully with as much patience as you can, rather than skimming through to get the gist.
Inside and outside. Read:
Useful background, if this is too fast:
Of historical interest:
Approximation and the problem of error.
Useful background:
Autonomy, creativity and critical engagement. Read:
Models and arguments for constructing meaningful robots, 1.
Perception and representation in cognitive architecture.
Read:
Models and arguments for constructing meaningful robots, 2.
Perception and representation in cognitive architecture.
Read:
No class: go to Oliver Sacks instead.
Models and arguments for constructing meaningful robots, 3.
Meaning in language.
Read:
Recap discussion: key imagery.
To prepare for this meeting, pick a key image or idea from the semester's reading or your own research. Think about what facets of representation that idea puts into relief, and why the idea is effective at conveying it. We'll assemble the results into a smorgasbord that we can continue to draw on in thinking about how to present our views about representation.
Recap discussion: vignettes, visualization and narrative.
To prepare for this meeting, draw on the key imagery that we discussed last time to prepare a sketch that might contribute to a presentation about representation. You might map out a fragment of narrated animation, a snippet of dialogue, or a story framework which might tie together different ideas. The point is to make the imagery more precise, more consistent, and more explicitly tied to our other ideas and the overall theme of representation.
Nearby talks on autonomous vehicles:
Changes since Feb 28, 2024 10:59 AM