Integrating Reasoning Systems for Trustworthy AI
October 13, 2024
Hybrid event
Held in conjunction with 40th International Conference on Logic Programming (ICLP)
Dallas, Texas, United States
The focus of the 2024 Logic and Practice of Programming workshop is integrating reasoning systems for trustworthy AI, especially including integrating diverse models of programming with rules and constraints.
Trustworthy AI requires programming with rules and constraints for expressing and solving knowledge-intensive inference and combinatorial problems. A wide range of programming models have been proposed, including but not limited to the following, and essentially all of them require or support imperative programming for use in practical applications.
Classical first-order logic (FOL), not supporting transitive relations,
with satisfiability (SAT) and satisfiability modulo theory (SMT) solvers
Deductive database (primarily Datalog) systems with fact-driven inference
Logic programming (dominantly Prolog) systems with goal-directed search,
extended with sophisticated tabling and well-founded semantics
Answer set programming (ASP) systems,
with sophisticated grounding and solving and stable model semantics
First-order logic (FOL) extended with inductive definitions (ID)
Constraint logic programming(CLP) extending Prolog systems with constraints
Constraint programming (CP), not supporting transitive relations,
with backtracking, constraint propagation, local search, and more for solving
Mathematical programming (MP), not supporting transitive relations,
with linear programming, nonlinear programming, and much more for solving
Co-inductive logic programming (s(ASP), s(CASP)) extending Prolog systems
with goal-directed search for ASP and co-ASP solutions
Advanced knowledge representation (KR) with higher-order, objects, updates,
defeasible reasoning, paraconsistency, uncertainty, and probability
Given any application problem---whether for planning or scheduling or regulatory compliance, requiring logical or probabilistic reasoning, or constraint satisfaction or optimization---how to best express and solve it using one or more of the models?
In recent years, AI systems built with large neural networks trained on massive data sets (such as GPT3 with 96 layers and 175-billion parameters on 570 GB of filtered data (https://arxiv.org/pdf/2005.14165)) have become increasingly capable in producing impressive outputs and beating humans in many applications. However, these systems may produce outputs that are not reliable, explainable, or aligned with intended uses.
The goal of the workshop is to bring together best people and best languages, tools, and ideas to discuss how to address these challenges with rigorous knowledge representation and reasoning, powerful and easy-to-use rule and constraint languages, and robust justifications and alignment checks. A wide variety of application problems will be used in the discussions. See here for some example application problems.
Potential participants are invited to submit a position paper (1 or 2 pages), and also to state whether they wish to present a talk at the workshop. Because we intend to bring together people from a diverse range of language and programming communities, it is essential that all talks be accessible to non-specialists.
The program committee will invite attendees based on their position paper submissions and will attempt to accommodate presentation requests, but in ways that fit with the broader organizational goals outlined above.
Please submit your position paper through this EasyChair submission URL
https://easychair.org/conferences/?conf=lpop2024
Position paper due Aug 16 18, 2024 (extended)
Attendee invitation Sep 6, 2024
Camera-ready Sep 20, 2024
Workshop Oct 13, 2024
Luc De Raedt KU Leuven, Belgium
Georg Gottlob Oxford University, UK
Henry Kautz University of Virginia, US
Anil Nerode Cornell University, US
Annie Liu Stony Brook University, US
Martin Gebser, University of Klagenfurtm, Austria
Michael Gelfond, Texas Tech University, US
Benjamin Grosof, DARPA, US
Gopal Gupta, UT Dallas, US
Michael Kifer, Stony Brook University, US
Marta Kwiatkowska, University of Oxford, UK
Fabrizio Riguzzi, University of Ferrara, Italy
Joost Vennekens, KU Leuven, Belgium
Toby Walsh, University of New South Wales, Australia
Jan Wielemaker, CWI, The Netherlands
Roland Yap, National University of Singapore, Singapore