Invited Speakers
Knowledge Graphs and the Evolving AI Landscape
Monday, May 15, 2023, 9:00 AM- 10:00 AM
Room: Grand Ballroom
Tetherless World Senior Constellation ChairProfessor of Computer and Cognitive ScienceDirector Rensselaer Web Science Research CenterRensselaer Polytechnic Institute
Abstract: The Artificial Intelligence landscape is changing at an unprecedented pace. Powerful AI tools and services have amazed both the general public as well as many seasoned AI researchers. Like all technologies, however, challenges remain. Many remaining challenges for large language models and generative AI align with strengths of knowledge graphs and semantic AI. In this talk, we will discuss topics including context, abstraction, and provenance. We hope to provide some useful directions for knowledge graph research and applications in today’s evolving landscape.
Bio:
Deborah McGuinness is the Tetherless World Senior Constellation Chair and Professor of Computer, Cognitive, and Web Sciences as well as Professor of Industrial and Systems Engineering at RPI. She is also the founding director of the RPI Web Science Research Center. Deborah has been recognized for her seminal work in explanation, knowledge engineering, ontologies, provenance, and methodologies for creating deployed applications. Her applications have touched a wide range of areas including precision health, exposure science, wireless spectrum, material science, as well as wine and food pairings, just to name a few. She is a fellow of the Association for the Advancement of Artificial Intelligence for , the American Association for the Advancement of Science, as well as the recipient of the Robert Engelmore Award from AAAI for leadership in Semantic Web research and in bridging Artificial Intelligence (AI) and eScience, significant contributions to deployed AI applications, and extensive service to the AI community. She is also the recipient of the lifetime achievement award from the Knowledge Graph conference for outstanding contributions to the field of Knowledge Engineering.
Deborah currently leads a number of large diverse data intensive resource efforts and her team is creating next generation ontology-enabled research infrastructure for work in large interdisciplinary settings. Prior to joining RPI, Deborah was the acting director of the Knowledge Systems, Artificial Intelligence Laboratory and Senior Research Scientist in the Computer Science Department of Stanford University, and previous to that she was at AT&T Bell Laboratories in Artificial Intelligence Research. Deborah also has consulted with numerous large corporations as well as emerging startup companies wishing to plan, develop, deploy, and maintain semantic web and/or AI applications. Deborah has also worked as an expert witness in a number of cases, and has deposition and trial experience. Some areas of recent work include: data science, next generation health advisors, ontology design and evolution environments, semantically-enabled virtual observatories, semantic integration of scientific data, context-aware mobile applications, search, eCommerce, configuration, and supply chain management. Deborah holds a Bachelor of Math and Computer Science from Duke University, her Master of Computer Science from University of California at Berkeley, and her Ph.D. in Computer Science from Rutgers University.
Bio:
Deborah McGuinness is the Tetherless World Senior Constellation Chair and Professor of Computer, Cognitive, and Web Sciences as well as Professor of Industrial and Systems Engineering at RPI. She is also the founding director of the RPI Web Science Research Center. Deborah has been recognized for her seminal work in explanation, knowledge engineering, ontologies, provenance, and methodologies for creating deployed applications. Her applications have touched a wide range of areas including precision health, exposure science, wireless spectrum, material science, as well as wine and food pairings, just to name a few. She is a fellow of the Association for the Advancement of Artificial Intelligence for , the American Association for the Advancement of Science, as well as the recipient of the Robert Engelmore Award from AAAI for leadership in Semantic Web research and in bridging Artificial Intelligence (AI) and eScience, significant contributions to deployed AI applications, and extensive service to the AI community. She is also the recipient of the lifetime achievement award from the Knowledge Graph conference for outstanding contributions to the field of Knowledge Engineering.
Deborah currently leads a number of large diverse data intensive resource efforts and her team is creating next generation ontology-enabled research infrastructure for work in large interdisciplinary settings. Prior to joining RPI, Deborah was the acting director of the Knowledge Systems, Artificial Intelligence Laboratory and Senior Research Scientist in the Computer Science Department of Stanford University, and previous to that she was at AT&T Bell Laboratories in Artificial Intelligence Research. Deborah also has consulted with numerous large corporations as well as emerging startup companies wishing to plan, develop, deploy, and maintain semantic web and/or AI applications. Deborah has also worked as an expert witness in a number of cases, and has deposition and trial experience. Some areas of recent work include: data science, next generation health advisors, ontology design and evolution environments, semantically-enabled virtual observatories, semantic integration of scientific data, context-aware mobile applications, search, eCommerce, configuration, and supply chain management. Deborah holds a Bachelor of Math and Computer Science from Duke University, her Master of Computer Science from University of California at Berkeley, and her Ph.D. in Computer Science from Rutgers University.
Reasonable, Trusted AI through Symbolic Ethico-legal Control and Reflection?
Tue, May 16, 2023, 9:10 AM - 10:00 AM
Room: Grand Ballroom
Professor, Chair for AI Systems Engineering, Otto-Friedrich-Universität BambergProfessor, Math and Computer Science, Frei University, Berlin, Germany
Abstract
In formal methods, symbolic specification and verification are prominent and successful means of achieving trust and security in software and hardware development.
Due to their opaque, statistical nature, combined with their dependence on typically imperfect data, modern subsymbolic AI models can be seen as antipodes to systems developed according to this formal methods paradigm. In subsymbolic AI, the sharp concept of 'trust' has thus been replaced by the fuzzy and inconclusive concept of 'trustworthiness', which seems to be insufficient for most critical applications. I will argue that appropriate hybrid (or neuro-symbolic) AI architectures are a promising option for overcoming this dichotomy. Not only could they reintroduce a sharper notion of trust, but they could enable run-time alignment between socially legitimated, regulative ethico-legal theories and trained models. The key idea is to focus less on explaining the imperfection of trained models, but rather on the independent evaluation and justification (with symbolic means) of their proposed actions in a given application context.
Adopting this perspective, I will outline recent collaborative research on (i) ethico-legal control and reflection architectures, (ii) logico-pluralistic normative reasoning using the meta-logical knowledge representation and reasoning methodology LogiKEy, and (iii) recent progress in the use of higher-order interactive and automated theorem provers to support the automation of (not only) normative reasoning in the LogiKEy framework.
Suggested Readings
Bio Prof. Christoph Benzmüller holds the Chair of Artificial Intelligence Systems Engineering at the University of Bamberg, is an Adjunct Professor at the Freie Universität Berlin, and collaborates closely with the University of Luxembourg, among others. He is a member of the Federation of German Scientists (VDW) and a German national contact for the Confederation of Laboratories for Artificial Intelligence Research in Europe (CLAIRE).
For Benzmüller, AI is less a technology than a scientific discipline that should conceptually focus more on the exploration and experimentation with the abstract representation of objects. For Benzmüller, the exploration and flexible processing of abstract representational objects within hybrid (or neuro-symbolic) AI architectures that merge symbolic and sub-symbolic techniques is a key challenge and opportunity for modelling (strongly) intelligent AI systems. Explicit, declarative representations are also particularly relevant for the realisation of trusted, controllable AI systems, since they not only make (normative and other) knowledge transparent and justifiable, but also enable efficient and precise communication between humans and machines.
Benzmüller's research therefore addresses topics such as the automation of rational and normative reasoning in computers, universal knowledge representation, computational metaphysics, and the mechanisation of mathematical reasoning. A particular research focus is on higher-order interactive and automated theorem proving as a backbone for the above activities.
In formal methods, symbolic specification and verification are prominent and successful means of achieving trust and security in software and hardware development.
Due to their opaque, statistical nature, combined with their dependence on typically imperfect data, modern subsymbolic AI models can be seen as antipodes to systems developed according to this formal methods paradigm. In subsymbolic AI, the sharp concept of 'trust' has thus been replaced by the fuzzy and inconclusive concept of 'trustworthiness', which seems to be insufficient for most critical applications. I will argue that appropriate hybrid (or neuro-symbolic) AI architectures are a promising option for overcoming this dichotomy. Not only could they reintroduce a sharper notion of trust, but they could enable run-time alignment between socially legitimated, regulative ethico-legal theories and trained models. The key idea is to focus less on explaining the imperfection of trained models, but rather on the independent evaluation and justification (with symbolic means) of their proposed actions in a given application context.
Adopting this perspective, I will outline recent collaborative research on (i) ethico-legal control and reflection architectures, (ii) logico-pluralistic normative reasoning using the meta-logical knowledge representation and reasoning methodology LogiKEy, and (iii) recent progress in the use of higher-order interactive and automated theorem provers to support the automation of (not only) normative reasoning in the LogiKEy framework.
Suggested Readings
- Designing Normative Theories for Ethical and Legal Reasoning: LogiKEy Framework, Methodology, and Tool Support. Artificial Intelligence, 2020. http://doi.org/10.1016/j.artint.2020.103348
- Reasonable Machines: A Research Manifesto. KI 2020, Springer, 2020. http://doi.org/10.1007/978-3-030-58285-2_20
- Universal (Meta-)Logical Reasoning: Recent Successes. Science of Computer Programming, 2019. https://doi.org/10.1016/j.scico.2018.10.008
- Who Finds the Short Proof? Logic Journal of the IGPL, 2023. http://doi.org/10.1093/jigpal/jzac082
Bio Prof. Christoph Benzmüller holds the Chair of Artificial Intelligence Systems Engineering at the University of Bamberg, is an Adjunct Professor at the Freie Universität Berlin, and collaborates closely with the University of Luxembourg, among others. He is a member of the Federation of German Scientists (VDW) and a German national contact for the Confederation of Laboratories for Artificial Intelligence Research in Europe (CLAIRE).
For Benzmüller, AI is less a technology than a scientific discipline that should conceptually focus more on the exploration and experimentation with the abstract representation of objects. For Benzmüller, the exploration and flexible processing of abstract representational objects within hybrid (or neuro-symbolic) AI architectures that merge symbolic and sub-symbolic techniques is a key challenge and opportunity for modelling (strongly) intelligent AI systems. Explicit, declarative representations are also particularly relevant for the realisation of trusted, controllable AI systems, since they not only make (normative and other) knowledge transparent and justifiable, but also enable efficient and precise communication between humans and machines.
Benzmüller's research therefore addresses topics such as the automation of rational and normative reasoning in computers, universal knowledge representation, computational metaphysics, and the mechanisation of mathematical reasoning. A particular research focus is on higher-order interactive and automated theorem proving as a backbone for the above activities.
AI and NLP Challenges, Solutions, and Gaps in the age of Chat GPT
Wednesday, May 17, 2023, 9:00 AM - 10:00 AM
Room: Grand Ballroom
Bonnie J. Dorr
Professor of Computer Science at the University of FloridaAbstract:
This talk presents challenges, solutions, and gaps in AI and Natural Language Processing (NLP), with an emphasis on the need for explainability in the era of ChatGPT. Examples include: machine translation of human languages, ask detection for defending against social engineering attacks, and stance detection for extracting attitudes from social media. Past, current, and future projects face several challenges: (a) brittleness of rule-based linguistic principles for large-scale processing; (b) shallowness of statistical methods and neural language models for understanding implicit information; and (c) lack of “explainability” amidst ever-increasing numbers of black-box models. A case is made for hybrid approaches that combine linguistic generalizations with statistical and neural models to handle implicitly conveyed information (e.g., beliefs and intentions), and also for the implementation of an “explainable” propositional representation that supports the ability of developers and end users to understand what is going on inside the AI system. Questions of interest range from “What is the social engineer’s underlying goal in a two-way interaction?” to “What beliefs support individuals’ attitudes regarding pandemic interventions?” to “How does targeted influence impact attitudes online?”. Such information is generally not extractable from large language models alone and, moreover, such models are hampered in that they are too large to retrain on a regular basis by the average researcher, developer, or customer. Representative examples of ChatGPT output are provided to illustrate areas where more exploration is needed, particularly with respect to extraction of intentions and task-specific goals.
Bio
Professor Dorr joined the Department of Computer and Information Science and Engineering at the University of Florida in 2022 where she directs the Natural Language Research (NLP) Group. Her research focuses on deep language understanding, semantics, language processing using linguistically informed machine learning models, large-scale multilingual processing, explainable artificial intelligence (AI), social computing, and detection of underlying mental states. Her recent contributions have fallen squarely in the realm of cyber-NLP, for example, responding to social engineering attacks and detecting indicators of influence. She has an affiliate appointment at the Institute for Human and Machine Cognition, is Professor Emerita at the University of Maryland, former program manager at the Defense Advanced Research Projects Agency (DARPA), and former president of the Association for Computational Linguistics. She is a Sloan Fellow, NSF Presidential Faculty (PECASE) Fellow, AAAI Fellow, ACL Fellow, and ACM Fellow. In 2020 she was named by DARPA to the Information Science and Technology (ISAT) Study Group. She holds a Master's and a Ph.D. in computer science from the Massachusetts Institute of Technology, with a Bachelor's degree in computer science from Boston University.
This talk presents challenges, solutions, and gaps in AI and Natural Language Processing (NLP), with an emphasis on the need for explainability in the era of ChatGPT. Examples include: machine translation of human languages, ask detection for defending against social engineering attacks, and stance detection for extracting attitudes from social media. Past, current, and future projects face several challenges: (a) brittleness of rule-based linguistic principles for large-scale processing; (b) shallowness of statistical methods and neural language models for understanding implicit information; and (c) lack of “explainability” amidst ever-increasing numbers of black-box models. A case is made for hybrid approaches that combine linguistic generalizations with statistical and neural models to handle implicitly conveyed information (e.g., beliefs and intentions), and also for the implementation of an “explainable” propositional representation that supports the ability of developers and end users to understand what is going on inside the AI system. Questions of interest range from “What is the social engineer’s underlying goal in a two-way interaction?” to “What beliefs support individuals’ attitudes regarding pandemic interventions?” to “How does targeted influence impact attitudes online?”. Such information is generally not extractable from large language models alone and, moreover, such models are hampered in that they are too large to retrain on a regular basis by the average researcher, developer, or customer. Representative examples of ChatGPT output are provided to illustrate areas where more exploration is needed, particularly with respect to extraction of intentions and task-specific goals.
Bio
Professor Dorr joined the Department of Computer and Information Science and Engineering at the University of Florida in 2022 where she directs the Natural Language Research (NLP) Group. Her research focuses on deep language understanding, semantics, language processing using linguistically informed machine learning models, large-scale multilingual processing, explainable artificial intelligence (AI), social computing, and detection of underlying mental states. Her recent contributions have fallen squarely in the realm of cyber-NLP, for example, responding to social engineering attacks and detecting indicators of influence. She has an affiliate appointment at the Institute for Human and Machine Cognition, is Professor Emerita at the University of Maryland, former program manager at the Defense Advanced Research Projects Agency (DARPA), and former president of the Association for Computational Linguistics. She is a Sloan Fellow, NSF Presidential Faculty (PECASE) Fellow, AAAI Fellow, ACL Fellow, and ACM Fellow. In 2020 she was named by DARPA to the Information Science and Technology (ISAT) Study Group. She holds a Master's and a Ph.D. in computer science from the Massachusetts Institute of Technology, with a Bachelor's degree in computer science from Boston University.
Special Track Invited Speaker:
Special Track Invited Speaker:
Applied Natural Language Processing
Knowledge Discovery in Aircraft Maintenance Records
Monday, May 15, 2023, 13:30 - 15:00
Room: Palm (ANLP session)
Abstract:
Aircraft maintenance records log day-to-day maintenance activities performed on in-service aircraft and possess a wealth of crucial knowledge related to parts and systems such as part names and associated issues. Mining such knowledge is critical for prognostics and health management, and in particular is essential to improve safety and quality, lower lifecycle maintenance cost, minimize downtime, improve parts inventory related to an aircraft, and improve manufacturing quality and rework. The maintenance records conveyed in unstructured text are very noisy with frequent use of local jargon, ad-hoc acronyms, misspellings, and non-standard abbreviations. In addition, general methods for preprocessing and knowledge extraction do not work effectively for domain-specific cases. Thus, automatically extracting knowledge from maintenance records is a daunting task. This presentation covers the practical constraints and challenges in discovering knowledge from aircraft maintenance records. It also covers the practical Natural Language Processing and Machine Learning methods to address the challenges.
Bio
Dr. Nobal Niraula is an Associate Technical Fellow at Boeing and works for Boeing Research & Technology AI & Analytics group. His areas of expertise include Natural Language Processing, Text Analytics, Machine Learning, Information Retrieval, and Dialog Systems. He has research experiences in these fields from both universities and research labs including Microsoft Research, AT&T Labs Research, INRIA and CNRS. He has received best research paper awards, authored more than 50 publications in top, peer-reviewed international conferences and journals. He earned a Ph.D. in Computer Science from The University of Memphis, USA, a M.Sc. in communication networks and services from Telecom SudParis, France, a M.E. in information and communication technology from Asian Institute of Technology, Thailand, and a B.E. in computer engineering from Tribhuvan University, Nepal.
Aircraft maintenance records log day-to-day maintenance activities performed on in-service aircraft and possess a wealth of crucial knowledge related to parts and systems such as part names and associated issues. Mining such knowledge is critical for prognostics and health management, and in particular is essential to improve safety and quality, lower lifecycle maintenance cost, minimize downtime, improve parts inventory related to an aircraft, and improve manufacturing quality and rework. The maintenance records conveyed in unstructured text are very noisy with frequent use of local jargon, ad-hoc acronyms, misspellings, and non-standard abbreviations. In addition, general methods for preprocessing and knowledge extraction do not work effectively for domain-specific cases. Thus, automatically extracting knowledge from maintenance records is a daunting task. This presentation covers the practical constraints and challenges in discovering knowledge from aircraft maintenance records. It also covers the practical Natural Language Processing and Machine Learning methods to address the challenges.
Bio
Dr. Nobal Niraula is an Associate Technical Fellow at Boeing and works for Boeing Research & Technology AI & Analytics group. His areas of expertise include Natural Language Processing, Text Analytics, Machine Learning, Information Retrieval, and Dialog Systems. He has research experiences in these fields from both universities and research labs including Microsoft Research, AT&T Labs Research, INRIA and CNRS. He has received best research paper awards, authored more than 50 publications in top, peer-reviewed international conferences and journals. He earned a Ph.D. in Computer Science from The University of Memphis, USA, a M.Sc. in communication networks and services from Telecom SudParis, France, a M.E. in information and communication technology from Asian Institute of Technology, Thailand, and a B.E. in computer engineering from Tribhuvan University, Nepal.