Architectures and Evaluation for Generality, Autonomy & Progress in AI

August, 2019, Macao, China

2ND International Workshop held in conjunction with IJCAI 2019

Call for Papers

Architectures and Evaluation for Generality, Autonomy & Progress in AI

Call for Papers

The proposed Second Workshop on Architectures and Evaluation for Generality, Autonomy and Progress in AI (AEGAP-19) focuses on the original grand dream of AI: the creation of autonomous agents with general intelligence comparable to or exceeding that of humans. “Generality” and “autonomy” are to be interpreted in their widest sense, the former: systems that can handle a wide variety of data, situations and tasks, that can explain themselves, and that are trustworthy; the latter: systems do not need constant attention and re-adjustment from their creators, can acquire knowledge cumulatively, and evaluate their own progress on complex tasks, to give some examples.

With the remarkable recent increase in interest in and work on AI, a growing number of researchers find it relevant to return to a domain-independent and problem-independent focus on intelligence and the development of AI systems with greater levels of autonomy. Sound engineering principles, as well as the history of AI, present good evidence that a general-purpose system cannot be built by bundling many special-purpose systems together, since design objectives, working environments, evaluation criteria, progress paths, emerging features, etc., may be very different for a fully autonomous system than for those developed to solve specific, fully specified problems.

Since there is still no widely accepted theory on this, AEGAP aims to bring together researchers from different sub-disciplines to discuss how the different approaches and techniques can contribute to the goal of building beneficial AI with high levels of generality and autonomy. To achieve this goal, we will likely need to build large-scale, complex and dynamic architectures that can integrate bottom-up and top-down approaches. One hopeful avenue may be to combine logic- or rule-based top-down approaches with neuroscience-inspired bottom-up approaches, so that intelligence might emerge from their interplay. Whatever paths forward will be explored, the search space for promising designs is likely a large one.

Progress must be evaluated, using methods that can accommodate and compare a diverse set of approaches, both those being developed currently and in the future. While benchmarking the AI systems performance in specific domains and on specific tasks is not too difficult, assessing progress in autonomous learning agents is much more challenging when the focus is on generality and autonomy, as real progress in this direction only takes place when a system exhibits enough autonomous flexibility to find a diversity of solutions for a range of tasks, some of which may not be known until after the system is deployed. Furthermore, autonomous system must also be responsible and explainable in a certain way for the AI to be beneficial to the human society.

The AEGAP-19 IJCAI Workshop welcomes regular papers, short position papers, and papers accompanied by demonstrations of all relevant theoretical and technical aspects of the Workshop’s theme, including (but not limited by) the following topics:

  • New theoretical insights relevant to generality and autonomy
  • Analysis of design requirements for generality and autonomy
  • New methodologies relevant to generality and autonomy
  • Design proposals for cognitive architectures targeting generality and/or autonomy
  • Analysis of the potential and limitations of existing and new approaches
  • Synergies and integration of AI approaches
  • New programming languages or architectural principles relevant to generality and autonomy
  • Novel network architectures for generality and autonomy
  • New learning and/or educational methods for generality and autonomy
  • Analysis, comparisons and proposals of AI/ML benchmarks and competitions
  • Tasks and methods for evaluating general and autonomous AI systems
  • Unified theories for evaluating general intelligence and cognitive capability
  • Evaluation of multi-agent systems in competitive and cooperative scenarios
  • Better understanding of the characterization of task requirements and difficulty (energy, time, trials needed...) beyond algorithmic complexity
  • The relation between autonomy and predictability
  • Emergence of (symbolic) logic from neural networks
  • Integration of top-down and bottom-up approaches (e.g. logic-based and neural-inspired)
IJCAI-18

Key Information:

When:
August 10th
Where:
To be Determined
Address:
Macao, China

Paper & Demo submission:

Due date:
April 30th, 2019
Notification date:
May, 10th, 2019
Camera-ready date:
To Be Determined
Submission system:
EasyChair

Contact:

xiangliAGI@temple.edu

KEYNOTE TALKS

Coming Soon

tentative

Program

Coming Soon

Pre-Proceedings

Coming Soon

Organization Committee


Dr. Kristinn R. Thórisson (contact)

Reykjavik University & Icelandic Institute for Intelligent Machines, Iceland

Dr. Hiroshi Yamakawa

University of Tokyo & Dwango AI Lab, Japan


Dr. Itsuki Noda

National Institute of Advanced Industrial Science and Technology, Japan

Dr. Ryutaro Ichise

National Institute of Informatics, Japan


Dr. Satoshi Kurihara

University of Electro-Communications, Japan
 

Xiang Li

Temple University, United States
 

Steering Committee


Dr. Kristinn R. Thórisson (contact)

Reykjavik University & Icelandic Institute for Intelligent Machines, Iceland

Dr. José Hernández-Orallo (contact)

Polytechnic University of Valencia, Spain

Dr. Pei Wang

Temple University, U.S.
 

Jordi Bieger

Delft University of Technology, The Netherlands &
Reykjavik University, Iceland

Dr. Hiroshi Yamakawa

University of Tokyo & Dwango AI Lab, Japan


Dr. Satoshi Kurihara

University of Electro-Communications, Japan
 

Program Committee


Eizo Akiyama Tsukuba University, Japan
Joscha Bach Harvard University, U.S.
Tarek Richard Besold City University of London, U.K.
Jordi Bieger Delft University of Technology, The Netherlands & Reykjavik University, Iceland
Selmer Bringsjord Rensselaer Polytechnic Institute, U.S.
Miles Brundage University of Oxford, U.K.
Lola Cañamero University of Hertfordshire, U.K.
Antonio Chella University of Palermo, Italy
Haris Dindo Yewno & University of Palermo, Italy
David Dowe Monash University, Australia
Kenji Doya Okinawa Institute of Science and Technology, Japan
Emmanuel Dupoux EHESS, France
Jan Feyereisl AI Roadmap Institute & GoodAI, Czech Republic
Patrick Hammer Temple University, U.S.
Helgi P. Helgason Activity Stream, Iceland
Bernhard Hengst University of New South Wales, Australia
Sean Holden University of Cambride, U.K.
Hidenori Kawamura Hokkaido University, Japan
David Kremelberg Icelandic Institute for Intelligent Machines, Iceland
Satoshi Kurihara Keio University, Japan
Othalia Larue Wright State University, U.S.
Ramon Lopez de Mantaras AI Research Institute or the Spanish National Research Council, Spain
Richard Mallah Future of Life Institute, U.S.
Tomas Mikolov Facebook AI Research, U.S.
Itsuki Noda National Institute of Advanced Industrial Science and Technology, Japan
Frans A. Oliehoek University Of Liverpool, U.K.
Satoshi Ono Kagoshima University, Japan
Laurent Orseau DeepMind, U.K.
Ricardo B.C. Prudencio Federal University of Pernambuco, Brazil
Gavin Rens KU Leuven,Belgium
Hiroyuki Sato University of Electro-Communications, Japan
Ute Schmid Universität Bamberg, Germany
Murray Shanahan Imperial College London & DeepMind, U.K.
Carles Sierra IIIA-CSIC, Spain & UT Sydney, Australia
Jim Spohrer IBM Research, U.S.
Bas Steunebrink NNAISENSE, Switzerland
Claes Strannegård Chalmers University of Technology & University of Gothenburg, Sweden
Reiji Suzuki Nagoya University, Japan
Tadahiro Taniguchi Ritsumeikan University, Japan
Kristinn R. Thórisson Reykjavik University & Icelandic Institute for Intelligent Machines, Iceland
Hiroaki Wagatsuma Kyushu Institute of Technology, Japan
Pei Wang Temple University, U.S.
Hiroshi Yamakawa Dwango Artificial Intelligence Laboratory, Japan
Masahito Yamamoto Hokkaido University, Japan

Submission

Papers should be between 2 and 12 pages (excluding references) and describe the authors' original work in full (no extended abstracts). Formatting Guidelines, LaTeX Styles and MS Word Template can be downloaded from here. Papers will be subjected to peer-review and can be accepted for oral presentation and/or poster presentation. For papers that have previously been submitted to IJCAI and rejected, we ask authors to append the reviews and their responses to aid our review process.

Proposals for Demonstrations should be accompanied with a 2-page description for inclusion in the workshop's pre-proceedings. Examples include, but are not limited to: (interactively) demonstrating new tests or benchmarks, or the performance of a robot, (cognitive) architecture or design methodology.

Oral presentations should be given by one of the authors during one of the Contributed Talks Sessions.

Accepted papers will be gathered into a volume of pre-proceedings and published on this website before the workshop. We are looking into the possibility of producing a special issue for an archival journal.