Artificial General Intelligence & Constructivist A.I.

Place: Reykjavik, Iceland
Dates: August 1 2012 – August 15 2012
Tentative Schedule

This summer school focuses on issues related to the original goal of artificial general intelligence, namely that of building machines capable of operating in a range of different environments and domains, and doing a range of unrelated tasks in a coordinated manner, with a special focus on architectural and integrative issues. What some have called “general AI”.

Fundamental AI challenges are presented and dissected, including:

• What kinds of methodologies will be required to achieve artificial general intelligence (AGI)?
• How different will they be from today's software development methods?
• What role can logic and reasoning play?
• How do we construct highly distributed architectures for these purposes?
• Do we need new programming languages?

Starting with the limitations of present software development methodologies based on hand-coded knowledge, we outline new methodologies involving self-organizing and self-programming systems - what we call constructivist AI - and present recent advances in reasoning-based and self-expanding architectures.

Hands-on work will include experiments in the newly-developed programming language Replicode, which has been explicitly created for AI systems capable of self-inspection.

Since the introduction of powered factories in the 1800's, the use of machines for automation of routine tasks has become a staple of modernity. By extension of this paradigm to electronic data, at the arrival of practical computing in the mid-1900's, automating routine information processing has seemed an obvious step in the same direction. But could that same paradigm be used to automate non-routine tasks as well? Could computing power be used to solve problems and tasks that required more than a narrow knowledge of a sequence of steps – tasks that we might say require insight, creativity, invention? This question has stirred a hot debate from the early days of what is now typically referred to by the term “artificial intelligence” (“artificial” because it is hand-made, “intelligence” because it requires more than blind execution of a series of simple steps).

With the advent of ever-more powerful search engines, classification algorithms and machine vision applications, the line between “dumb automation” and “intelligent behavior” seems more blurry than ever. It seems that with the current trends we should eventually see machines have the power of a human mind. However, a growing criticism of the current practices of the field of A.I. has centered on the issue that current research is too narrow, instead of focusing on the core question that started the field, namely, Can we make machines that have *general* intelligence, in the sense that most people understand it? No matter how the problem is addressed, one thing is clear: human-level intelligence calls for a complex system. Are the current software development in popular use up to this task? Arguments are accumulating that new methodologies are needed to lead us to machines that rival the human mind. Chief among the functions that are difficult to address with present software design methods are transversal cognitive functions: functions of thought that seem highly flexible and able to operate for a wide range of human skills, such as general learning abilities, attentional control, and explanatory and introspective abilities.

The summer school is organized around three key threads. The first, and main one, relates to constructionist versus constructivist AI methodologies, self-programming, introspection and logic. The second is the theoretical foundation of AI, the difference between cognitive science and AI, and the basic principles on which AI currently stands. The third and last is the role of AI in society, current academic and industrial methods used to advance it, and societal changes that may be expected from its development in the next 20 years.

The summer school is targeted to those with a background in artificial intelligence and computer science, and may also be of interest to students in philosophy, psychology and cognitive science with a background in software developments or mathematics. It is open to all graduate students in all countries; the teaching language is English. The upper limit of attendees is 35, accepted on a first-come, first-serve basis.

To apply for admission, send an email with a short cover letter and a CV to the following address: summerschool-2012@cadia.ru.is. Acceptance decisions will be sent out via return email, at the latest March 1st 2012. Further details will be announced on this website. Questions about the summer school can be sent to thorisson-<at>-ru.is.

The tuition for the event is 250 Euros. Living cost for a 2-week stay in Reykjavik ranges from 1300 Euro to 2600 Euro, depending on location, quality and amenities. A limited amount of support may be provided to a subset of applicants; to apply, please indicate this in your cover letter, and clearly state reasons for requesting the support.

Here are some suggested places that offer reasonably-priced rooms for rent.

* Capital Inn www.capitalinn.is has available budget accommodation:

• Single room 9.500,-kr.
• Twin room 12.400,-kr.
• Triple room 15.300,-kr.
• Hostel sleeping bag accommodation 3.590,-kr.
• Hostel accommodation including bed-linen 4.090,-kr.

* Kex Hostel www.kexhostel.is offers budget accommodation for these prices:

• Single room 13.400,-kr.
• 4 bed dorm 5.100,-kr.
• 6 bed dorm 4.900,-kr.
• 10 bed dorm 4.300,-kr.
• 16 bed dorm 3.200,-kr.

Taxi from either of these to RU should take approximately 10 minutes with price around ISK 2000 (taxi rates are regulated so any variation in prices reflect traffic or time of day).

If you prefer public transport - bus Nr. 19 stops in front of RU (for bus schedule & map please visit: http://www.straeto.is/english).

• Tentative Schedule

Dr. Thórisson Dr.Thórisson has been researching artificial intelligence for two decades, in academia and industry. His research centers on realtime interactive intelligences, complex cognitive systems and mind models. At MIT he pioneered new ideas in the area of communicative, multimodal intelligent agents. Recent projects include developing a cognitive architecture for the humanoid robot ASIMO by Honda Motor Corporation. He is the co-founder of CADIA, Iceland’s first AI lab, and Radar Networks, a Semantic Web company in San Francisco, and is the Founding Director of the Icelandic Institute for Intelligent Machines. He has taught advanced AI courses at Columbia University, KTH and Reykjavik University, and consulted for NASA and British Telecom, among others. Kris has authored numerous scientific papers and sits on the editorial board of the Journal of Artificial General Intelligence and the LNCS Transactions on Computational Collective Intelligence. Dr. Thórisson is PI of the HUMANOBS FP7 project with Eric Nivel, which is poised to break new ground in the field of AGI in the next 12 months.

• Cognitive architecture for ASIMO
• Co-founder of Center for Analysis & Design of Intelligent Agents, Reykjavik University
• Founding Director, Icelandic Institute for Intelligent Machines
• PI of the HUMANOBS Project
  
• Editorial boards:
  • LNCS Transactions on Computational Collective Intelligence
  • Journal of Artificial General Intelligence

Dr. Wang is an Associate Professor at Temple University and an Affiliate Researcher at IIIM in Reykjavik. He is the author of Rigid Flexibility: The Logic of Intelligence (Springer 2006, Applied Logic Series). He received his Ph.D. in Computer Science and Cognitive Science from Indiana University, and his M.S. and B.S. in Computer Science, both from Peking University. His research interests include artificial intelligence and cognitive science, especially on unified theories of intelligence, formal models of rationality, reasoning under uncertainty, learning and adaptation, knowledge representation, and real-time decision making. Dr. Wang is the Chief Executive Editor of the Journal of Artificial General Intelligence.

• Author of NARS (Non-Axiomatic Reasoning System)
• Editor-in-Chief: Journal of Artificial General Intelligence
• Home Page

Mr. Nivel is a senior R&D engineer at Reykjavik University's Center for Analysis & Design of Intelligent Agents (CADIA). Eric's wide range of technical expertise includes evolutionary computing and architectures for autonomous systems to real time distributed systems and high-performance computing. He has developed a self-rewriting computational substrate for general autonomous systems, some of which has been used as primary technical frameworks for professional theatrical art works. His work was awarded sponsorship by the French Ministry of Culture and the French Agency for Research. Eric has extensive experience in research and project management. He was the lead software engineer at Holografika in Hungary, where he designed the software for a 2.5 x 2 x 1.5 meter realtime hologram generation and transmission system, and participated in various related FP6 projects like HOLOVISION, for which he coordinated the software engineering. He also consulted for research, telecom and aerospace concerns like Institut Pasteur, France Telecom, and Dassault Aviation. Eric currently leads the HUMANOBS FP7 project with Dr. Thórisson.

• Co-Director of science and engineering, HUMANOBS FP7 project
• Extensive experience in software engineering and advanced A.I. projects
• Author of the Ikon Flux AGI architecture

Dr. Bjornsson is the director and co-founder of CADIA (with Kristinn R. Thórisson). His main research focus is in developing search techniques as a general problem-solving/inference mechanism, and applying them to a wide variety of problem domains. At Reykjavik University he is heading a project into building general game-playing agents; such agents can automatically learn how to play many different games at an expert level without any human intervention. A successful realization of this task poses interesting research challenges for artificial intelligence sub-disciplines such as knowledge representation, logic inference, agent-based reasoning, heuristic search, and machine learning. Yngvi's agent, CADIA-Player, won the third and fourth International General Game Playing Competitions (2007, 2008). Before joining Reykjavík University in 2004 he was a research associate with the GAMES research group at University of Alberta, Canada, working as a part of a team developing techniques or solving large-scale search problems, such as solving the game of checkers (named by Science Magazine as one of top ten scientific achievements of the year 2007). Yngvi's extensive experience in building high-performance search/inference systems will be valuable to the project in various ways, especially for the development of the reasoning reand learning capabilities of the system.

• Author of CADIA-Player, two-time winner, GGP Competition at Stanford
• Co-founder of CADIA

Dr. Sanz is professor in Automatic Control and Systems Engineering at the University of Madrid (Universidad Politécnica de Madrid) and specializes in the frontier between control, computing and intelligence–embedded systems, real-time distributed systems, software engineering, artificial intelligence and cognitive systems. He has been involved in, and led, numerous EU-funded research projects in the field of real-time distributed systems and complex intelligent controllers. He is the chairman of the International Federation of Automatic Control Technical Committee on Computers and Control and directs the Autonomous Systems lab at University of Madrid. He has been evaluator and reviewer for the European Commission in the fields of real-time embedded systems, cognitive systems, dependability and complex systems.

• Director of Autonomous Systems lab

Dr. Dindo is Assistant Professor in Computer Engineering at the University of Palermo. From 2009 he is member of the scientific board in the EU-funded project “HUMANOBS: Humanoids that Learn Socio-Communicative Skills by Observation”. He has consulted for 3rdi Technology Inc. by promoting technology transfer in the field of semantic search engines, user profiling and intelligent advertising. He is member of AAAI and AI*IA. His research focus is in the field of social robotics and human-robot interaction, social learning and language acquisition. In particular, he concentrated on the problem of transferring advanced cognitive skills from human to robots in a holistic approach involving several complex abilities, such as vision, cross-modal coordination, motor control, extraction of affordances, knowledge representation, learning and symbol manipulation. Haris Dindo has published over 40 articles in international journals and peer-reviewed conference proceedings, and participated in numerous national and international projects.

• Member of the Robotics Lab, University of Palermo

Home Page

Dr. Kühnberger is a Professor of Artificial Intelligence at the Institute of Cognitive Science, University of Osnabrück (IKW), Germany. He obtained a PhD in computational linguistics of the University of Tübingen. He has co-authored or co-edited over 100 publications in the areas of analogical reasoning, ontology design, neural-symbolic integration, cognitive architectures, and other fields. He co-organized several conferences and workshops including the CogSci’07 Workshop on Analogies: Integrating Multiple Cognitive Abilities in Nashville Tennessee, the Processing Text Technological Resources conference at the ZiF in Bielefeld, and the European Cognitive Science Conference 2003 in Osnabrück. In 2009 Kai-Uwe Kühnberger was visiting fellow of the Scottish Information and Computer Science Alliance (SICSA). He is survey editor of the Elsevier Journal Cognitive Systems Research, editor of the book series Thinking Machines (Atlantis/Springer), and Executive Editor of Journal of Artificial General Intelligence.

Home page

Dr. Goertzel is the founder of AGI conference and the AGI Society. He is the author of numerous books on AGI and related topics including Probabilistic Logic Networks (Springer-Verlag 2008) and Advances in Artificial General Intelligence (Springer-Verlag 2007). He currently leads Novamente LLC, a privately held software company that attempts to develop a form of strong AI. He is also the CEO of Biomind LLC, a company that markets a software product for the AI-supported analysis of biological microarray data; and he is an advisor to the Singularity Institute for Artificial Intelligence, and formerly its Director of Research. He is also Executive Editor of Journal of Artificial General Intelligence.

Home page

Dr. Bringsjord specializes in the logico-mathematical and philosophical foundations of artificial intelligence (AI) and cognitive science, and in collaboratively building AI systems on the basis of computational logic. He is author of What Robots Can & Can't Be (Kluwer 1992) and Superminds: People Harness Hypercomputation, and More (Kluwer 2003). He is the Chair of the Department of Cognitive Science at Rensselaer Polytechnic Institute and the Director of the Rensselaer AI & Reasoning (RAIR) Lab. (RPI).

Center for Analysis and Design of Intelligent Agents
Associate Professor, School of Computer Science, Reykjavik University
Director, Icelandic Institute for Intelligent Machines
Menntavegur 1
101 Reykjavik, Iceland
thorisson at ru.is
+354 898 0398

Temple University
USA

Center for Analysis and Design of Intelligent Agents
and School of Computer Science, Reykjavik University
Menntavegur 1, Venus, 2. h.
101 Reykjavik, Iceland

Icelandic Institute for Intelligent Machines
and SCS, Reykjavik University
Menntavegur 1
2. h. Uranus
101 Reykjavik
Iceland