Research on Icelandic Banking System

A student (B.Sc or M.Sc) is needed with interest in the Financial System to assist with research on Index linked and housing association loans in Iceland. The research will be to establish the exact terms of loans made, and also to determine their accounting treatment by the Banks, and will assist with building a model of the Icelandic banking system and preparing a paper describing its behaviour over time. Estimated time 1-2 months at 50%, but with flexible scheduling.

Graphical Displays

A student (M.Sc or advanced B.Sc) with good programming skills is needed to take an existing simulation of the Fractional Reserve Banking System and extend it to have an animated display of the money flows created by the debt created by the system over time. Project is envisioned as an opengl extension written in python; but java or C++ could also be used. Estimated time 2-3 months at 50% effort. There is scope for extending the project beyond the initial task to create a graphical construction tool that would allow users to easily setup banking networks and debt flows to explore specific scenarios, and providing a web based implementation in javascript.

Economic History/Simulation

A student with interest in the financial system and programming (M.Sc or Ph.D) is needed to research the gold standard system, determine the regulatory framework it used, and extend an existing simulation of the fractional reserve banking system written in python to be a full reproduction of the gold standard framework. Project would be the first simulation of the gold standard framework, and would contribute to understanding of historical financial instability in the 19th and early 20th centuries. Estimated time 3-6 months at 50%.

Create a Simple Model of Currency Flows

Using java or C++/QT, create a simulation that allows currency flows to be modelled between countries, by assuming different rates of expansion of the underlying money supply and varying numbers of exchange transactions. Graphical display of resulting network flows. Estimated time, 3 months @ 50%.

Spatio-semantic Navigation

We are looking for a student(B.Sc or M.Sc) to extend an existing software package for performing robot navigation combined with semantic reasoning to produce a demo and help prepare a paper on the system. Student must have strong C++ programming skills, particularly on Windows. This task should take approximately 2-3 months at 50% effort, but there is some flexibility in scheduling.

Collaboration Project With One of Our Industry Partners

Starting probably in February, we need one student (B.Sc or M.Sc) to perform basic experiments with supervised learning algorithms on a small, but potentially complex data set. Student should be familiar with basic machine learning algorithms, such as neural networks, decision trees, etc. This task should take approximately one month at 50% effort, and the student will need to devote at least 50% time to the project. This work could possibly extend into a second phase lasting several more months.

Learning Natural Turn-taking Behavior

Starting sometime aroundthe summer, we need a graduate student (M.Sc or Ph.D) to work on developing and testing machine learning methods in a human-computer dialog system. Requires strong C++ programming skills, and student should ideally be prepared to stay with the project for a period of at least one year.
Radio Show Host
We need one student (B.Sc or M.Sc) to work on various data gathering modules for an artificial radio show host. There are various potential sources of knowledge that the radio host could use such as musicbrainz database and Google queries but the results need to be formatted into text that can be used for spoken dialogue or monologue. This task should take 4-6 months at 50% effort, but there is some flexibility in scheduling.

Creating an Annotated Database of Spoken Sentences From Human Speakers

People give indications at the end of each sentence of whether they are giving or holding a turn, by modifying their pitch slightly. We need a student (B.Sc or M.Sc) to gather audio samples of human speakers into a database of dialogue speech samples. On top of that a program module has to be created that can fetch and play audio clips from this database given some search criteria to simulate a human speaker. This task should take 2-3 months at 50% effort.

For the past few years Kristinn R. Thorisson has been collaborating with the excellent team at Honda Research Institute USA and Communicative Machines in developing integrated cognitive architectures for humanoid robots; their developed Cognitive Map Architecture (PDF) which, for example, enables ASIMO to play board games with kids.

Underlying Technology & Approach

The work relies on Psyclone, a powerful middleware that mediates between disparate modules; vision system, speech recognition, and spatio-temporal understanding mechanisms, as well as Asimo’s body. The software supports Thórisson’s Constructionist Design Methodology, which was the underlying approach to develop the Cognitive Map Architecture (PDF) which coordinates Asimo’s interpersonal behaviors.

Find the video news clip on the website of the Icelandic National Broadcasting Service (Note: audio is in Icelandic).

About ASIMO

ASIMO is a humanoid robot created by Honda to help people who lack full mobility. It was first introduced in year 2000 and has been improved ever since. It can now run (at 9 kph), hop on one or both legs, and pour drinks. Find more information about ASIMO on HONDA’s official website for ASIMO.

[[Note: Post updated with additional links and information, Dec-7-2011]]

WEDNESDAY, NOVEMBER 16
we will present few short talks about
Simulation and Artificial Intelligence Research at the Icelandic Institute for Intelligent Machines

Time: 12:00 – 13:00. Place: Reykjavik University, V102

• Robust Continuous Machine Learning by Dr. J. Deon Garrett, IIIM
• Building Artificial Economies by Dr. Jacky Mallett, IIIM
• Socially Intelligent Characters by Claudio Pedica, IIIM

We welcome you to visit us and hear our researchers present their projects and explain how collaboration with our partners contributes to better results.

IIIM is one of several participants in Global Entrepreneurship Week at Reykjavík University. Full schedule is available here.

From Global Entrepreneurship Week web site:

Global Entrepreneurship Week is the world’s largest celebration of the innovators and job creators who launch startups that bring ideas to life, drive economic growth and expand human welfare.

During one week each November, GEW inspires people everywhere through local, national and global activities designed to help them explore their potential as self-starters and innovators. These activities, from large-scale competitions and events to intimate networking gatherings, connect participants to potential collaborators, mentors and even investors—introducing them to new possibilities and exciting opportunities.

IIIM Projects at Vísindavaka

• Research Scientist J. Deon Garrett presented Augmenting Spatial Navigation Structures with Semantic Information. The goal of this project is to develop a mechanism by which a robot can annotate physical environment to give it greater intelligence, behavior and capabilities. The robot might be told, “This is the kitchen”, and then be able to reason from that information to answer spatial questions like, “How far is it to the kitchen now?”
• Research Scientist Jacky Mallett was presenting her recent work on Banking System modeling and simulation that raises questions about the accuracy of current monetary and economic theory.
• Senior Programmer Guðný R. Jónsdóttir’s project Conversation with Polite Computers was presented in the form of a poster and short sample video. The main goal of this project is to create autonomous computer characters that can interact with people in a humanlike manner.

Thank you for coming and see you at next Vísindavaka!

Links

• Official website of Vísindavaka 2011

Discussion and Presentations in a Nutshell

The presentations covered general issues such as what distinction, if any, should be made between self-programming and more conventional ideas of learning (Goertzel), and the appropriate role of self-programming in the development of AGI systems; e.g. can self-programming be used to help generate AGI systems, or does a human-level intelligence need to be in place for the system to meaningfully utilize these capabilities.

Other presenters focused on particular aspects or implementations of self-programming systems. Such as the role of creativity (Leijnen), imitation (Hall) and program-space search (Skaba). Several existing architectures were discussed as they partained to self-programming capabilities. One in particular was BECCA, a general system for feature creation and general reinforcement learning (Rohrer). Also NARS, an AI system based on non-axiomatic reasoning (Wang), and a proposed system based on the idea of “emergent inference” (Pissanetzky).

Videos, Papers and Resources

IIIM offers full papers and extra materials for downloading.

Time & Location

Date: Monday, August 29, 2011
Time: 16:00 – 17:00
Location: Reykjavik University, Room V1.08 (Map)

The Center for Analysis & Design of Intelligent Agents at Reykjavik University and the Icelandic Institute for Intelligent Machines offer a lecture by Pei Wang (Temple University)

Lecture Abstract
When applying mathematical logic to explain and reproduce cognition and intelligence, many challenges are encountered. These challenges all have a common nature, that is, they all exist outside mathematics, the domain for which mathematical logic was designed. This suggests that the logic of cognition and the logic of mathematics may be fundamentally different, and the former cannot be obtained by partially revising or extending the latter. In this talk a new cognitive logic is introduced. The logic uses a categorical language which is close to natural languages, and an experience-grounded semantics which covers several types of uncertainty. The logic has inference rules for deduction, induction, abduction, revision, analogy, etc., all within the same framework. When implemented in a reasoning system, such a cognitive logic provides unified solutions to fundamental problems in artificial intelligence and cognitive science.

Pei Wang – Bio

Dr. Pei Wang is an Associate Professor at Temple University. 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.

Brought to you by the Icelandic Institute for Intelligent Machines and
the Center for Analysis & Design of Intelligent Agents

The HUMANOBS Workshop is focused on the impending paradigm shift from constructionist to constructivist methodologies for building artificial intelligence. Constructionist refers to traditional top-down software design, which places serious limitations on scaling, while Constructivist refers to self-organizing systems that manage their own growth.

Update, August: Please note that the final program agenda is subject to change. Final details are kept up to date on the official HUMANOBS workshop website.

Incentives for Constructivist AI

The development of artificial intelligence systems has to date been largely one of manual labor. This Constructionist approach to A.I. has resulted in a diverse set of isolated solutions to relatively small pieces of the much larger question about how to artificially replicate the fundamental principles of intelligence.

Small success stories of putting these pieces together in robotics, for example, has made people optimistic that continuing on this path would lead to artificial general intelligence. This is unlikely. Standard software development methods, with their standard top-down design approach, put serious limitations on scaling; in A.I. the Constructionist approach results in systems with limited domain application and severe performance brittleness. Genuine integration, as required for general intelligence, is both practically and theoretically precluded.

Going beyond current A.I. systems requires significantly more complex integration than attempted to date, especially regarding transversal functions such as attention and learning. The only way to address the challenge is replacing top-down architectural design as the main development methodology with methods focusing on self-generated code and self-organizing architectures. We call this Constructivist A.I.

Support Stipend Program

A Support Stipend Program is being offered for 5 students and/or junior scientists (PhD students or post-docs from less than two years) attending HUMANOBS workshop, to be held in Reykjavik, Iceland, on September 1st and 2nd, 2011. The selected individuals will receive 750 Euro stipend to support them in their attendance; attendance on both days of the HUMANOBS workshop is required. Funds will be issued after the workshop. The individuals will be asked to present their work in the form of a poster.

For more information on how to apply, visit HUMANOBS workshop website.

TENTATIVE AGENDA

SEPTEMBER 1st

• 09:00 – 09:25 — Welcome, introduction to the workshop / Day 1
• 09:30 – 10:10 — Presentation 1

Jörg Siekman
AUTOMATED REASONING: FROM SEARCH-BASED TECHNIQUES TO PLANNING
Presentation of challenges 1 & 2

• 10:10 – 10:20 — Team discussion rules presented
• 10:20 – 11:05 — Team collaboration on challenges 1 & 2
• 11:05 – 11:20 — COFFEE BREAK
• 11:20 – 12:05 — Challenges 1 & 2: Team conclusions and solutions
• 12:05 – 13:30 — LUNCH
• 13:30 – 14:00 — Presentation 2

Eric Baum
WHAT HUMAN GO PLAYERS DO THAT MY SYSTEM DOESN’T
Presentation of challenges 3 and 4

• 14:00 – 14:45 — Team collaboration on problems 3 & 4
• 14:45 – 15:30 — Challenges 3 & 4: Team conclusions and solutions
• 15:30 – 15:50 — COFFEE BREAK
• 14:45 – 15:15 — Presentation 3

[Author TBD]
[Topic TBD]
Presentation of challenges 5 & 6

• 15:15 – 16:00 — Team collaboration on challenges 5 & 6
• 16:00 – 16:45 — Challenges 5 & 6: Team conclusions and solutions
• 16:45 – 17:00 — Conclusions of Day 1

SEPTEMBER 2nd

• 09:00 – 09:15 — Introduction to schedule for Day 2
• 09:20 – 09:50 — Presentation 1

Pei Wang
HOW TO CONTROL A ROBOT USING A REASONING SYSTEM
Presentation of challenges 7 & 8

• 09:50 – 10:35 — Team collaboration on challenges 7 & 8
• 10:35 – 11:20 Challenges 7 & 8: Team conclusions and solutions
• 11:20 – 11:40 — COFFEE BREAK
• 11:40 – 12:10 — Presentation 2

[Author TBD]
[Topic TBD]
Presentation of challenges 11 & 12

• 12:10 – 13:30 — LUNCH
• 13:30 – 14:15 — Team collaboration on challenges 9 & 10
• 14:30 – 14:45 — Challenges 9 & 10: Team conclusions and solutions
• 14:45 – 15:15 — Presentation 3

Kristinn R. Thórisson & Eric Nivel
FROM CONSTRUCTIONIST TO CONSTRUCTIVIST AI: ARCHITECTURE MATTERS
Presentation of challenges 9 & 10

• 15:15 – 16:10 — Team collaboration on challenges 11 & 12
• 16:10 – 16:30 — COFFEE BREAK
• 16:30 – 17:20 — Challenges 11 & 12: Team conclusions and solutions
• 17:20 – 17:35 — Conclusions of Day 1
• 17:35 – 17:45 — Summary of Workshop / Introduction to Day 3
• 19:00 – 21:30 — Dinner

SEPTEMBER 3rd

• 08:00 — Half-day excursion to Iceland’s countryside

About HUMANOBS

The HUMANOBS project started in January 2009 and runs until year-end 2011. The goal of the project is to create an artificial intelligence that can learn socio-communicative skills by observing and imitating people. The three main objectives of the project are to build auto-reconfiguring architecture, behavior observation mechanisms and behavior generation and coordination mechanisms. As a result demonstration a virtual humanoid television host will be created, capable of taking interviews and conducting a 30-minute TV program.

More detailed information on the project can be found on HUMANOBS official web site.

You can jump right to the interview at PhysOrg.com titled “Dr. Ben Goertzel on Artificial General Intelligence, Transhumanism and Open Source“.

About Ben Goertzel

Dr. Ben Goertzel is a prominent leader in the wave of advances in fields of artificial general intelligence. He is CEO of Biomind LLC, a company providing AI-supported analysis of biological data, as well as an open-sourced version of many of the company’s software tools—aptly named OpenBiomind. He is an advisor to the Singularity Institute for Artificial Intelligence, and its former Director of Research.

You can find out more about Ben on his Official Website.

Time & Location

10th of June, 2011 at 13:00, in Reykjavik University, Betelgás (Venus, 1st floor)

Dr. Jacky Mallet, Ph.D. graduate from MIT, will be holding a talk at Reykjavik University on her research on modern banking systems and economic theory. Using complex dynamic systems, her computer simulations have yielded evidence challenging some of the assumptions on which modern economic theories are based.

A number of deficiencies in the textbook model of the banking system’s behaviour will be demonstrated in the talk, titled Analysing the Behaviour of the Textbook Fractional Reserve Banking Model as a Complex Dynamic System.

Brought to you by the Icelandic Institute for Intelligent Machines and
the Center for Analysis & Design of Intelligent Agents

Dr. Jacky Mallet

Dr. Mallet has worked, amongst other things, as a Research Assistant at MIT’s Object Based Media Group and as a Team Leader of R&D at Sony. She currently works as a Distributed Systems Architect at CCP Games, Inc. She is an affiliate researcher of the Icelandic Institute for Intelligent Machines.

Brief Overview of the Presentation

Modern banking systems are based on a recursive relationship between lending and deposits that is generically known as fractional reserve banking. The simplified model of the system presented in introductory economic textbooks is notably deficient, in that it fails to include either loan defaults or loan repayments.

In this talk will be briefly discussed the historical context of economic theory on the operation of the banking system, and reviewed the different forms it has taken over the several centuries it has been in use. A simple computer simulation of the textbook model will then be presented which includes loan repayments and interbank lending, and a number of deficiencies in the textbook models predictions will be demonstrated.

For more information, please contact us.