Human-Like Computing Machine Intelligence Workshop

Lead Research Organisation: Imperial College London
Department Name: Dept of Computing

Abstract

Recent prominent advances in statistical AI, such as the success of
driverless cars and Alpha Go's recent defeat of
the 9-dan ranked South Korean Go player Lee Se-dol have considerably
increased public awareness of the potential for computers to achieve
human-level competence in settings which were previously the exclusive
realm of human judgement and skill. However, in human terms
statistical AI techniques still act as skilled zombies, having
unconscious skills detached from the ability to express goals or intentions.
The lack of ability to communicate in a co-operative fashion with
human beings limits the long-term potential of these approaches
in many commercially and socially valuable settings.

The workshop will address issues related to interaction in which human co-operative and communicative skills
can be studied and modelled computationally. While many of these issues have
been prominent for some time in symbolic AI, we are yet
to see successful integration of statistical and symbolic AI approaches
which achieves the broad range of phenomena present in human behaviour.
This workshop aims to contribute to the objectives of the EPSRC's Human-Like
Computing priority area by crystalising the key issues which need to
be prioritised in this area.

Planned Impact

It became clear from discussions at the EPSRC Bristol workshop on Human-Like Computing that
different communities of researchers were out of touch with latest
research and trends within areas far from their own. This led
to the suggestion of the value in holding an inter-disciplinary
workshop in which leading experts in these fields provided overviews
and descriptions of latest research findings to those from other
fields. It was agreed that the format and UK tradition of the Machine
Intelligence series would provide an ideal forum for holding such a workshop,
and that the EPSRC should be approached for funding.
As a consequence the proposed workshop will
have the theme of ``Human-Like Computing''. It was further agreed that
invited papers from the workshop will be published in a special
issue of the journal Topics in Cognitive Science. Ulrike Hahn,
a member of the editorial board of the journal, has been tasked with
organising this special issue.

For the limited cost such a workshop is expected to have considerable
long-term impact in helping form a community of researchers working on some
of the hardest unsolved problems in bot Artificial Intelligence and Cognitive
Science.

Publications

10 25 50
 
Title BSD 3-Clause for the Metagol system. 
Description Metagol is an inductive logic programming (ILP) system based on meta-interpretive learning. Please contact Andrew Cropper (andrew.cropper@cs.ox.ac.uk) with any questions / bugs. If you use Metagol for research, please use this citation or cite the relevant paper. 
Type Of Art Artefact (including digital) 
Year Produced 2017 
Impact The system source code has been made available on a Github from which it has been widely distributed. The impact, measured by Google Scholar citations to the related papers was 54 citations in 2017, and a total of 219. 
URL https://github.com/metagol/metagol
 
Description The Machine Intelligence 20 workshop was supported by the EPSRC grant and was held at Cumerbland Lodge during 23-25 October 2016. Human-Like Computing (HLC) research aims to endow machines with human-like perceptual, reasoning and learning abilities which support collaboration and communication with human beings. Such abilities should support computers in interpreting the aims and intentions of humans based on learning and accumulated background knowledge to help identify contexts and cues from human behaviour. Techniques of this kind are required in applications in which close interactions are required between computers and human users. The workshop brought together leading experts in AI and Psychology to investigate areas of fruitful interaction in which human co-operative and communicative skills can be studied and modelled computationally. While many of these issues have been prominent for some time in symbolic AI, we are yet to see successful integration of statistical and symbolic AI approaches which achieves the broad range of phenomena present in human behaviour. The workshop helped identifying the objectives of the EPSRC's HLC area by crystalising the key research issues to be prioritised in this area.

In particular, it was agreed at the meeting that the following eight topics were of particular common interest for ongoing interactions between the AI and Psychology communities.

1. Representation Change - implementation, role and psychology
of creativity
2. High to low-level learning - bridging the gap
3. Value of forgetting in people and AI systems
4. Comprehensibility, language, meaning and learning, explanation,
accountability
5. Verbal versus non-verbal communication
6. Small data learning- AI challenge and connection to cognition
7. Social Robotics - theory of minds - Self Perception - Context cues
Spatial reasoning
8. Automated programming - psychology and application
Exploitation Route We have planned a publication as a special issue in the Journal "Topics in Cognitive Science". The special issue is being coordinated by Ulrike Hahn.

A roadmap document for the area has been developed by Nigel Birch with input from the HLC MI20 community.
Following discussions with EPSRC we are planning the submission of a Network Plus grant application.
Sectors Agriculture, Food and Drink,Chemicals,Construction,Creative Economy,Digital/Communication/Information Technologies (including Software),Environment,Financial Services, and Management Consultancy,Pharmaceuticals and Medical Biotechnology

URL http://mi20-hlc.doc.ic.ac.uk/
 
Description Discussions leading up to the MI20 HLC were fed into the House Commons Select Committee on Artificial Intelligence and Robotics. In particular, the Select Committee with contributions from a number of those involved in the MI20 meeting, including Profs Muggleton, Bundy and Cohen. At the Committee discussion the views on the importance of Explainable AI were presented by Prof Muggleton. Prof Muggleton has been awarded a £1.7M EPSRC NetworkPlus Grant (2018-2023) based on following up work discussed on the workshop.
Sector Aerospace, Defence and Marine,Creative Economy,Electronics,Environment,Financial Services, and Management Consultancy
Impact Types Cultural,Societal,Economic

 
Description Human-Like Computing initiative
Geographic Reach National 
Policy Influence Type Membership of a guideline committee
Impact Discussions have been held between EPSRC and the advisory committee members of the Human-Like Computing workshop on next steps. These have led to an agreement for submission of a Network Plus project to development of the community in this area. Additionally there has been agreement by EPSRC to fund a £1-2M call in this area in September 2017.
URL https://www.parliament.uk/business/committees/committees-a-z/commons-select/science-and-technology-c...
 
Description Logical Vision partnership 
Organisation University of New South Wales
Department School of Computer Science and Engineering
Country Australia 
Sector Academic/University 
PI Contribution This partnership involves looking at logic-based forms of machine learning for Robocup agents. Prof Sammut's group is a world leader in Robocup competitions and we have been exploring the application of our Logical Vision technology with his group.
Collaborator Contribution Prof Sammut has provided both data and expertise for training recognition of positions of the ball, players and field markings. Prof Sammut's group has also provided financial support for two way visits between the UK and Australia to discuss the ongoing project. He is in the process of applying for Australian Research Council funding to further support this work.
Impact There is a paper being development for submission to the Machine Learning journal.
Start Year 2016
 
Description Nanjing-Imperial Machine Learning Centre 
Organisation Nanjing University (NJU)
Country China 
Sector Academic/University 
PI Contribution This is the outcome of two years collaboration with the University of Nanjing, and has involved multiple bilateral visits. Joint research on the development of techniques for integrating Statistical and Logical Machine Learning has led to an early report on a new technology called Logical Vision. Major conference and Journal submissions are in progress. Our contribution has been in providing expertise and research in Logic-Based Machine Learning.
Collaborator Contribution Nanjing University is China's top centre for research in Statistical Machine Learning. They have developed the code base for the LogVis system which was recently released on GitHub under a BSD open source license. Nanjing University has just agreed to fund the new centre at a level of £60K per year. This will provide funding for travel and RA time to support the ongoing collaboration.
Impact W-Z Dai, S.H. Muggleton, and Z-H Zhou. Logical Vision: Meta-interpretive learning for simple geometrical concepts. In Late Breaking Paper Proceedings of the 25th International Conference on Inductive Logic Programming, pages 1-16. CEUR, 2015.
Start Year 2015
 
Description Submission to House of Commons Select Committee on Artificial Intelligence 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Policymakers/politicians
Results and Impact This involved development of contribution to a report on Artificial Intelligence to the House Commons Select Committee on this topic during 2016.
The submission was made by Prof Tony Cohn (University of Leeds) with contributions from various other leading groups in Artificial Intelligence.
Prof Cohn asked me to present the report to the Committee at the Houses of Parliament, which I did in May 2016.
Year(s) Of Engagement Activity 2016
URL https://www.parliament.uk/business/committees/committees-a-z/commons-select/science-and-technology-c...