Robot Cognition Workshop Invitation Letter

This was the formal invitation letter sent to all previously agreed participants in the 1998 International Workshop on Robot Cognition.

On behalf of the Future Technologies Group at British Telecom Laboratories, I am pleased to invite you to attend an international workshop on engineering cognitively sophisticated robots.

The workshop, open to selected international and BT invitees from 23 to 25 April 1998, will explore the feasibility of a new research project aimed at engineering an intelligent and possibly self aware machine. We will identify key technological requirements, when such a machine could be available, what types of tasks it might undertake, and what spin-off applications might be expected from building initial prototypes. This project clearly has ambitious goals, similar to those propounded at the start of research into artificial intelligence nearly 50 years ago. This immediately raises several questions: why should BT undertake such a project now, and what do we expect to gain from it? What does BT expect to hear or learn at the workshop, and how does it intend to proceed in the future?

This letter is intended to describe the rationale behind the workshop, its format, what we would like the presenters to cover, and the various ways we foresee the project developing in the future.

Background

The telecommunications industry is undergoing radical change. Developments in optical fibre systems mean that it will soon be possible to move Tbits/s of information down a single fibre and hundreds of Mbits/s into every household. Traditional methods by which telecommunications companies have sold services and earned revenue will vanish: those left simply shuffling bits around will be sidelined as suppliers of a low margin commodity. At the same time, the control software for our networks and systems displays increasingly daunting complexity; this complexity will soon become overwhelming. Apart from the need to reduce costs as rapidly as possible, there are thus two drives within the industry. One is to find innovative ways to deliver new services without being overcome by the systems’ complexity; the other is to identify ‘value added’ services which may entice customers to pay for something other than bandwidth. It is our opinion that both of these will require cognitively sophisticated machines, albeit for different tasks.

The first drive suggests the need for intelligent and automated network management, service control, and workforce allocation. We also must bypass today’s mammoth, monolithic software control systems to deliver new services rapidly and cost effectively and to put in place flexible billing systems personalised to individual customers. The key issues are speed to market, cost, and service quality. The right intelligent, adaptive, and self aware system could progressively meet these needs, whilst learning online and slashing ongoing code writing and maintenance costs. Overall savings could approach 50-70%.

Intelligent services are of central importance to the second drive. We will require not merely the simple re-direction features now labelled as ‘intelligent network services’, but the ability to understand and generate information. Ultimately, intelligent services will include reading, summarising and discussing documents, identifying and predicting people’s needs, and becoming an integral part of the lives of those using the technology. Such systems will be much more than the large operating systems and search engines of today; they will have moved beyond simply processing information to understanding it. Those providing the most intelligent of such machines will sell the highest value added services.

We do not believe that such machines will be created simply by incrementally expanding and tweaking traditional hardware and software systems. Today’s most powerful supercomputers and the most astounding AI programs lack the cognitive sophistication of a six week old puppy, and it is not apparent how merely increasing their raw power might overcome that.

Rather, we believe that the underlying architectures must be radically rethought. This workshop will identify key issues along possible paths to an embodied cognitive architecture whose capabilities are more in line with those of the products of evolution by natural selection – Nature’s own 3.8 billion year R&D project.

Workshop Format and Agenda

The workshop has been designed to give leading international contributors sufficient time to explain their ideas, whilst allowing the audience plenty of time to explore the issues. Thus, speakers have been allocated 45 or 60 minute slots with an additional 15 minutes for questions. Discussion forums may be partially structured in advance, and time keeping will be strictly enforced. Some speakers will be asked to summarise and comment on events during the day as part of their presentation. The workshop was planned by identifying a number of topic areas and relating this to work progressing in various laboratories. As a result, most topics are covered by two or three speakers, and some speakers cover more than one topic. The suggested nature of each presentation is noted in the draft agenda.

We are NOT after talks which only summarise the published work of the presenters. This is available in the open literature. We are much more interested in presentations addressing the following issues from the perspectives of presenters’ own fields:

1. What has been learnt so far about how to build such a machine?
2. What has been learnt about how not to build such a machine?
3. What theory and engineering practise already exists?
4. What key scientific questions must be answered or development issues overcome to allow such a machine to be built?
5. What are the next steps we need to take to address these?
6. What sources of inspiration could be tapped for new ideas?
7. What are the shorter term spin-off benefits likely from these research topics?
8. How much would it cost to start addressing these issues?

The Presenters

A list of presenters is available from the blurb about the workshop.

Future Developments

This workshop will launch a project which we hope will expand with better focus as a result of the workshop. As such this is part of the first year’s work in identifying those institutions and individuals with whom we would like to work, and those topic areas of greatest importance. If it is then agreed that a worthwhile and feasible collaborative project exists, we would approach the identified parties and discuss possible arrangements in detail.

Support

This workshop has been arranged by Dr. Greg Mulhauser with the technical and managerial backing of :

• Professor Chris Winter, Head of Future Systems Group
• Dr. Alan Steventon, Head of Systems Research Unit
• Professor Peter Cochrane, Head of Applied Research and Technology (ART), the research arm of BT Laboratories

It is funded by BT’s Corporate Research Program Office (CRPO).

BT attendees consist of members of ART — particularly researchers in AI, Artificial Life and Novel Hardware Technologies — and the CRPO.

Presentation Arrangements

All technical speakers should allow 15 minutes for questions and discussion. The chairman will enforce time keeping ruthlessly.

Audio-visual equipment will be provided, including projection facilities for Macs and PCs, plus overhead projector, slide projector, and s-VHS video player capable of s-VHS, VHS, PAL and NTSC standards. Please indicate any other special needs.