|Information Society Technologies - Future and Emerging Technologies
IST-FET Workshop for Preparing for FP7 'New Directions for ICTs in FP7:
Grand Challenges for Basic Research'
Brussels, 21-22 April 2004
In preparation of the 7th Framework Programme, the Future and Emerging Technologies Unit of the Directorate General Information Society has organised a 2-day workshop, on the 21st and 22nd of April in Brussels. Due to the importance of this event, ERCIM News publishes the full executive summary of the report.
The aim of this event was to discuss the new grand challenges that lie ahead of us and to identify future key and visionary basic research directions in Information and Communication Technologies (ICTs), as well as in their combination with other disciplines for addressing the new challenges. The event was targeted to a wide spectrum of leading scientists, industrialists and science policy makers and was attended by more than 120 participants.
The workshop comprised a set of plenary sessions and four parallel thematic sessions, covering issues on 'Strategy and policy for ICT-based collaborative basic research', 'Components Research', 'Systems Research', and, 'Intelligence and Cognition'. During the event, the participants exchanged and shared their knowledge, experience and visions and drew a set of recommendations and conclusions on all the above issues. These are highlighted in the next sections.
Thematic Session on Strategy and Policy for ICT-Based Collaborative Basic Research in FP7
The high added value and importance of collaborative basic research in ICTs are now widely recognised in Europe. Defining and promoting future basic research directions for collaborative basic research in ICTs at EU scale, developing strategies for fostering research excellence beyond national boundaries, motivating and enabling effective university-industry collaboration and well integrating the new member states into the EU's basic research and innovation fabric are some of the new challenges lying ahead of us in the coming FP7. These are the main issues discussed in this session. Its findings are summarised below.
Concerning the definition of future basic research directions, it was emphasised that the boundaries of ICT research are now further expanding and ICTs are increasingly cross-fertilising with many other science and technology fields. In this framework, three mechanisms were discussed for selecting future basic research directions.
First, through identification of grand challenges, ie, visionary themes demanding breakthroughs in basic research and engineering in many key technologies focusing 10+ years in the future. Grand challenges should lead to 'pictures of the future' and lie at the edge of what 'just might be possible', so as to inspire researchers beyond the boundaries of ordinary thinking. Second, through identification of key technological issues of major importance to economic growth. And third, derived from major social and societal drivers.
Irrespective of the mechanism used, a basic research direction selected would have to clearly address the following four major aspects: what is the issue to be addressed, how can it be best addressed, for whom is it addressed, and by whom.
For the development of research directions it was also found beneficial to foresee a 'planning phase' during which relevant stakeholders are brought together to build a consensus on the primary motivation and basis for a new research programme. Multi-disciplinary representation in this planning phase is a key issue but multi-disciplinarity does not necessarily imply that in a research programme there must be equal participation across all relevant disciplines concerned.
Research excellence beyond national boundaries is often a prerequisite to ensure excellence. New programmes must be designed to attract the best researchers in each of the disciplines concerned and provide methods to train new generations of researchers but also researchers in industry. Education and training mechanisms have to become an integral part of future EU programmes.
Excellent research will often be characterised by co-sponsoring from national funding agencies. Such co-sponsorship is highly desirable and mechanisms for achieving it need to be put in place. Co-sponsored programmes need to remain open to participation from all member states and not limited only to the states whose agencies are co-financing them.
Aiming for excellence in new research programmes calls for a definition of excellence. While a number of methods now exist on how to measure excellence, individual projects need also to define metrics through which their excellence can be measured. Metrics of progress should also be defined at the level of a new programme, from its beginning.
Stronger emphasis needs to be placed on the dissemination and promotion of results achieved in research programmes. Multifaceted mechanisms need to be put in place for going beyond standard academic publications and conferences by including measures for dissemination of results to industry to ensure take-up and to society at large.
In terms of academia/industry collaboration: strong industrial involvement in new research programmes is aimed. This involvement may take many different forms, from project observers to 'matching of research funds'. Explorative research should primarily be carried out in an open environment. At the same time, industrial involvement can leverage the research directions and the later take-up of results. However, the rules for participation and the associated rules for IPR ownership must be clear from the start and at the level of a new programme.
The ten new member countries have a long tradition of excellent research. There is however a strong need there to consider initiatives for added infrastructure support (eg structural funds, new equipment through research projects, etc) and methods to get researchers more tightly integrated into the EU research community, in particular by better exploiting the existing networks of excellence and co-ordination actions.
Thematic Session on Components Research
In the coming decade, application driven research in components will be guided by ambient intelligence applications focusing on health, comfort and leisure, communication, mobility, safety and security. Opportunity-driven innovations will be increasingly based on novel nanodevice building blocks and related nanofabrication techniques, on new component design and architectures, and likely on bio-inspired approaches and concepts. Such developments will complement the mainstream RTD efforts outlined in the ITRS roadmap and were the subject of discussions held in this thematic session.
With nanoscale devices reaching characteristic dimensions of 10 nanometres (nm) in 2015-2020, new opportunities will emerge to combine ultimate 'top-down' semiconductor platforms with 'bottom-up' developments in materials, physics, chemistry and biology. A set of grand challenges was identified that summarises the multidisciplinary research themes required for bringing these visions into tomorrow's innovations.
Advances in nano-scale materials science will provide the basis to add significant extra functionality and performance to devices and systems on mainstream silicon platforms. These in turn will bring along and require new system architectures that will underpin new applications in ICTs and the life sciences. Two main issues demand particular attention, namely heat generation and dissipation as dimensions shrink, and architectures. Inorganic nanowires and nanotubes will likely have a prominent position in these developments.
The second grand challenge is to enable the combination and interfacing among a growing diversity of materials, functions, devices and information carriers. These 'information carriers' could be electrons, photons, spins, ions, etc.
New materials, devices and circuits will require cost-effective fabrication techniques for complex systems with deep nanometre scale devices. Nanoscale components must be grown and patterned at scales around and below 10 nm, going far beyond the current limitations of lithography. Self-assembly of nano-objects mediated by (bio)chemical interactions appears as one of the promising routes for a sustainable manufacturing of downscaled nano-components.
Pushing the limits of miniaturisation to the one nanometre scale requires new methods and tools to accurately model, manipulate, fabricate and characterise nano-objects down to the atomic scale. It also requires new paradigms to exchange information with single atoms or molecules.
Denser integration and combination of top-down, bottom up and self-organised devices will vastly increase the complexity of ICT components and architectures. These require methods and tools to master giga-complexity ICT architectures, integrating billions of devices with nano-scale dimensions and coping with variability, defects and energy-dissipation issues. Inspiration from bio-systems is likely to lead towards innovative and lower cost solutions.
A number of new physical phenomena or properties of matter at the meso-scale have recently been discovered or demonstrated. These should be further investigated and, as appropriate, developed into new functions or technological developments for the ICT. Research for the discovery and further investigation of such new phenomena needs also to be supported.
The increasingly fading boundaries between ICT and other related fields such as materials sciences, physics, chemistry, biochemistry and life sciences were stressed all along the discussions held in this session. Future research is thus expected to become more multidisciplinary and be based on strong and effective integration of excellent researchers coming from all these different disciplines.
Thematic Session on Systems Research
Large scale systems like communication networks, large databases and software systems, the Internet, large distributed control systems, businesses and the global economy, are examples of huge, interdependent open information-processing systems with behaviour that is increasingly difficult to predict and control. Modelling, simulation, design and control of such large scale systems in technology, business and the sciences are major issues to address in the coming years and were the subject of discussions held in this thematic session.
Present and prospective developments were discussed as a basis for understanding how to address the challenges of building systems that are robust, resilient, dependable and secure, exhibit multi-purpose functionality, and guarantee operation in mission critical tasks. It was recognised that interdisciplinary research drawing on results from complex systems research is now essential if we are to establish new paradigms within which to address the challenges ahead. Five research directions were identified.
The first research direction addresses the need to infer system models even when only inconsistent and incomplete information is available about their functioning and interactions. We need to develop techniques for inferring the dynamics of complex systems, the laws governing their interaction, and ways to describe their behaviour, in order to simulate many systems for which there is at present insufficient direct knowledge. Such systems occur especially in ecology, medicine, molecular biology, certain technological-information processes such as the internet, and systems in management, finance, and economics whose behaviour is very dependent on the human in the loop.
The second research direction addresses the design of human responsive ICT systems that integrate well with humans and adapt well to human needs. It was recognised that today we can no longer treat ICT systems as separate from their human users and others affected by them, and that an ICT system and the context in which it is deployed together form a system. Research is required to establish new design principles that accommodate the changing needs and desires of human participants in complex systems, rather than treating them as outside the system, and rather than presuming to know what they need. The research would need to acquire better understanding of human behaviour - especially regarding group man-machine interactions from which generative theories should inform the organisation of architectures that support and sustain participation, including participative co-design of the systems themselves.
Three complementary research directions address the need to ensure that systems built have the properties we demand. All of them insist on the need to underpin the formal description of such systems.
The first of this set of research directions is to develop underpinning 'foundations' for software-intensive systems. The aim is to make a fundamental leap in the scientific basis of software engineering technologies to capture evolution and dynamics, selfish interests of individual entities, various levels of bounded rationality, learning aspects and self-emerging behaviour, in a strict, yet tractable way. Advances required include new algorithmic techniques for distributed systems and property-aware compilation and implementation techniques for non-functional properties such as security, safety, scalability, resource optimisation, quality of service, and efficiency, in order to be able to guarantee these non-functional properties.
The other two research directions in this set, namely 'Control and Integrated Design of Complex Systems' and 'Many level, Many Unit, Many Purpose Systems' are closely related. It was recognised that for systems comprising many interacting elements we currently lack techniques for modelling that enable prediction of the types of structures, the dynamic behaviour, and the properties that are likely to emerge at higher, collective levels of activity. A key characteristic of many such systems is that they involve components with uncertain operation and interactions, which are at least partially unpredictable and often not yet formalised in algorithmic terms. A number of biological, social, management and economic systems exhibit these properties, yet they must continue to function adaptably, malleably and resiliently, in the face of such unpredictability. The ambition is on the one hand to establish engineering guidelines that draw inspiration from complex natural, social, technological and economic systems and on the other to establish an underpinning framework of formal or mathematical techniques. Together, they should enable us to find cost-effective solutions to problems that cannot be solved with current techniques.
Thematic Session on Intelligence and Cognition
Achieving true machine intelligence remains an illusive challenge for the perception, cognition and AI research communities. Panel discussions in this session focussed on new promising research directions for achieving leap progress in this area through the understanding of the processes underpinning intelligence and cognition in living organisms.
The vision that was promoted is the one Toward Natural Cognition, where the goal is to build artificial cognitive systems inspired by biology, in particular neuroscience, under the following two assumptions: Cognition by systems interacting with the real world is depending on and is facilitated by their body. The structure of this body, the environment and the body-environment interaction are inseparable from one another.
Traditional cognitive science, cognitive psychology and AI make no commitment to the form of a cognitive system's implementation. Today, especially in cognitive neuroscience and robotics, the infrastructure (ie embodiment) is considered much more crucial to the understanding of cognition. One obvious difference between IT systems and biological cognition is the extent to which biology is self-programming, has adaptive configuration of sensors and effectors, and has extendable processing able to make analogies and cope with novel percepts.
Toward Natural Cognition is aimed at taking a relatively radical step away from classical AI-based IT approaches to cognition toward research on self-organisation and development as a natural framework for cognition. In this context, cognition is seen as more than just an inferential process. It is a property that results from the interaction of an organism with its environment. Embodiment, as the central notion in this vision, is characterised by a number of attributes, such as: Embodiment is intrinsically developmental and is structured by interaction with the environment. It enables affective interaction, the acquisition of meaning from percepts created through sensory-action integration and the grounding of 'concepts' in the agent's sensory-motor and social interaction (which provides the basis for natural language). It facilitates learning by formation of cross-modal associations through induction and generation of correlations. It includes continuous dynamics with discrete attractor states, provides the basis for grounding and maintains the distinction between the description of cognition by external observers, and its underlying mechanisms.
The underlying research challenges of this vision include:
- the exploration of non-classical computation and the development of robust scalable self-constructing/repairing architectures
- the exploitation of phylogenetic (evolutionary) and ontogenetic (individual) development and the development of cognitive systems with self-regulation and self-maintenance ('homeostasis') properties
- the investigation and subsequent exploitation of emergent properties of large-scale structures (hardware and simulation) for cognitive processes
- the attainment of high-level cognition by bottom-up organisation.
For addressing the above challenges, relevant research topics to investigate include: new forms of biology-inspired analogue processing and related tools for implementing such large-scale biology-inspired systems; body construction or development and underlying materials and morphological issues; exploring 'how much embodiment is sufficient', and the relationship between top-down knowledge, innateness and learning.
Toward Natural Cognition and its underlying challenges would require not only inspiration from biology (in particular, neuroscience) but also exploring the relationship between cognition and architecture. Trans-disciplinarity is also a must, offering a synergistic interaction of IT with neuro- or, more generally, life sciences.