Automating scientific Discovery with Artificial Intelligence Technologies

Luger, E. (Contributor)

Activity: Consultancy typesContribution to the work of national or international committees and working groups

Description

Invitation to EPSRC ‘Automating scientific Discovery with Artificial Intelligence Technologies’ framing workshop (ongoing: shaping roadmap and contributing to subsequent white paper) Background Progress in the advancement and use of AI technologies is happening apace. These developments are now being increasingly turned to the needs of other areas of science. While AI techniques, as in the case of machine learning, have an extensive track record in being used for scientific data analysis, perhaps their most exciting potential impacts are in formulating and testing hypotheses and choosing experiments that are optimally informative. In automating the process of science discovery, these techniques have the potential to irrevocably change the way that science is done. The application of AI methodologies in this way could be envisaged as the next stage of a journey that has seen computers become integral to the advance of experimental science. In particular, this has the potential to transform every area of science and social science that involves experimental measurements or computational modelling. In experimental science, it is could be argued that the early applications will be found where there are costly experiments that will benefit from optimised choices about what to measure next. The first experiments to benefit may be both those that are already fully computer-controlled and whose results are digital, avoiding the need to develop mechanical robots. In theoretical modelling, the first beneficiaries could be envisaged as those in which the choices of what to calculate are constrained by computational resources, such as selection and size of basis sets, precision of calculations, and speed of convergence. The greatest benefits will likely be found in research involving iterative interaction between experiment and theory, since AI methodologies and algorithms offer the prospect of optimising the interplay between them. The UK is in a position to build upon its established expertise in this field and take a leading role in this new area. There is a window of opportunity for the UK to apply its leading and most advanced AI techniques for this purpose, and it is believed there is an appetite in the research community to implement these methods in cutting edge experimental science. Purpose This workshop will bring together those versed in the methodological foundations, and will aim to identify a small initial set of key exemplar areas of science for the deployment of the methods. We will aim to build an understanding of the landscape, engaging key stakeholders including AI and machine learning researchers, research scientists who can consider the ethical, societal and philosophical implications, and industrial end-users. We will consider a list of questions including but not limited to those below. • What are the obstacles to be overcome for the AI techniques? • What are the competing techniques? • Who are the key research communities in the UK and internationally? • Who are the important industrial players in the UK and globally? • Which demonstration applications will be most impactful? • Is the breadth or the depth of demonstration applications more important? • How to ensure that research is conducted in a responsible and ethical manner, and that political/ societal implications of automation are considered and usefully communicated?
Period28 Mar 2017
Event titleAutomating scientific Discovery with Artificial Intelligence Technologies
Event typeWorkshop
LocationBirmingham, United Kingdom