HPC at Imperial
Powered by Imperial Tech Foresight
Three Imperial academic foresighters joined industry experts, policy-makers, entrepreneurs and members of the Imperial Business Programme, senior College staff, researchers and postgraduates for HPC at Imperial, a one day event exploring the applications that will exploit the power of high performance computing in the coming decade and the potential shift from niche uses to broader adoption.
Download the full presentation from the day:
Answers to all questions that were not addressed at the event during Q&A panel discussion:
1. If parallel was the answer to 3GHz clock rate, what will be the answer when we hit the limits of parallel? (Or can parallel go on forever?)
For parallel computing, the real barriers are likely to be algorithmic rather than physics or engineering. Amdahl's law shows us that even if we have perfect and infinite parallelism, the speed of execution is limited by “quantum’s” of work that are inherently serial. Additionally, there are a wide range of problems (such as the traveling salesman problem) that are described in complexity theory as NP-complete, meaning that there is no fast way known to solve the problem and the time required to solve a problem increases very quickly with the problem size - quickly outgrowing the capability of any conceivable computer. Determining whether or not it is possible to solve these problems quickly is one of the big unsolved problems in computer science today.
2. How will companies be able to access the best compute power without having to invest in their own clusters?
That strongly depends on what “best” means. Certain problems require specialised computer hardware for reasons of economy and time to solution and it is critical companies get expert advice specific to their actual requirements. There is no one-size-fits-all solution. Having said that Cloud today has made it possible for small companies to easily access effectively limitless computing resources without the need for upfront capital investment - even Wired has quipped Cloud is supercomputing for the 99%.
3. With accelerated innovation in an HPC world, how do we avoid bias and maintain ethics?
We need to think very carefully about inclusiveness in the digital economy. Open innovation and open access are critical in removing barriers and widening participation in technology development. More needs to be done to foster open innovation and open access by both government and business. Not only is this important for the wellbeing of society, be increasingly we are seeing that successful technology companies only thrive when they are able to truly engage with their user communities.
4. Will building the world's most powerful supercomputer become more of a national flag waving device than having the biggest army, or tallest building?
Undoubtedly some of the largest supercomputers in the world do have strong national agendas. However, back in 2013 Amazon put 26,496 CPU cores together to build a Top 100 supercomputer in one its data centres. So in the same way that the space race shifted from being a race between nations to being one that is now arguably lead by the private sector, we are increasingly seeing more leadership from the private sector in the private sector which is of course highly international.
1. Will network access become a determinant of life expectancy?
If you ask my teenage sons, definitely yes! But seriously, research shows that information on how to prevent and treat disease is affecting behaviour of individuals, both in their choices of prevention and also in their decisions to access care. For example, greater awareness about the symptoms of stroke and heart attacks is responsible for some of the increase in attendances at Accident and Emergency departments; newspaper reports on the swine flu pandemic has led to a temporary, but drastic reduction in people using public transport in Hong Kong. Researchers are now trying to understand why there are some people who are not receptive to information and education, and also, how behaviour can change the prevalence of disease in societies.
2. Emerging Markets Do Policy and Infrastructure collide with Local Governance?
This is an interesting question. The answer of course depends to a great extent on the decentralization of decision making in countries, which can vary quite markedly. We see that in countries with weak governance decision making is often quite decentralized, to regional administration. It is also often influenced by external organizations, such as international donor organizations or non-governmental organizations in low-income countries. International donors often take considerable influence on national decisions, for example, by making funding conditional on that certain policies are implemented. We can argue whether that is a good or a bad thing. International organizations may have insufficient consideration for national or cultural characteristics that affect priorities, but on the other hand, they may enforce that policies follow proven evidence on what are robust decisions.
3. Will AI enable me to cheat death? Deep learning the micro-determinants that help me live longer? Cheat death?
Hey, give it a try! There is a lot of information out there that will help you to lead a healthy life, although sometimes it is difficult to stick to it. See my comment above. A comedian said that the problem with extending lives is that we get the extra years tagged on at the end of our lives, when we are old and possibly demented! My (unscientific) recommendation is to enjoy life now.
HPC Team at Imperial:
Head of Dept. of Materials
HPC Systems & Training Specialist
HPC Systems Support Specialist
HPC Systems & Support Analyst