Note: Due to COVID-19 the BoF program was significantly reduced and the BoF on Ethics in HPC did not take place.
Ethical concerns in science can be traced back for centuries. With acceleration of innovation and an increasing impact of HPC on daily life, however, it is more important than ever to have these discussions. The goal of this BoF is to foster this community-wide discussion on ethics in HPC on issues such as what responsibilities we have as practitioners in the field and how best to ensure that ethical concerns are considered in the development and implementation of HPC applications. The BoF deliberately does not restrict or proscribe particular areas of discussion. However, the following provides suggestions for possible topics.
With the COVID-19 pandemic, HPC found itself in the spotlight of no less than two prominent controversies: The potentially exclusive access to HPC supported insights for treatments or vaccines. The other is related to the implementation of disease surveillance as well as tools to inform shelter in place orders, often conflicting with existing legislation protecting data privacy or freedom of speech. The spectrum of compromises found by different organizations and nations highlights how sensitive to cultural context ethical judgments are. Consider, for example, attempts to find decentralized solutions or the moratorium on the use of facial recognition technology for law enforcement by various cloud providers.
HPC as a widely applied computational tool is ethically neutral, although some applications rest in gray areas, such as oil and gas exploration, maintaining nuclear weapons, mining personal information for targeted ads or mass surveillance. In addition, HPC could be leveraged for clearly unethical ends such as designing an optimally destructive pathogen or engineering an effective disinformation campaign. As a technology that can be abused, we must consider what, if anything, the community should do to ward against this.
HPC may intensify economic disparity as illustrated by the example that barely 1% of Top500 supercomputing systems, reside in South America and Africa, despite being home to approximately 20% of the world’s population. Just as nations use HPC to increase national productivity and protect their citizen’s welfare, businesses use HPC to manifest a competitive advantage.
We should also consider whether HPC is perpetuating inequality. Women and people of color remain greatly underrepresented also within HPC communities, perhaps this lack of diversity is an inhibitor to the field.
The environmental impact of HPC is another matter of importance, considering that the Top 4 supercomputers already consume 60MW in electricity, the equivalent of 50,000 American households. This number, represents only a small fraction of the environmental footprint of HPC, as is particularly important relative to climate change.
With many workloads incorporating machine learning directly affecting people’s lives, participants may wish to address some of the ethical issues raised by HPC enabled ML/AI. This includes the impact a new generation of automation capabilities might impose on labor forces and society, the risks of amplifying bias, or the application in killer robots, as well as more speculative topics such as the implications of superintelligence.
This BoF hopes to discuss issues such as these, and to consider what values we as a community should be promoting and how best to do so.
Location & Time
SC20, Dallas, USA
Margaret Lawson, (Sandia National Laboratories, University of Illinois)
Jakob Luettgau, (German Climate Computing Center, University of Hamburg)
Jay Lofstead (Sandia National Laboratories)
Randy Rannow (IEEE, Silverdraft Supercomputing)