Centre for the Study of Existential Risk – Written evidence (RSK0063)
Prepared by Shahar Edgerton Avin, Lalitha Sundaram, Jess Whittlestone, Matthijs M. Maas, and Tom Hobson[1] on behalf of The Centre for the Study of Existential Risk at the University of Cambridge, 16 Mill Lane, Cambridge, CB2 1SB
The Centre for the Study of Existential Risk (CSER) is an interdisciplinary research centre within the University of Cambridge dedicated to the study and mitigation of risks that could lead to human extinction or civilisational collapse.
In this submission we:
(1) introduce our approach to defining and classifying extreme risks, including global catastrophic risks and existential risks (addressing question 1: "What do you understand the term ‘extreme risk’ to mean?");
(2) highlight two important cross-cutting themes emerging from our research, which are key to grounding risk assessment and management: (a) the systemic nature and political, technological and environmental context of extreme risks, and (b) the relationship between extreme risk and questions of global justice (addressing questions 2: "Are there types of risks to which the UK is particularly vulnerable or for which it is poorly prepared?" and question 6: "How effectively do current ways of characterising risks support evidence-based policy decisions?"); and
(3) summarise our research findings on key tools for appropriately responding to extreme risks, including tools for foresight (addressing question 5: "How can the Government ensure that it identifies and considers as wide a range of risks as possible?"); tools for crafting intervention policies and ensuring their effective implementation (addressing question 7: "How effectively do Departments mitigate risks?"); tools for resilience (addressing question 10: "What challenges are there in developing resilience capability?"); and tools for governance (which go beyond the questions posed by the committee to consider the UK's role in the global governance required to address extreme risks).
1.1. Global catastrophic risks (GCRs) are those which could lead to significant loss of life or value across the globe, impacting all of humanity. While a clear delineation of the category has yet to emerge in the academic field, key works refer to disasters that inflict a loss of 10% or more of the human population, or (on lower thresholds) to more than 10 million deaths.[1] While such extreme scenarios have not been experienced in living memory[2], they are certainly not unprecedented: for example the Late Pleistocene population bottleneck or the Black Death would be considered global catastrophes. Several scientifically-plausible scenarios have been identified which could lead to such losses today or in the future, including the use of nuclear or biological weapons in warfare, catastrophic climate change, and pandemics.[1][2]
1.2. Existential risks are those which could lead to ‘the premature extinction of earth-originating intelligent life, or the permanent and drastic destruction of its potential for desirable future development’.[3] Unlike global catastrophic risks, existential risk scenarios do not allow for meaningful recovery, and are therefore, by definition, unprecedented in human history. Nonetheless, our knowledge of the extinction of other species, as well as the potential for drastic changes in the long-term trajectory of species, indicate that such risks are plausible and merit significant attention, especially as any response to them must occur before the loss is irreversible.[4][5][6][7][8]
1.3. In what follows, we use ‘extreme risks’ to refer to global catastrophic risks and existential risks. The nature of these extreme risks (complex and unprecedented) make them difficult to assess and address, in comparison to more regularly occurring events such as floods, earthquakes or terrorist attacks. For that reason, we recommend the committee pay special attention to ensuring that any government risk assessment process is also able to handle global catastrophic and existential risks. In the remainder of this submission, we draw on our research to make recommendations for how to do this.
1.4. To understand how extreme risks can arise we need to consider the interaction between many critical systems and threats. We therefore recommend breaking the analysis of extreme risks into three components:
● (i) the critical system (or systems) whose safety boundaries are breached by a potential threat
● (ii) the mechanisms by which this threat might spread globally and affect the majority of the human population, and
● (iii) the manner in which we might fail to prevent or mitigate both (i) and (ii).[9]
1.5. A critical system is one whose ordinary operation plays a crucial role in supporting humanity’s ability to survive in its current form. There are seven levels of critical systems, from physical systems (fundamental physical features of a habitable environment) up to socio-technological systems (such as institutions and technologies). When the safety boundaries of critical systems are breached, this can have cascading effects with the potential to spread disruption. For example, a disease outbreak could lead to a global catastrophe if it threatens some crucial part of the anatomical system resulting in the death of a large percentage of hosts (critical system affected), if it is able to spread to the whole world through transport networks and human contact (global spread mechanism), and if we fail to identify and contain the threat quickly enough (prevention and mitigation failures). Extreme impacts of climate change could arise if greenhouse gas emissions severely affect the critical system of agriculture, are spread by the global weather system, and if these impacts persist long-term due to conflicting interests.
1.6. The pandemic scenario, which used to be mainly the concern of experts and professionals, has, in the very recent past, become an immediate concern for the global population. Other scenarios, such as extreme climatic effects of a nuclear exchange, or extreme impacts of climate change may today seem to some as remote as the pandemic scenario used to be prior to 2020. The possibility of a crisis much more lethal than Covid-19 highlights the importance of preparing for extreme risks. Classifying risks by their critical systems, spread mechanisms and prevention and mitigation failures provides an analytical tool for studying systemic risks without reducing their inherent complexity and helps identify policy levers, and other opportunities, for mitigating them.
In our research, two key themes have emerged which we often see neglected: (1) the need to take a systemic approach to identifying and mitigating extreme risks, and (2) the importance of considering the interaction between extreme risks and global justice. We expand on these points and how we recommend the Committee address them below.
2.1.1. When thinking about the extreme risks facing the UK and the world, it is natural to focus on individual hazards which could precipitate a catastrophe, such as asteroids, new technologies, or climate change, and how those hazards can be mitigated. However, this is dangerously simplistic: as the framework in 1.4 shows, the extent to which a specific hazard poses a risk depends on how it interacts with various features of- and systems within our society.[10][11] For example, a major earthquake only becomes a risk if our critical infrastructure is built in such a way that it can be destabilised by earthquakes.
2.1.2. Taking a more systemic approach which explicitly considers these interactions can both (a) help us to identify a wider range of both emergent and structural risks, as well as their drivers, and (b) enable us to find more effective mitigation strategies for reducing risk.[10]
2.1.3. We recommend three specific ways that the Committee can take a more systemic approach to extreme risk assessment and management:
● 2.1.4. Technological risks should be assessed in their social, political and environmental contexts. For example, when considering the risks of new technologies, it is key to consider their interaction effects with surrounding institutional and political factors: are there competitive dynamics or conflicting incentives which make it more likely harmful or unsafe technologies will be deployed?[12][13] Engaging with context also applies to the framing and assessment of ongoing or past disasters. Incidences such as the Fukushima Daichi nuclear accident are usually understood as “natural" disasters, but we should take care to attend to the important social, political, governance and legal lessons that can be learned to increase future resilience.[14]
● 2.1.5. Extreme risks tend to be complex, with a significant potential for indirect harm, which should be assessed. Whether or not a specific hazard ‘poses a risk’ in itself is often less important than understanding its possible impact in a wider context: rather than debating whether or not the various direct impacts of climate change (such as heat stress or sea level rise) are individually extreme enough to pose an ‘extreme risk’, we find it more useful and meaningful to consider how each impact interacts with other factors such as food insecurity, international conflict, or future geoengineering technologies to increase aggregate global risk to an extreme level.[15][16][17]
● 2.1.6. Mitigation strategies work better when they address society’s structural vulnerability to catastrophes. Direct, often technological, solutions to specific hazards may address part of the threat, but do surprisingly little to reduce risk if underlying cross-domain vulnerabilities are not addressed. For instance, one such cross-cutting vulnerability which we have identified is the challenge of reaching informed collective action during times of crisis, which is a present and growing concern for democratic societies that have in recent years seen their ‘epistemic security’ undermined by political polarization and emerging technologies.[18]
2.2.1. Our research has found that extreme risk and global justice are closely related, and that tackling extreme risk requires tackling many questions of global justice.
● 2.2.2. Global injustice is a driver of extreme risk: it is both an exacerbating factor for specific hazards like climate change and global conflict, and a systemic factor driving societal vulnerability and hampering efforts to address risks.
● 2.2.3. Addressing extreme risks requires us to consider moral issues of global justice: how our (in)action today may harm peoples and values unable to demand justice for themselves, including globally disempowered groups; our unrepresented future generations;[19] and our past heritage.[20]
● 2.2.4. Global justice raises important questions about (risk) distribution and the demands we make, or the sacrifices we expect, of different parties in confronting shared extreme challenges. For example, who should shoulder the burden of preventing climate harm, when not all parties are subjected to the same risks, and not all parties contribute equally to the hazard?[21]
● 2.2.5. In our work, we have proposed various concrete policies for addressing global justice concerns around extreme risk, including the formation of an All-party Parliamentary Group on Future Generations, and the inclusion of obligations to consider the long-term risks or impacts of governmental policies.[22] We also emphasize the importance of promoting dialogues on extreme risks across diverse groups who may represent or emphasize different conceptions of justice or ethics in confronting existential risks.[23]
● 2.2.6. We also particularly recommend that the Committee consider:
○ How rising inequality and power differentials might fatally undercut programs and efforts to mitigate extreme risks;
○ How national or global policies to mitigate extreme risks can be developed in a way that is legitimate and able to elicit meaningful and authentic support from across society and diverse stakeholders.
Finally, we recommend the Committee explores and considers the following toolbox of institutional, analytical, and policy instruments for appropriately identifying and responding to these extreme risks.
Identification and assessment of extreme risks is a core activity at CSER; we have gained a better understanding of the challenges involved in foresight for extreme risks,[24][25] and have developed a range of methods to overcome these:
● 3.1.1. For critical systems that are essential for survival yet subject to constantly-evolving transformations and threats (such as critical ecosystems), we adopt a horizon scanning method based on the investigate, discuss, estimate, aggregate (IDEA) protocol.[26][27][28]
● 3.1.2. To help direct research activities towards the most pressing topics, we use modified expert elicitation to identify specific questions that are of sufficient breadth and importance to set field-wide research agendas, as for biosecurity in the UK.[29]
● 3.1.3. For exploration of near-term developments in technological domains, such as biotechnology or misuse of artificial intelligence, we use regular expert elicitation exercises which emphasise a diversity of experts, and incorporate a "red team" approach to increase the range and creativity of scenarios considered.[30][31][32]
● 3.1.4. To assist in the exploration of longer-term technological developments we combine theoretical analysis and survey work to identify key themes and milestones that can structure future foresight exercises.[33][34]
● 3.1.5. To keep track of the fast-expanding and intrinsically interdisciplinary literature on global catastrophic and existential risks, we have developed a scientific literature crawling system which combines crowdsourcing and machine learning elements, to identify and curate potentially relevant scientific work as soon as it is published.[35]
3.1.6. A prior publication sponsored by our centre, and informed by in-depth discussions with UK government departments engaged in risk assessment and foresight, summarises many of our foresight findings as they relate to current government processes.[36] We therefore recommend that the Committee expand capacity for regular, transparent and formal foresight and horizon-scanning activities, both within risk areas, but more importantly across risk areas.
In order to effectively prevent or mitigate a broad spectrum of extreme risks, there is a need to develop robust, evidence-based and innovative policies to enhance resilience, reduce vulnerabilities and to directly address the underlying drivers of extreme risks. CSER has developed a number of policy recommendations that should inform the UK’s policy positions on such risks.
● 3.2.1. The UK should develop policies both to address specific risks and to reduce systemic vulnerabilities. Risk-specific policies should include those that address underlying drivers of extreme risk, whether environmental (e.g. climate change), technological (e.g. artificial intelligence or biotechnology) or social (e.g. inequality or polarization).
● 3.2.2. CSER has developed policy recommendations that address a range of specific risks, including biotechnology and the life sciences,[30][37] artificial intelligence and machine learning,[38][39] climate change and diplomacy,[40] disaster response,[41] finance and responsible investment,[42] and defence procurement processes.[43] These recommendations advance a forward looking approach to understanding and preparing for extreme risk.
● 3.2.3. As many extreme risks are global in nature, national risk mitigation efforts should also include pursuing international agreements and action, which the UK is in a strong position to do. The UK has led on a number of crucial risk specific international agreements and policies, particularly in the field of arms control, through the Biological Weapons Convention.[44] and also in relation to emerging technologies and nuclear weapons. We recommend that the UK has the opportunity to take an early lead on pursuing multilateral prohibitions and arms control agreements on military AI applications, especially in relation to the intersection of emerging technologies and nuclear forces.[12]
Even when based on sound foresight, risk assessments and preparedness plans do not save lives unless their recommendations and provisions are implemented well. Implementation, however, is not straightforward, and often clashes with existing cultures and incentives within the organisations responsible. Our work has highlighted challenges to the implementation of policies that target extreme risks in a range of domains. While the responsibility for risk prevention and mitigation extend far beyond government, the government still has a significant role to play in education, communication, standard setting, incentive shaping and oversight. Some specific ways the government could improve implementation of risk management include:
● 3.3.1. Integration of risk assessment into the earliest stages of developing and procuring novel technologies, especially for safety-critical or defence-related systems, and
● 3.3.2. Ensuring throughout-lifetime accountability for high-technology systems, particularly those used by the military.[45]
● 3.3.3. Informing and instilling a culture of systemic risk awareness amongst ‘universal owners’, the class of institutional investors that by their nature cannot stock-pick their way out of a crisis, thus aligning significant financial interest and resources with broad risk management priorities.[46]
● 3.3.4. Investing in regulation, auditing and support for academic research to build an ecosystem that is able to hold developers of emerging technologies accountable, thus creating an environment where user trust can be placed in trustworthy actors.[38]
● 3.3.5. Exploring ways in which emerging technologies such as artificial intelligence can support crisis response, while also investing in mechanisms to speed up or pre-prime ethical review processes (‘doing ethics with urgency’) for the rapid rollout of such tools during crises.[47]
The risk classification system presented under 1.4. above also applies to our thinking on the topic of resilience: instead of a hazard-centric approach, we recommend a systems-based approach to resilience that emphasises the advantages of targeting investment, design and preparedness efforts into key systems that sit at the intersection of multiple risks. For example, our work has highlighted the importance of building resilience in the power system,[48] the food system,[49] the systems that support information sharing and collective decision making,[18] and in supporting mental health in times of crisis.[50]
Extreme risks challenge structures, norms and instruments of existing governance at the exact time we need them most. Accordingly, extreme risk mitigation requires us to reassess both national and international governance structures. This involves questions of adequate instrument choice and design, as well as in which institutions, or at what level, risk governance is best formulated, coordinated, and implemented. We therefore recommend the Committee considers:
● 3.5.1. How UK national risk policies can be embedded into a broader multi-level global governance architecture for extreme risks. To do this, it will be important for the UK to better map the existing global governance architecture for different extreme risk areas. This can help better understand these regimes’ maturity, overlaps, and gaps, and identify opportunities to patch, support or strengthen this architecture.[51]
● 3.5.2. How to guarantee sufficient institutional flexibility to ensure governance can evolve as the character (or our understanding) of extreme risks changes. In particular, the UK can play a role in organizing the emerging international ‘regime architecture’ around new technological risks. In doing so, it should take stock of various trade-offs (such as political power; inclusiveness; adaptiveness; brittleness) in considering the merits of centralizing governance in single institutions. Taking such action is especially urgent in areas such as in the governance of AI technology, where the global governance architecture is currently in a window of opportunity to set norms and organize cooperation frameworks.[52]
● 3.5.3. How to work effectively with different and diverse stakeholders on governance for emerging risk areas. UK action can play a key role during the current window of opportunity, to set down the appropriate norms and collaboration frameworks amongst global and national stakeholders, in order to foster cross-cultural cooperation on AI governance,[53] or to promote productive debates amongst different stakeholders in the AI epistemic community, in ways that ensure productive policies around both existing and future AI policy issues.[54][55][56]
References
[1] Global Challenges Foundation. (2017). Global Catastrophic Risks 2017.
[2] Global Challenges Foundation. (2016). Resetting the frame.
[3] Bostrom, N. (2013). Existential risk prevention as global priority. Global Policy, 4(1), 15-31.
[4] Rees, M. (2004). Our final century. Arrow.
[5] Ó hÉigeartaigh, S. (2016). Technological Wild cards: existential risk and a changing humanity.
[6] Currie, A., & Ó hÉigeartaigh, S. (2018). Working together to face humanity’s greatest threats: Introduction to The Future of Research on Catastrophic and Existential Risk.
[7] Häggström, O. and Rhodes, C. (2019). Guest editorial. Foresight, 21(1), 1-3.
[8] Beard, S., & Kaczmarek, P. (2019). On the Wrongness of Human Extinction.
[9] Avin, S., Wintle, B. C., Weitzdörfer, J., Ó hÉigeartaigh, S. S., Sutherland, W. J., & Rees, M. J. (2018). Classifying global catastrophic risks. Futures, 102, 20-26.
[10] Beard, S., & Torres, P. (2020). Identifying and Assessing the Drivers of Global Catastrophic Risk: A Review and Proposal for the Global Challenges Foundation. Global Challenges Foundation.
[11] Liu, H. Y., Lauta, K. C., & Maas, M. M. (2018). Governing Boring Apocalypses: A new typology of existential vulnerabilities and exposures for existential risk research. Futures, 102, 6-19.
[12] Amadae, S. M., & Avin, S. (2019). Autonomy and Machine Learning as Risk Factors at the Interface of Nuclear Weapons, Computers and People. In Boulanin, V. (Ed.). (2019). The Impact of Artificial Intelligence on Strategic Stability and Nuclear Risk: Euro-Atlantic Perspectives.
[13] Currie, A. (2018). Geoengineering tensions. Futures, 102, 78-88.
[14] Hörhager, E., & Weitzdörfer, J. (2018). From natural hazard to man-made disaster. In Amelung, I., Bälz, M., Holbig, H., Schumann, M., & Storz, C. (Eds.). (2018). Protecting the Weak in East Asia: Framing, Mobilisation and Institutionalisation. Routledge.
[15] King, D., Schrag, D., Dadi, Z., Ye, Q., & Ghosh, A. (2017). Climate change: a risk assessment.
[16] Beard, S., Holt, L., Tzachor, A., Kemp, L., Avin, S., Torres, P., & Belfield, H. (2021) Assessing Climate Change’s Contribution to Global Catastrophic Risk. Futures.
[17] Tzachor A. (2020). Famine Dynamics: The Self-undermining Structures of the Global Food System. Global Relations Forum Young Academics Program Analysis Paper Series No.8.
[18] Seger, A., Avin, S., Pearson, G., Briers, M., Ó hÉigeartaigh, S., & Bacon, H. (2020). Tackling threats to informed decision- making in democratic societies. The Alan Turing Institute.
[19] Beard, S. (2019). What Is Unfair about Unequal Brute Luck? An Intergenerational Puzzle. Philosophia, 47(4), 1043-1051.
[20] Kaczmarek, P., & Beard, S. (2020). Human Extinction and Our Obligations to the Past. Utilitas, 32(2), 199-208.
[21] McLaughlin, A. (2020). The limit of climate justice: unfair sacrifice and aggregate harm. Critical Review of International Social and Political Philosophy, 1-22.
[22] Jones, N., O’Brien, M., & Ryan, T. (2018). Representation of future generations in United Kingdom policy-making. Futures, 102, 153-163.
[23] Beard, S. (2020). Climate Justice: Integrating Economics and Philosophy, Ravi Kanbur and Henry Shue (Eds.). Oxford University Press, 2018, 288 pages. Economics and Philosophy, 36(1), 176-182.
[24] Beard, S., Rowe, T., & Fox, J. (2020). An analysis and evaluation of methods currently used to quantify the likelihood of existential hazards. Futures, 115, 102469.
[25] Currie, A. (2019). Existential risk, creativity & well-adapted science. Studies in History and Philosophy of Science Part A, 76, 39-48.
[26] Hanea, A. M., McBride, M. F., Burgman, M. A., Wintle, B. C., Fidler, F., Flander, L., ... & Mascaro, S. (2017). Investigate Discuss Estimate Aggregate for structured expert judgement. International Journal of Forecasting, 33(1), 267-279.
[27] Hemming, V., Burgman, M. A., Hanea, A. M., McBride, M. F., & Wintle, B. C. (2018). A practical guide to structured expert elicitation using the IDEA protocol. Methods in Ecology and Evolution, 9(1), 169-180.
[28] Jucker, T., Wintle, B., Shackelford, G., Bocquillon, P., Geffert, J. L., Kasoar, T., ... & Mukherjee, N. (2018). Ten‐year assessment of the 100 priority questions for global biodiversity conservation. Conservation Biology, 32(6), 1457-1463.
[29] Kemp, L., Aldridge, D. C., Booy, O., Bower, H., Browne, D., Burgmann, M., ... & Sutherland, W. J. (2021). 80 questions for UK biological security. Plos one, 16(1), e0241190.
[30] Wintle, B. C., Boehm, C. R., Rhodes, C., Molloy, J. C., Millett, P., Adam, L., ... & Sutherland, W. J. (2017). Point of View: A transatlantic perspective on 20 emerging issues in biological engineering. Elife, 6, e30247.
[31] Brundage, M., Avin, S., Clark, J., Toner, H., Eckersley, P., Garfinkel, B., ... & Amodei, D. (2018). The malicious use of artificial intelligence: Forecasting, prevention, and mitigation. arXiv preprint arXiv:1802.07228.
[32] Kemp, L., Adam, L., Boehm, C. R., Breitling, R., Casagrande, R., Dando, M., ... & Sutherland, W. J. (2020). Point of View: Bioengineering horizon scan 2020. Elife, 9, e54489.
[33] Martınez-Plumed, F., Loe, B. S., Flach, P., O hEigeartaigh, S., Vold, K., & Hernández-Orallo, J. (2018). The facets of artificial intelligence: a framework to track the evolution of AI. In International Joint Conferences on Artificial Intelligence (pp. 5180-5187).
[34] Cremer, C. Z., & Whittlestone, J. (2020). Canaries in Technology Mines: Warning Signs of Transformative Progress in AI.
[35] Shackelford, G. E., Kemp, L., Rhodes, C., Sundaram, L., ÓhÉigeartaigh, S. S., Beard, S., ... & Sutherland, W. J. (2020). Accumulating evidence using crowdsourcing and machine learning: A living bibliography about existential risk and global catastrophic risk. Futures, 116, 102508.
[36] Centre for the Study of Existential Risk. (2019). Managing global catastrophic risks Part 1: Understand.
[37] Rhodes, C. (2020). Scientific freedom and responsibility in a biosecurity context. In The freedom of scientific research. Manchester University Press.
[38] Brundage, M., Avin, S., Wang, J., Belfield, H., Krueger, G., Hadfield, G., ... & Anderljung, M. (2020). Toward trustworthy AI development: mechanisms for supporting verifiable claims. arXiv preprint arXiv:2004.07213.
[39] Ovadya, A., & Whittlestone, J. (2019). Reducing malicious use of synthetic media research: Considerations and potential release practices for machine learning. arXiv preprint arXiv:1907.11274.
[40] Kemp, L. (2019). Mediation without measures: conflict resolution in climate diplomacy. In Research Handbook on Mediating International Crises. Edward Elgar Publishing.
[41] Weitzdörfer, J., & Beard, S. (2019). Law and Policy Responses to Disaster-Induced Financial Distress. In Governance, Risk and Financial Impact of Mega Disasters (pp. 47-80). Springer, Singapore.
[42] Chambers, D., Dimson, E., & Quigley, E. (2020). To Divest or to Engage? A Case Study of Investor Responses to Climate Activism. The Journal of Investing, 29(2), 10-20.
[43] Belfield, H., Jayanti, A., & Avin, S. (2020). Written Evidence-Defence industrial policy: procurement and prosperity.
[44] Rhodes, C. (2017). Workshop Report – Eighth Review Conference of the Biological Weapons Convention: Where Next?, CSER Workshop Report Series, No.1.a.
[45] Jayanti, A., & Avin, S. (2020). It Takes a Village: The Shared Responsibility of 'Raising' an Autonomous Weapon. Submitted to and Presented at RSIS Military Transformations Programme Workshop.
[46] Quigley, E. (2020). Universal Ownership in Practice: A Practical Positive Investment Framework for Asset Owners. Available at SSRN 3638217.
[47] Tzachor, A., Whittlestone, J., & Sundaram, L. (2020). Artificial intelligence in a crisis needs ethics with urgency. Nature Machine Intelligence, 2(7), 365-366.
[48] EIS Council. (2017). “Black Sky” Infrastructure and Societal Resilience Workshop.
[49] Tzachor, A. (2019). The future of feed: Integrating technologies to decouple feed production from environmental impacts. Industrial Biotechnology, 15(2), 52-62.
[50] Beard, S., Brierton, K., Gilbert, P., & Huppert, F. (2020). Public Mental Health and COVID-19: a compassion based approach to recovery and resilience. Prepared for the Association of Liberal Democrat Engineers and Scientists.
[51] Kemp, L., & Rhodes, C. (2020). The Cartography of Global Catastrophic Governance.
[52] Cihon, P., Maas, M. M., & Kemp, L. (2020). Fragmentation and the Future: Investigating Architectures for International AI Governance. Global Policy, 11(5), 545-556.
[53] ÓhÉigeartaigh, S. S., Whittlestone, J., Liu, Y., Zeng, Y., & Liu, Z. (2020). Overcoming barriers to cross-cultural cooperation in AI ethics and governance. Philosophy & Technology, 33(4), 571-593.
[54] Cave, S., & ÓhÉigeartaigh, S. S. (2019). Bridging near-and long-term concerns about AI. Nature Machine Intelligence, 1(1), 5-6.
[55] Prunkl, C., & Whittlestone, J. (2020, February). Beyond near-and long-term: Towards a clearer account of research priorities in AI ethics and society. In Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (pp. 138-143).
[56] Stix, C., & Maas, M. M. (2020). Bridging the gap: the case for an ‘Incompletely Theorized Agreement’on AI policy. AI and Ethics, 1-11.
28 January 2021
[1] SEA (sa478@cam.ac.uk) led the submission, all other authors contributed equally (order randomized).
[2] At the time of writing, the global death toll of Covid-19 is just over 2 million. While shocking, that is still 0.025% of the world’s population.