The Centre for the Study of Existential Risk strongly welcomes the recently announced partnership on AI to benefit people and society (current partners: DeepMind/Google, Facebook, Microsoft, Amazon, IBM). Increasingly powerful AI systems are becoming used in an ever-wider range of real-world settings. This offers wonderful opportunities for helping us with many global challenges – for example, DeepMind have recently developed tools to aid doctors in the NHS, and massively improved the energy efficiency of Google’s servers with a version of DQN (the Atari-beating algorithm), which has very beneficial implications for climate change. Similarly, Microsoft Research are making great progress on applying AI to cancer diagnosis and prevention.
However, the widespread use and further development of these systems will also throw up challenges – including fairness and potential biases in algorithms or the data they generate their results from; our ability to understand how these algorithms function and the settings in which they may not perform as well, and the impact of AI on job markets. In the longer-term, AI is set to be such a transformative technology that it is prudent to think carefully about its safe development, the potential impacts and risks of long-term advances, and the global challenges to which it can be applied beneficially. These challenges will require deep interdisciplinary and cross-sector collaboration between technology research leaders, scholars across disciplines, and policymakers who seek to stay up to date with a rapidly progressing technology. Cambridge is taking a leading role in these discussions; in addition to CSER’s work, research leaders in Cambridge’s machine learning department have been organising workshops at the major machine learning conferences on the societal impacts of AI , legal and policy challenges that AI will present, and the technical design of AI systems so as to be reliable ‘in the wild’. Cambridge has also recently partnered with Oxford, Berkeley and Imperial on a new centre to study the long-term opportunities and challenges of AI, supported by the Leverhulme Foundation – the Centre for the Future of Intelligence.
The research leaders in companies such as DeepMind, Facebook, Microsoft, Amazon and IBM are among the best placed to think in a long-term manner about these issues, in their deep understanding of the current state of the art, their unique insights into where the field will be in ten years’ time, and the ways in which their advances will change the world. They also have a unique opportunity to play a guiding role, in collaboration with others. This partnership is a tremendously positive step, and demonstrates laudable responsibility and leadership from the companies involved. We strongly welcome it, and look forward to opportunities to collaborate on many of the research issues the Partnership highlights.
Seán Ó hÉigeartaigh,
Executive Director, Centre for the Study of Existential Risk