How to apply for a SAINTS studentship

Research statement [1000 words MAX] 

During the first year of your PhD, you will be supported in developing a research proposal. Therefore at this stage, we are not asking for a full research proposal. Instead, we would like you to outline your initial research interests and ideas and how these relate to the topic of AI Safety.

In the research statement you should:

• Make it clear which of the SAINTS themes (see below ‘Research Themes’) your research interests align to (you do not need to address both)
• Justify why this research will make a significant and original contribution to multidisciplinary AI safety research. (Where appropriate, references can be included to motivate and contextualise your research ideas).

We believe that AI safety will require multidisciplinary solutions. Your statement should therefore also:

• Explain why the research you are interested in requires a multidisciplinary perspective, and how your research may benefit from engagement with industry.
• Explain how your disciplinary/professional expertise will contribute to the SAINTS multidisciplinary research environment.

After candidates have been accepted into the SAINTS programme, we will identify and assign two or more supervisors who are aligned with your research interests. You therefore do not need to identify a supervisor as part of your application.

Personal statement [500 words MAX] 

Your Personal Statement should outline your academic background and interests, explaining why you wish to pursue a PhD within SAINTS and how this aligns with your broader career goals and aspirations. We encourage you to use this statement to give us a sense of who you are as a researcher and how your experiences have prepared you for doctoral study. You should include:

• Your educational background
• Any relevant professional or voluntary experience
• How you hope to contribute to and engage with the SAINTS research community

Ultimately, the aim is to produce self-driving vehicles that can go anywhere with no human intervention. In the interim, vehicles will have a level of autonomy but there will be a combination of human and AI-based control. Research in this context will include: demonstrating safety of AI-based functionality, in dynamic and uncertain contexts; safety of transitions of control between humans and AI, especially in emergency situations; the legal liability of drivers vs vehicle manufacturers; the ethics of placing responsibility on drivers to monitor systems (something that humans are poor at); ensuring safety and ethical acceptability as the contexts (including regulations and human perceptions of risk) change. 

These include AI-enabled navigation systems for commercial ships and robots that explore ocean depths where humans can’t easily reach, for example to monitor infrastructure such as pipelines and offshore power generators such as wind turbines. Research in this area can explore how these robots can autonomously navigate, monitor and manage their own “health” (they may be on missions for months), collect data or samples without harming marine life or doing environmental damage. Other issues include operating in multiple legal jurisdictions, cultures and societies; the ability to manage and recover the robots following a failure (noting the rules of salvage); the ability to remotely monitor and control the robots. 

Drones equipped with AI are used for surveillance in relatively open areas like borders, powerlines or forests as well as in urban environments which are much more densely populated. Both urban and rural environments are dynamic and have uncertainties, e.g., movement of trees and pylons in high winds, putting up cranes and scaffolding, etc. Safety research focuses on ensuring these drones can navigate autonomously without colliding with obstacles, collect data in privacy respecting ways, and make real-time decisions in response to changing environmental conditions. Other issues include weighing the benefits, e.g., identifying powerline problems to enable rapid repair, versus the risks, e.g., the impact with objects; emergency management including recovery of a drone and data it may have captured; risks of normalisation of mass surveillance or social rejection of operations, especially in urban environments. 

AI is increasingly integrated into military operations for decision support and surveillance. Research is crucial to ensure that AI tools enhance the safety and effectiveness of personnel without leading to unintended consequences. Issues include disaster relief, emergency planning, accessible rescue services, designing to ensure compliance with legal and ethical rules; achieving effective and responsible operation; achieving good situational awareness for operators; understanding liability in the event of accidents; and understanding the ability of operators to monitor more than one system at once. 

In factories, AI systems and robots work alongside human workers. Safety research is crucial to prevent accidents and to ensure efficient and effective collaboration. Issues include establishing confidence in cobots motion planning; balancing the benefits of increased efficiency with the impact on jobs (loss thereof or reduction in quality/value); understanding the change in the nature of jobs on safety and operator mental health; liability in the event of accidents.

AI systems assist in disease diagnosis, analysing medical images and recommending treatment plans. Research in this area includes designing so as to maximise the effectiveness of the human-AI team, considering it as a joint cognitive system. Issues include establishing trust in interactions and rebuilding trust following adverse events; understanding and managing moral and legal responsibility, especially following an incident; understanding how to use explanations (of AI-based decisions or recommendations) to assure safety and preserve meaningful human control.