Qut Phd Scholarship: Preferences, Rewards And Value Alignment

The QUT Generative AI Lab (GenAI Lab) is a new, specialist research initiative focused on addressing the emerging social and cultural challenges and possibilities of Generative AI. 

Staffed by a multidisciplinary team led by Distinguished Professor Jean Burgess, the lab aims to develop, disseminate and apply new sociotechnical research capabilities specific to Generative AI, combining critical, technical, and externally-engaged approaches. 

The GenAI Lab is aligned with QUT’s Digital Media Research Centre (DMRC) and the ARC Centre of Excellence for Automated Decision-Making and Society (ADM+S).

This scholarship supports one of a cohort of three PhD students who will commence together in the lab’s first year of operation.

The successful candidate will undertake research on the preference and reward models used to achieve ‘value alignment’ in GenAI systems, using a combination of computational, critical, and qualitative methods. 

The proposed project may cover issues such as:

• the apparent trade-off between helpfulness, honesty, and harmlessness in value alignment
• institutional, geo-political, and algorithmic assumptions and arrangements that underlie and frame contemporary approaches to value alignment such as Reinforcement Learning from Human Feedback (RLHF)
• issues of representation, fairness, observability and interpretability in the design and evaluation of reward and preference models.

The candidate will be supervised by Dr Aaron Snoswell and Distinguished Professor Jean Burgess and will have the opportunity to engage in a dynamic environment with members of the QUT School of Communication and Digital Media Research Centre, as well as the national ADM+S Centre.  

You need to meet the entry requirements for a QUT Doctor of Philosophy, including any English language requirements.

We are seeking to recruit a student to take up the scholarship and begin full-time study in July 2024.

You should have:

• a background in a relevant humanities and social science discipline, such as communication and media studies, STS or information studies, digital sociology, or internet studies
• recently completed a first-class honours degree, a research master degree, or a coursework master degree with a significant research component from a recognised institution and in a cognate discipline
• a strong interest in undertaking a three-year research project on preference and reward models used to achieve 'value alignment' in GenAI systems
• demonstrated excellent capacity and potential for research.

• The first step is to email Dr Aaron Snoswell detailing your academic and research background, your motivation to research in this field and interest in this scholarship and include your CV, full academic transcript and details of three referees (email and contact number) as soon as possible.
• Shortlisted applicants will be invited to an online interview with members of the supervisory team.
• If supported to apply, you will be invited to submit an expression of interest (EOI) following the advice at how to apply for a research degree.
• In your EOI, nominate Dr Aaron Snoswell as your proposed principal supervisor, and copy the link to this scholarship website into question 2 of the financial details section.

Please visit https://www.qut.edu.au/study/fees-and-scholarships/scholarships/preferences-rewards-and-value-alignment-qut-generative-ai-lab-phd-scholarship

The closing date for applications is 31 May 2024.

Dr Aaron Snoswell at **********@qut.edu.au