Applied indicators for consciousness & pain in AI
- commitment
- 10 hrs/week
- flexibility
- No mentee proposal
mentored by
Riley HarrisUniversity of Oxford
The project
Work with the lead to develop techniques for detecting consciousness indicators in AI systems. Ideally, the team would consist of a few people with different profiles, all of the following skills would be useful:
- Mechanistic interpretability (ability to carry out or replicate mechanistic interpretability papers)
- Software engineering skills (ability to write complex code without AI assistance)
- Strong analytical skills (e.g. significant training in philosophy or a related discipline)
- Strong writing skills
- Neuroscience of consciousness background or related consciousness background
- Ability to work on messy empirical questions without clear answers
The mentee's role
- Designing new software tools with clear guidelines
- Mechanistic analysis outside of standard techniques
- Applying neuroscience and philosophy research
- Writing and analysis
- There may also be scope for a particularly motivated candidate to work on a philosophical paper concerning pain in AI
Who I'm looking for
Nice-to-haves:
- Strong technical skills
- Background understanding of AI consciousness, animal consciousness, mechanistic interpretability
- Strong writing and analytic skills
- Ability to carry out research projects independently
I will be happy to receive applications from motivated candidates that do not satisfy these criteria.
Questions for applicants
- Why do you want to work on this project?
- What achievement or project of yours are you most proud of and why?
- Which skills and abilities would most help you make progress on the project?
Support offered
- Shaping direction
- Drafting
- Conceptual clarity
- Knowledge of literature
- format
- Research & writing
- topic
- Artificial sentience
Riley Harris
University of Oxford
Third year PhD student in philosophy at Oxford with interests in AI consciousness, AI pain, and moral uncertainty.