Gracie Green - Operations Lead
Hiya, I'm Gracie. I've just graduated in Psychology and Philosophy
at Oxford where I generally focussed on AI, consciousness, and
ethics. Within AI safety, I've generally taken on operations roles
and have worked at OAISI since my first year of university (when we
were still OxAI's SGT) as well as working at the Cambridge AI Safety
Hub and ARENA. I've carried out research into philosophical theories
of alignment, inductive bias, critical thinking, and the
non-identity problem. I'm always around for a chat and very up for
meeting new people, so feel free to reach out if you'd like to get
to know OAISI a little better!
Nick Marsh - Strategic Lead
Hi, I'm Nick! I'm a third-year CS and Philosophy student. I'm
interested in a bunch of areas in AI safety - I've done some
mechanistic interpretability and priorities work previously. I'm a
research assistant in the Human Information Processing lab here in
Oxford, and I run [orchard], a group for people to work on their
passion projects. Feel free to reach out if you want to chat about
AI safety, philosophy, or anything else!
Louis Thomson - Committee Member
Hi, I'm Louis! I'm in my masters year doing Computer Science and
Philosophy at Oxford, and my main interests include AI [control /
safety / ethics], game theory, and cooperative AI. I've previously
worked on the following projects (in order of recency): cooperation
and alignment in multi-agent games (current project); AI-Control
games; behavioural evaluations of deception in LLMs; and active
learning in LLMs.
I'm always up for talking about all things CS, philosophy (particularly AI-relevant topics but also continental philosophy), and other entirely-not-work-related stuff like music and games!
I'm always up for talking about all things CS, philosophy (particularly AI-relevant topics but also continental philosophy), and other entirely-not-work-related stuff like music and games!
Rohan Selva-Radov - Committee Member
I'm Rohan, a second-year undergraduate studying PPE at Merton.
In the past I've worked with Daniel Kokotajlo's team developing concrete scenarios about what AI developments over the next five years might mean for the world, and currently I'm spending some time thinking about how we can reduce the risks of malevolent actors misusing future generations of frontier AI models.
I'd be excited to chat about what policy & governance levers seem most promising, and to get a better understanding of the theories of victory are from different approaches in technical alignment.
In the past I've worked with Daniel Kokotajlo's team developing concrete scenarios about what AI developments over the next five years might mean for the world, and currently I'm spending some time thinking about how we can reduce the risks of malevolent actors misusing future generations of frontier AI models.
I'd be excited to chat about what policy & governance levers seem most promising, and to get a better understanding of the theories of victory are from different approaches in technical alignment.
James Lester - Committee Member
Hi, I’m James, a first-year MPhil in Economics at Balliol.
I did my undergrad, also in Economics, at the other place, where I found out about AI safety and catastrophic risk through the uni’s effective altruism community. Through the ERA:AI fellowship, I pursued a project on “risk compensation in AI deployment” – the possibility that safety-promoting measures might have limited effect if they incentivise greater or more reckless deployment of new models. Aside from that, I’ve done short projects on modelling AI’s effect on the labour market for software engineers and considering the possibility for “races to the bottom” between AI evals orgs, drawing on the auditing literature.
At OAISI, I help run programmes which introduce people to these ideas. Please reach out to chat about anything that takes your fancy! />
I did my undergrad, also in Economics, at the other place, where I found out about AI safety and catastrophic risk through the uni’s effective altruism community. Through the ERA:AI fellowship, I pursued a project on “risk compensation in AI deployment” – the possibility that safety-promoting measures might have limited effect if they incentivise greater or more reckless deployment of new models. Aside from that, I’ve done short projects on modelling AI’s effect on the labour market for software engineers and considering the possibility for “races to the bottom” between AI evals orgs, drawing on the auditing literature.
At OAISI, I help run programmes which introduce people to these ideas. Please reach out to chat about anything that takes your fancy! />