AI Alignment Landscape with Thomas Larsen (Center for AI Policy)
Show Notes
In this episode, Soroush explores crucial technical research directions in AI alignment with Thomas Larsen, the Director for Strategy at the Center for AI Policy in Washington, DC. Thomas, who dedicated approximately 75 hours in 2022 to compile an extensive overview of the technical alignment landscape, provides listeners with an updated snapshot of the diverse research directions within the field.
The episode touches on a long list of research areas including: model splintering, out-of-distribution (OOD) detection, low impact measures, threat modeling, scaling laws, brain-like AI safety, adversarial training and brain-machine interfaces (Neuralink).
Episode 8: Getting started in AI safety & alignment with Jamie Bernardi (AI Safety Lead, BlueDot Impact)
Show Notes
In this episode, Jamie Bernardi – Co-Founder & AI Safety Lead at BlueDot Impact – joins the program. BlueDot offers top-tier AI safety courses in collaboration with experts like Richard Ngo from OpenAI and Professor David Kreuger from the University of Cambridge.
Jamie discusses AI safety’s significance, its historical context, current challenges and ways for listeners to get involved in shaping a safe and positive future with advanced AI and AGI.
Episode 7: Responding to a world with AGI - Richard Dazeley
Show Notes
In this episode, Soroush interviews Professor Richard Dazeley – Deputy Head of School at the School of Information Technology at Deakin University in Melbourne, Australia. Professor Dazeley shares his vision of what AGI could look like through a thought-provoking exploration into the potential future and capabilities of the technology.
The discussion delves into the practical aspects of AGI and examines the technical and governance measures that need to be implemented – both now and into the future – to ensure that AGI serves us positively and safely.
Broaden your horizons in this episode by learning more about the challenges and opportunities we face in this AI-driven era.