Christina Lu

Christina Lu is a member of technical staff working on interpretability and a doctoral student in computer science at Oxford. Her research examines how language models represent personas and how those representations shape behavior; she also develops speculative projects about AI and ecological data, including the collaborative Whole Earth Codec. Previously, she was a software engineer at Google DeepMind and an affiliate researcher at Antikythera.

Career

Cosigned byLoading...

CosignedSign in to view

Dartmouth College

BA in Computer Science and Studio Art

Products & media

matsprogram.org

The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models

A research paper examining how language models represent default assistant personas and how shifts in that representation relate to persona drift.

Writing

opensystems.sg

Whole Earth Codec

An interactive art-and-research project proposing a multimodal foundation model that represents ecological data at planetary scale.

medialab-matadero.es

Synthetic Minds Podcast #4: Toy Worlds

Christina Lu discusses interoperable simulated worlds, embodied AI, and how humans and AI might learn to coexist.

Media appearance

thegoodrobot.co.uk

The Good Robot LIVE! From Berlin

Lu discusses sociotechnical research and critical approaches to technology in an interview recorded at the AI Anarchies conference.

Event

cs.ox.ac.uk

Subverting Machines, Fluctuating Identities: Re-learning Human Categorization

A co-authored paper critiquing static identity categories in machine learning and proposing relational approaches to identity.

Writing

Profile name

Company name

Company name

Profile name

Company name

Company name

Profile name

Company name

Company name

Profile name

Company name

Company name

Profile name

Company name

Company name

Christina Lu

Dartmouth College

Products & media

The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models

Whole Earth Codec

Synthetic Minds Podcast #4: Toy Worlds

The Good Robot LIVE! From Berlin

Subverting Machines, Fluctuating Identities: Re-learning Human Categorization

Curated Lists

List Inclusions