Introduction
Troy Tian
I'm an undergraduate at UCLA studying Computer Science and Asian American Studies. My work and interests lie in AI safety, including technical research, policy/governance, and their intersection.
Selected Writing
A sample of technical, policy, and humanities writing across several domains. View all →
AI Liability Caps: A Price-Anderson Analysis
Should frontier AI deployment be governed by a liability cap? A structured policy argument with a tiered insurance recommendation, the strongest objection to it, and the empirical evidence that would reverse the position.
Three Failure Modes of OpenAI's Preparedness Framework
A technical critique of OpenAI's safety framework — covering misaligned deployment incentives, the limits of capabilities-threshold evaluation, and the dangers of RLHF as the primary alignment mechanism.
Martingale Evaluation of Human–LLM Interaction
A proposed first-step experiment for detecting sycophancy and epistemic drift in LLM interactions using a martingale-based scoring system — framed as a stochastic decision process.
Algorithmic Monocultures in LLM Political Framing
URC-Sciences program research proposal: do LLM alignment techniques systematically degrade epistemic diversity? A research design using adversarial prompts, LIME/SHAP interpretability, and latent variable decomposition.
The Semantic Space Between: American Chinese Cuisine
A senior project paper tracing how the American Chinese food industry has mirrored, shaped, and been shaped by the Chinese American immigrant experience — from the Gold Rush to Panda Express.