AI access inequities, bad philosophical inquiries, and research iniquities

Links 5/14

1. Anton Leicht makes a strong case that we need be worried about differential access to frontier models as an equity issue. This is a strong case for building as many data centers as we can. Although this is counter to much of the anti-AI discourse, this seems right to me. Though given the increasing security risks, it seems obviously true that widespread availability of frontier models is ending.Cut Off — Anton Leicht

2. Kasra with a great piece on why philosophical training can make you easier to fool: skill at constructing and deconstructing arguments doesn’t track truth without introspective attunement to the emotional motivations behind belief. My uncle calls this phenomenon being “too smart to learn.” Easier to fool — Kasra

3. The Scientist surveys eight research-integrity stories from 2025 — I hadn’t even heard of most of these stories, and this is a topic I’m following closely on a regular basis! Inside the Scientific Community’s Research Integrity Crisis — The Scientist