Geoffrey Irving: ML lacks obstacle culture compared to theoretical CS
geoffreyirving · x · 2026-08-25
Geoffrey Irving notes that machine learning has a much weaker culture around obstacles and negative results compared to theoretical computer science. This is partly because it is too easy to get negative results in ML for mundane reasons, such as insufficient hyperparameter tuning. He suggests that understanding the learning path in time may be impossible, and expresses optimism about theoretical computer scientists joining AI alignment due to the field's rich "obstacle culture" of theorems showing formal blocks to specific algorithms or proofs.
Related event: Irving: Alignment Research Needs TCS Barrier Culture(3 posts)→
More from Research
- U-Mich Kozik Lab case study: SeqHub untangles the understudied respiratory microbiome — anshulkundaje · 2026-10-16
- Arc Institute's Patrick Hsu on CRISPR, AI-designed organisms and betting on people — thesephist · 2026-10-16
- SHOES: 10 minutes of real-robot RL lets quadrupeds walk in shoes, CoRL 2026 paper — kscottz · 2026-10-16
- Post-trained NVIDIA Nemotron-3.5-Super-VL hits 35% on AutoResearchExam, matching a 20× larger model — AlexGDimakis · 2026-10-16
- U of T team mines 122,194 patient reviews across 45 Ontario hospitals with NLP to expose satisfaction and equity patterns — ishtiaqueSIA · 2026-10-16
- University of Toronto hiring AI postdoc: builders of interventions, not just benchmarkers, due Oct 30 — ishtiaqueSIA · 2026-10-16