Uploaded April 2025 | Updated September 2026, 2 weeks ago
How causal representation learning is fixing AI’s vision blind spots! In this video, we explore real-world examples of causal ML in computer vision, like phenotypic imaging at Recursion’s Valence Lab, where scientists train models on cell images before/after gene knockouts to isolate biological causes from noise. We also break down the classic “wolves vs. dogs in snow” problem, why traditional CV models fixate on spurious correlations (like backgrounds), and how causal representations could force AI to focus on actual causal features.
#CausalVision #RecursionAI #MedicalAI #ComputerVision #MachineLearning
How causal representation learning is fixing AI’s vision blind spots! In this video, we explore real-world examples of causal ML in computer vision, like phenotypic imaging at Recursion’s Valence Lab, where scientists train models on cell images before/after gene knockouts to isolate biological causes from noise. We also break down the classic “wolves vs. dogs in snow” problem, why traditional CV models fixate on spurious correlations (like backgrounds), and how causal representations could force AI to focus on actual causal features.
#CausalVision #RecursionAI #MedicalAI #ComputerVision #MachineLearning
