How to Make AI Forget скачать в хорошем качестве

How to Make AI Forget

Episode Notes: https://thedataexchange.media/hirundo... In this episode, Ben Lorica speaks with Ben Luria, CEO and co-founder of Hirundo, about the emerging necessity of machine unlearning for enterprise AI. *Sections* The Core Problem: AI Models Can Learn but Cannot Forget - 00:00:50 Team Origins and Identifying the Scientific Gap in AI Infrastructure - 00:02:45 Why Guardrails and Context Engineering Fail at the Core - 00:05:58 Defining Machine Unlearning as AI Neuro-surgery - 00:08:02 Common Representations in Enterprise: Security, Bias, and PII - 00:10:13 Strategies for Behavioral vs. Data Unlearning - 00:11:41 The Risk of PII Leaks in Customer Interaction Models - 00:14:04 Measuring Success through Red Team Benchmarks and Utility Preservation - 00:16:39 The Flexibility of Weight Editing: Re-learning and Pattern Recognition - 00:22:45 Security Vulnerabilities in Reasoning Models and Recursive Addressing - 00:25:49 The Unlearning Workflow: Customization vs. Predefined Categories - 00:32:05 The Open Weights Landscape: US vs. Chinese Foundation Models - 00:34:33 Applying Unlearning to Autonomous Agents and Coding Models - 00:39:23 Future Roadmap: Productization, Multi-modality, and Industry Standards - 00:42:31 Establishing Trust through Explainability and Visualization Tools - 00:47:04 Support our work: subscribe to our newsletter 📩 https://gradientflow.substack.com/ leave us a tip 💰 https://buymeacoffee.com/gradientflow