Uploaded December 2025 | Updated September 2026, 2 weeks ago

---

At #FTC2024, Peter Triantafillou highlighted a core challenge in generative AI: we often don’t know what data large language models are trained on. 🤖⚠️ Asking a model to “not generate this again” isn’t simple—its weights encode knowledge across many cases, some to unlearn, others to retain. This makes unlearning an alignment problem, requiring outputs to reflect human values and avoid harmful, explicit, or copyrighted content. 🔍🛡️

Watch the full video on our Youtube channel @SAIConference

#FTC2024 #PeterTriantafillou #GenerativeAI #MachineUnlearning #AIAlignment #ResponsibleAI #AIEthics #TrustworthyAI #AIIntegrity #DigitalSafety #SafeAI #AIResearch #FutureOfAI