AI Needs to Feel Pain скачать в хорошем качестве

AI Needs to Feel Pain

Without feeling, our decision making is paralyzed. This is the full history of how we taught machines to feel using synthetic pain (reinforcement learning). Take control of your digital identify for free at https://ace.me Get access to all top AI models https://higgsfield.ai/nano-banana-2-i... use promo code: AOTP-10-OFF I explore the history from 1961 through the humanoid robots of today. I cover Value Functions (which we can think of as primitive emotions), Q learning and Policy Functions. This video features insights from Ilya Sutskever on why emotion is essential for intelligence (in order to care about outcomes). This is a full history of reinforcement learning: from games to physical control, from simulation to reality. I show how AI develops human-like behaviors, the relationship between actions and language, and cutting-edge embodied AI from Figure, Boston Dynamics, and Physical Intelligence. Let me know what you think in comments! JOIN my email list for occasional updates: https://forms.gle/TGgmnqXcP4KUFBRY6 Consider supporting this original work:   / artoftheproblem   00:00 - Introduction 00:32 - Learning Tic Tac Toe 02:00 - Learning Cart and pole 04:20 - Shannon & Chess 06:50 - Samuel's Checkers 09:25 - TD Gammon (Gerald Tesaruo) 11:00 - TD Learning 14:30 - Learning Atari (DQN) 17:28 - DIrect Policy Gradiant 19:40 - Domain Randomization Featuring insights from: Claude Shannon Arthur Samuel Gerald Tesauro Richard Sutton David Silver Deep Mind/Open AI etc.