Скачать или смотреть Core AI Principles 2: Training on Colossal Data Sets | Applied AI Literacy [2025]

The second principle of AI is Training. How does an AI know how to speak, reason, and answer questions? This video explores the massive scale of data used to train modern models and how they learn connections and language through unsupervised training.

🎓 In this video, you'll learn about:
-Colossal Data Sets: How AI models are trained on essentially all the information on the internet that companies can acquire.
-Unsupervised Learning: How models are left to "figure it out" on their own, internalizing connections and developing language without constant human supervision.
-Reinforcement: How human feedback (thumbs up/thumbs down) helps refine the AI's understanding of "good" answers versus "bad" answers.
-System Instructions: How companies like Anthropic or OpenAI give their models specific rules to live by (e.g., "You are a helpful assistant") before they interact with users.

The reason AI is able to generate anything is because it has been trained on a colossal amount of data. Unlike older systems that used supervised learning to find specific patterns, modern Generative AI utilizes unsupervised learning. The model consumes vast amounts of information (roughly the entire internet), internalizes the connections between data points, and teaches itself to imitate human language. This process is followed by reinforcement training and the application of system instructions to ensure the AI is helpful and safe.

📚 ABOUT THIS COURSE: This video is part of the Applied AI Literacy Course developed by Tulane Professor Knud Berthelsen for Newcomb-Tulane College Career Services. Tulane students, faculty, and staff can enroll in the complete course at: https://careerengagement.tulane.edu/r...

📖 RELATED PLAYLISTS:
Why AI Literacy Matters:    • Why AI Literacy Matters  

🔗 RESOURCES:
CAIDS Website: https://datainstitute.tulane.edu/
Professor Berthelsen's Website: https://knudberthelsen.com/

📍 ABOUT CAIDS: The Connolly Alexander Institute for Data Science (CAIDS) at Tulane University advances AI literacy and data science education through research, teaching, and community engagement.

👨‍🏫 ABOUT THE INSTRUCTOR: Professor Knud Berthelsen holds a joint appointment in the Strategy, Leadership & Analytics Minor (SLAM) at the School of Liberal Arts and at the Connolly Alexander Institute for Data Science (CAIDS). His work focuses on generative AI tools, their societal impact, and increasing applied AI literacy.

📧 CONNECT WITH CAIDS: Subscribe for more AI education content Questions? Leave them in the comments Visit: https://datainstitute.tulane.edu/
#ArtificialIntelligence #MachineLearning #LargeLanguageModels #TulaneUniversity #AILiteracy #AITraining #UnsupervisedLearning