GPT-3 - Language Models are Few-Shot Learners | Paper Explained

❤️ Become The AI Epiphany Patreon ❤️ ► https://www.patreon.com/theaiepiphany ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ In this video, I cover the famous GPT-3 model. I first give you some context about the stuff that happened since the paper was first published in May 2020 (hype, anti-hype, limitations, and cool apps), and then I dive deep into explaining the paper. You'll learn about: ✔️ Useful resources on GPT-3 ✔️ Main takeaways from the paper ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ ✅ "anti-hype" blog: https://lacker.io/ai/2020/07/06/giving-gpt-3-a-turing-test.html ✅ Gwern's blog: https://www.gwern.net/GPT-3 ✅ My transformer implementation: https://github.com/gordicaleksa/pytorch-original-transformer ✅ Cool "GPT game": https://play.aidungeon.io/ ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ ⌚️ Timetable: 00:00 GPT (anti)hype, Gwern, prompt programming 04:30 Abstract of the paper 06:50 Architecture, data, compute 12:15 Zero-shot, one-shot, and few-shot learning 18:45 Power-law chart (more compute please) 20:35 Results (machine translation) 23:05 NLI (reasoning is hard) 24:40 Arithmetic 26:25 Word unscrambling 28:40 SAT analogies (how smart are humans?) 30:45 Fake news generation 32:05 Data contamination 35:05 Limitations of the model 37:35 Bias, fairness (broader impact) 44:30 Final thoughts, are we going towards an AGI? ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ 💰 BECOME A PATREON OF THE AI EPIPHANY ❤️ If these videos, GitHub projects, and blogs help you, consider helping me out by supporting me on Patreon! The AI Epiphany ► https://www.patreon.com/theaiepiphany One-time donation: https://www.paypal.com/paypalme/theaiepiphany Much love! ❤️ ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ 💡 The AI Epiphany is a channel dedicated to simplifying the field of AI using creative visualizations and in general, a stronger focus on geometrical and visual intuition, rather than the algebraic and numerical "intuition". ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ 👋 CONNECT WITH ME ON SOCIAL LinkedIn ► https://www.linkedin.com/in/aleksagordic/ Twitter ► https: