The Future of Language Models
summary
Survey of current LMs: transformer-based models such as GPT-1/2/3 and BERT and Google’s LaMDA demo with Pluto and paper plane power chatbots and autocomplete but produce conspiracy and toxic outputs (examples: “vaccines sterilize,” Rothschild conspiracy) when trained on noisy corpora. Models scale to trillions of parameters, use Hugging Face for deployment, and the “stochastic parrots” critique (Mar 2021) claims bigger models alone do not fix bias or misinformation; mitigation strategies and data curation are needed.
AI-written summary · about
chapters (7)
- 0:00Dr. Ash Pahwa
- 4:39What is a Language Model?
- 6:19NLP Tasks Where a Language Model is Used
- 11:48How to Build a Language Model?
- 18:02Leading Language Models for NLP Applications
- 22:52How to Access Language Models (Transformers)
- 24:14Controversy Regarding Language Models
related talks
51mLarge Language Models: Friend, Foe, or Otherwise
43mBeyond the Hype: A Realistic Look at Large Language Models
51mTowards a Pragmatic Understanding of Large Language Models
01h 00mInside GPT – Large Language Models Demystified
30mLarge Language Models and Prompt Engineering for Developers
26mModern NLP for Proactive Harmful Content Moderation
52mUnleash the Power of Large Language Models (LLMs)
15mFuture with Large Language Models