Главная
Study mode:
on
1
- Intro & Overview
2
- Language Models
3
- Language Modeling Datasets
4
- Model Size
5
- Transformer Models
6
- Fine Tuning
7
- In-Context Learning
8
- Start of Experimental Results
9
- Question Answering
10
- What I think is happening
11
- Translation
12
- Winograd Schemes
13
- Commonsense Reasoning
14
- Reading Comprehension
15
- SuperGLUE
16
- NLI
17
- Arithmetic Expressions
18
- Word Unscrambling
19
- SAT Analogies
20
- News Article Generation
21
- Made-up Words
22
- Training Set Contamination
23
- Task Examples
Description:
Dive into an in-depth exploration of GPT-3, a groundbreaking language model, in this comprehensive video lecture. Examine how scaling up language models significantly improves task-agnostic, few-shot performance, potentially rivaling state-of-the-art fine-tuning approaches. Learn about the model's architecture, training process, and its impressive capabilities across various NLP tasks, including translation, question-answering, and complex reasoning. Discover the model's strengths in generating human-like text and its performance on challenging tasks such as arithmetic expressions and word unscrambling. Explore the broader implications of GPT-3's capabilities, including potential societal impacts and methodological challenges related to training on large web corpora. Gain insights into the future of natural language processing and the potential of large-scale language models to revolutionize AI applications.

GPT-3 - Language Models Are Few-Shot Learners

Yannic Kilcher
Add to list
0:00 / 0:00