GPT: Difference between revisions

From AI Wiki
No edit summary
No edit summary
Line 11: Line 11:
! Words
! Words
! Equivalent
! Equivalent
! hello
! BookCorpus
! hello
! WebText
! hello
! CommonCrawl
! hello
! hello
! hello
! hello

Revision as of 22:00, 2 March 2023

Main GPT Models

Model Parameters Context Window Training Data
Tokens Words Equivalent BookCorpus WebText CommonCrawl hello hello
GPT-1 117 Million 512 358
GPT-2 1.5 Billion 1024 716
GPT-3 175 Billion 2,048 1,433
GPT-3.5 xxx 4,000 2,800
ChatGPT xxx 4,096 2,867
GPT-4
(v1)
xxx 8,000 5,600
GPT-4
(v2)
xxx 32,000 22,400

What is GPT

GPT, which stands for Generative Pre-trained Transformer, is a type of language model developed by OpenAI. Based on the Transformer architecture and utilizes unsupervised learning, GPT is able to generate text indistinguishable from text written by humans.

How does GPT work?

Using unsupervised learning, GPT is able to generate text by predicting the next word in a sentence based on the context of the words that came before it. The model is trained on a large corpus of text, such as the entire internet, to learn the patterns and relationships between words and how they are used in context.

The Transformer architecture is used in GPT because it is well-suited for processing sequential data, such as text. The Transformer consists of self-attention mechanisms, which allow the model to focus on specific words in a sentence and learn the relationships between them.

Once the model is trained, it can be used for a variety of natural language processing (NLP) tasks, such as text generation, translation, and summarization. The model can also be fine-tuned for specific tasks by training it on smaller, task-specific datasets.

Applications of GPT

GPT has a wide range of applications in the field of NLP. Some of the most notable applications include:

Explain Like I'm 5 (ELI5)

GPT is a computer program that can write sentences like a person. It learned how to do this by reading lots of text on the internet. Now it can write new sentences and answer questions by using what it learned. It's like a smart robot that can talk and write in a way that sounds like a person!