AI GlossaryGPTENTRY — Generative AI
[Generative AI]
GPT.
Generative Pre-trained Transformer, a family of large language models trained to generate text.
In-depth explanation
01GPT models are autoregressive transformers trained to predict the next token. GPT-2 (2019) demonstrated remarkable text generation. GPT-3 (2020) with 175B parameters showed emergent capabilities like few-shot learning. GPT-4 (2023) is multimodal and powers ChatGPT. These models learn patterns from vast internet text and can perform diverse language tasks.
Examples
02EX. 01
GPT-3
EX. 02
GPT-4
EX. 03
ChatGPT
More in Generative AI
0301Diffusion ModelGenerative models that learn to create data by reversing a gradual noising process.02Generative Adversarial Network (GAN)Two neural networks competing against each other to generate realistic synthetic data.03HallucinationWhen AI models generate plausible-sounding but factually incorrect or fabricated information.04Large Language Model (LLM)AI models trained on vast text data that can generate and understand human-like text.05Prompt EngineeringThe practice of crafting effective inputs to get desired outputs from AI models.06RAG (Retrieval-Augmented Generation)Combining retrieval systems with language models to generate responses grounded in external knowledge.
[NEXT] — APPLY THE CONCEPT
Master GPT.
Learn how to apply this concept with hands-on projects in our comprehensive AI programs.