GPT (Generative Pre-trained Transformer)

GPT (Generative Pre-trained Transformer) is OpenAI's series of language models that have defined the current AI era. The name describes three key aspects: they generate text, they're pre-trained on vast data before deployment, and they use transformer architecture.

Evolution of GPT:

• GPT-1 (2018): Proved the pre-training approach works
• GPT-2 (2019): Demonstrated coherent long-form text generation
• GPT-3 (2020): Revealed emergent capabilities and few-shot learning
• GPT-3.5/ChatGPT (2022): Made AI conversational and accessible
• GPT-4 (2023): Multimodal, stronger reasoning, longer context
• GPT-4o (2024): Optimised for speed and multimodal interaction

Key capabilities include:

• Natural language understanding and generation
• Code writing, review, and explanation
• Analysis and reasoning
• Translation and summarisation
• Following complex multi-step instructions

GPT models set the benchmark that other models are compared against. Their API is the most widely-used for AI application development.