Large Language Models

Deep learning models (Transformer architecture) with billions of parameters, pre-trained on massive text corpora to execute complex natural language processing (NLP) and generative AI tasks.

Large Language Models (LLMs) are the core engine of generative AI: massive, Transformer-based neural networks (like GPT-4, LLaMA) trained on petabytes of text data. These models, often featuring billions of parameters (e.g., GPT-3 has 175 billion), function as sophisticated statistical prediction machines. They excel at a wide range of tasks: summarization, translation, code generation, and conversational AI (ChatGPT). While powerful, their output requires careful validation: they are prone to 'hallucination' (generating false information) and can reflect biases present in their training data.

https://www.ibm.com/topics/large-language-models