9 terms
Showing all terms starting with G
A framework where two neural networks (generator and discriminator) compete - one creating synthetic data, the other judging its authenticity.
AI systems that create new content - text, images, audio, video, or code - based on patterns learned during training.
Generative Pre-trained Transformer - OpenAI's series of LLMs trained to generate coherent, contextually relevant text from prompts.
An optimisation algorithm that iteratively adjusts model parameters in the direction that reduces the loss function during training.
Constraints built into AI systems to prevent harmful, biased, or off-topic outputs and keep models within safe operating boundaries.
A training paradigm where a model first learns language patterns unsupervised on large corpora before being adapted for downstream tasks.
A class of neural networks designed to operate on graph-structured data, useful for social networks, molecules, and knowledge graphs.
Connecting AI outputs to verifiable real-world information or data sources to reduce hallucinations and improve factual accuracy.
A post-training quantisation method for LLMs that compresses model weights to 4-bit or lower with minimal accuracy loss.
