Generative AI
AI models that can create original content (text, images, code) from textual descriptions.
AI models that can create original content (text, images, code) from textual descriptions.
Définition détaillée
Generative AI refers to deep learning models capable of generating new, coherent content: text, images, code, music, videos. Unlike discriminative models that classify or predict, generative models learn the underlying data distribution to produce novel samples. Common architectures include Transformers (GPT, Claude) for text and diffusion models (Stable Diffusion, DALL-E) for images.
Cas d'usage
Document analysis report generation: generative AI synthesizes hundreds of document pages into a structured report with executive summary and actionable recommendations.
Termes associés
Questions fréquentes
What are the risks of generative AI?
Hallucinations (made-up information), training data biases, copyright violations, energy costs. These can be mitigated through fine-tuning and human validation.
Can generative AI run locally?
Yes, open-source models like Llama, Mistral, or Phi can be deployed locally to ensure data privacy.