Overview
Generative AI learns patterns in data and uses them to synthesize new outputs. The research frontier includes controllability, quality, provenance, interaction design, evaluation, and the effects of deploying generators in creative and knowledge work.
What it is
Generative models estimate or approximate a data distribution so they can sample, complete, transform, or edit examples. They may use autoregressive prediction, diffusion, latent-variable models, or other approaches.
How it works
Training exposes a model to examples and an objective for reconstructing, predicting, denoising, or matching data. At generation time, conditioning information and a sampling process guide output creation. Product systems add filters, editing controls, retrieval, and human review.
Key concepts
- Autoregressive modeling
- Diffusion and denoising
- Latent representations
- Conditioning and control
- Data provenance
- Quality and safety evaluation
Current research questions
- How can creators control structure, style, and factual content together?
- How should synthetic data and generated media be identified and governed?
- What training data is memorized, and how can unwanted reproduction be reduced?
- How can evaluation reflect originality, usefulness, and downstream impact?
Applications
- Design and creative production
- Code and document drafting
- Simulation and synthetic data
- Education and accessibility
- Scientific hypothesis generation
Relevant research papers
A live selection from the ScholarLens research index. Open any result to read its full paper page and use the Reading Assistant where available.
Keep exploring
Search the literature with your own question, or open a paper and read it closely with ScholarLens.