SIGNAL//DESK
AI/MLsrc: curated AI glossary

Generative AI

Generative AI is a type of computer program that can create brand-new things like stories, pictures, or music from scratch, rather than just sorting or organizing information that already exists.

Generative AI refers to models designed to generate novel data outputs—such as text, images, or code—by learning the underlying patterns and structures of a training dataset, distinguishing them from discriminative models that focus solely on classification or regression tasks.

AI that produces novel content — text, images, code, audio, video — by sampling from a learned probability distribution, as opposed to discriminative models that only classify or predict.

evolution

  1. 2014 · history
    Generative Adversarial Networks (GANs)

    Ian Goodfellow introduced GANs, enabling neural networks to generate realistic synthetic data by pitting two models against each other.

  2. 2017 · history
    Transformer Architecture

    Google researchers published 'Attention Is All You Need,' providing the foundational architecture for modern large language models.

  3. 2020 · history
    GPT-3 Release

    OpenAI released GPT-3, demonstrating that massive scale in transformer models leads to emergent capabilities in text generation and reasoning.

  4. 2021 · history
    Diffusion Models

    The emergence of high-quality image generation models like DALL-E and Stable Diffusion marked a breakthrough in text-to-image synthesis.

  5. 2022 · history
    ChatGPT Launch

    OpenAI released ChatGPT, bringing generative AI to the mainstream through a conversational interface optimized for human interaction.


← all terms