# The Generative and Agentic Era (2020–present)

From 2020, AI shifted toward large generative models and autonomous agents. This sub-article expands [Section 7 of History](../history.md) and complements [Recent Topics](../recent.md).

## Large Language Models
- **GPT-2, GPT-3, GPT-4** and open models like Llama demonstrated few-shot learning, code generation, and general text understanding.
- **Scaling laws** showed predictable gains from compute, data, and parameters.

## Multimodal and Generative Models
- Models now process text, image, audio, and video together.
- **Diffusion models** (DALL·E, Stable Diffusion, Sora) generate images and video from text.

## Agents and Tool Use
- Models increasingly act as **agents**: planning, calling tools, browsing, and taking actions.
- Frameworks (LangChain, ReAct, computer-use agents) connect models to the world. See [Agents & Tooling](agents-tooling.md).

## Safety and Governance
- Frontier labs publish safety evals; the **EU AI Act** (2024) and international summits address risk. See [AI Safety](ai-safety.md).

## Why It Matters
This era made AI broadly accessible and consequential, shifting the debate from capability to alignment, safety, and societal impact.

## Sources
- OpenAI. *GPT-3 Language Models are Few-Shot Learners* (2020).
- European Parliament. *AI Act* (2024).

*Back to [History](../history.md) · See [Recent Topics](../recent.md).*