---
title: Foundation models 101
description: "In the rapidly evolving AI landscape, this is a big paradigm shift you should know about. So here is a quick 101: #### What is a Foundation Model? Foundation models are a class of models that meet two general criteria:…"
url: "https://members.botnirvana.org/playbook/foundation-model-101/"
updated: "2023-10-03"
category: Generative AI
---

# Foundation models 101

In the rapidly evolving AI landscape, this is a big paradigm shift you should know about. So here is a quick 101:

#### What is a Foundation Model?

Foundation models are a class of models that meet two general criteria:

- **Pretrained**: They are trained on a dataset that is both broad in scope and massive in size.
- **Adaptable**: It's designed to be general-purpose, and capable of handling a wide variety of downstream tasks.

#### Capabilities of Foundation Models

These models acquire various capabilities that can power your applications:

🗨️ Foundation models can process different styles, dialects, and languages, helping you communicate effectively with diverse audiences.

👁️ These powerful models can process and interpret visual data, enabling them to understand and generate images.

🤖 Foundation models can help develop "generalist" robots capable of performing myriad tasks across physically diverse environments.

🔍 Their multi-purpose nature along with their strong generative and multimodal capabilities offer new leverage for reasoning & search.

🤝 Foundation models can transform the developer and user experience for AI systems, making it easier to prototype and build AI applications.

#### Example Applications of Foundation Models

- Healthcare and Biomedicine: Foundation models can improve patient care and biomedical research by leveraging vast amounts of data across many modalities.
- Law: Foundation models can help attorneys read and produce long coherent narratives that incorporate shifting contexts and decipher ambiguous legal standards.
- Education: Foundation models can leverage relevant data from outside the domain and make use of data across multiple modalities to improve educational tasks.

#### Foundation Models Available

- Text Generation: GPT-4, PaLM 2, Alpaca-7B, FLAN-T5 XL, LLaMA-2
- Speech Recognition: Whisper
- Image Generation: Stable Diffusion
- Text-to-Speech: Bark
- Image Classification: CLIP

Foundation models have demonstrated raw potential, but we are still in the early days. It’s probably a big opportunity if we are to use it responsibly. What do you think? 💭