Understanding the concepts of Foundation Models, Large Language Models and Artificial General Intelligence

Decoding the AI jargon

Valentina Alto
5 min readMay 3


Since the advent of ChatGPT in November 2022, Generative AI became one of the hottest topics to talk about. It is worth noticing that research in the AI and generative AI domains has been going on for years; yet, the incredible usability and performance of ChatGPT filled the gap between research and the general user.

In the rapidly evolving world of AI, it’s easy to get lost in a sea of buzzwords and jargon. Three terms that have gained significant attention in recently are Foundation Models, Large Language Models (LLMs), and Artificial General Intelligence (AGI). In this article, we will go through the meaning of all of them as well as the relationship among them.

Foundation Model

A foundation model is a type of pre-trained AI model that can be fine-tuned for various specific tasks. These models are trained on massive amounts of data and can be adapted to perform a wide range of tasks, such as natural language processing, translation, and content generation.

Some of the main features of foundation models include:

  1. Pre-training: Foundation models are pre-trained on massive amounts of data, often from diverse sources, allowing them to learn general patterns, structures, and relationships within the data. This pre-training phase enables the models to acquire a broad understanding of various domains and develop a strong base for further fine-tuning.
  2. Fine-tuning: After the pre-training phase, foundation models can be fine-tuned for specific tasks using smaller, task-specific datasets. This adaptability allows developers to create specialized AI applications built on top of the foundation models, leveraging their pre-trained knowledge to achieve high performance with relatively less data and training time.
  3. Transfer learning: Foundation models are designed to leverage transfer learning, which means they can apply knowledge learned during pre-training to new, related tasks. This ability to transfer knowledge across tasks makes them highly versatile and efficient, as they can quickly adapt to new tasks with minimal additional…



Valentina Alto

Data&AI Specialist at @Microsoft | MSc in Data Science | AI, Machine Learning and Running enthusiast