Understanding the Fundamentals of AI

Understanding the Fundamentals of AI

What is Artificial Intelligence?

Artificial Intelligence (AI) is a broad field of study, much like physics. Machine learning is a subfield of AI, similar to how thermodynamics is a subfield of physics. Deep learning is a subset of machine learning, and within deep learning, we have discriminative and generative models. Large Language Models (LLMs) fall under deep learning, and the intersection of generative models and LLMs powers applications like ChatGPT and Google Bard.

What is Machine Learning?

Machine learning involves training a program with input data to make predictions on new, unseen data. For example, a model trained on Nike sales data can predict Adidas shoe sales based on Adidas sales data.
Supervised Learning: Uses labeled data (e.g., historical data with known outcomes like tip amounts based on bill amounts and order type) to make predictions.
Unsupervised Learning: Uses unlabeled data to identify patterns and group data (e.g., grouping employees based on their income and tenure).

What is Deep Learning?

Deep learning is a type of machine learning that utilizes artificial neural networks, inspired by the human brain. These networks consist of multiple layers, with more layers generally indicating a more powerful model.
Semisupervised Learning: A combination of supervised and unsupervised learning where a deep learning model is trained on a small amount of labeled data and a large amount of unlabeled data. This is useful in situations where labeling all data is impractical, such as fraud detection in banking.

What is Generative AI?

Unlike discriminative models that classify data (e.g., cat vs. dog), generative models learn patterns in the training data and generate new content based on those patterns.
Types of Generative AI:
Text-to-Text: ChatGPT, Google Bard
Text-to-Image: Midjourney, DALL-E, Stable Diffusion
Text-to-Video: Google Imagen Video, CogVideo, Make-a-Video
Text-to-3D: OpenAI's Shape-e
Text-to-Task: Models trained to perform specific tasks (e.g., summarizing emails).

What are Large Language Models?

LLMs are a subset of deep learning. They are pre-trained on massive datasets and then fine-tuned for specific purposes. This is analogous to a general-purpose dog being trained for a specific role like police work or guiding the visually impaired.
Real-world Applications: Hospitals can fine-tune pre-trained LLMs with their own medical data to improve diagnostic accuracy.