Decoding the AI/ML Jargon

Introduction

The other day, I was discussing some AI/ML concepts with a friend, and I realized that these concepts aren't always easy to explain. AI and ML have a language of their own, filled with technical terms that can be quite confusing. So, I thought it would be interesting to decode some of the most commonly used jargons in a way that’s easy to consume. By no means this is a complete list, but I have tried to capture some of the most commonly used/heard terms.

If you like this content please contribute by adding a Jargon and it's explanation in the comments.

With that background enjoy the read

Artificial Intelligence (AI)

Definition: AI is the simulation of human intelligence in machines that are programmed to think, learn, and solve problems. AI has a variety of methods and technologies like machine learning and more. The ultimate goal of AI is to create systems that can perform tasks that normally require human intelligence and decision making.

Think of AI as a smart assistant that can understand and perform tasks, just like a human but without getting tired or needing a break as long as it is powered on.

Artificial Narrow Intelligence (ANI)

Definition: ANI, also known as weak AI, is designed to perform a narrow task (e.g., facial recognition, sorting baggage in airports etc) and does not possess general intelligence. ANI systems are focused and optimized for specific problems.

A calculator is an example of ANI that is extremely good at performing arithmetic tasks but cannot do anything else.

Artificial General Intelligence (AGI)

Definition: AGI is a type of AI that can understand, learn, and apply knowledge across a wide range of tasks, similar to human intelligence. Unlike narrow AI, which is specialized for specific tasks, AGI aims for broad cognitive abilities.

It’s like having a robot with the ability to learn any skill or knowledge area just like a human as long as it is exposed to variety of tasks and trained to perform many things.

Artificial Super Intelligence (ASI)

Definition: ASI refers to AI that surpasses human intelligence across all fields, from creativity to problem-solving. ASI represents a point where machines not only perform tasks at human levels but exceed them.

Imagine an AI humanoid that not only learns and performs every possible task but does so far better and faster than any human could.

Machine Learning (ML)

Definition: ML is a subfield of AI that involves the use of algorithms and statistical models to enable computers to learn from and make predictions based on data. Unlike traditional programming, where a programmer explicitly codes all instructions, ML systems improve their performance through experience via a process called training and fine tuning.

Think of the animals in the circus which are trained to perform tricks. The more you train them (provide data), the better it gets at performing the tricks (making predictions).

Neural Networks

Definition: Neural networks are computing systems inspired by the biological neural networks of human brains. They consist of interconnected nodes (neurons) that work together to process input data and produce outputs. These networks can learn and adapt by adjusting the connections (weights) between neurons based on the data they process.

It is like a network of highways connecting various cities. Vehicles(information) travels through these highways (neurons) to reach its destination (output).

Deep Learning (DL)

Definition: Deep learning is a subset of ML that uses neural networks with many layers (deep neural networks) to analyze various factors of data. These multiple layers allow the model to learn increasingly abstract features, enabling it to perform complex tasks such as image and speech recognition.

An archaeologist uncovering an ancient artifact buried deep underground. The artifact has multiple layers of dirt and rock covering it. To uncover the artifact, you need to carefully dig through each layer. Each layer you remove reveals more details about the artifact until you can finally see and understand its complete form.

Natural Language Processing (NLP)

Definition: NLP is a field of AI that focuses on the interaction between computers and humans through natural language. It involves the ability of machines to read, understand, and generate human language. Applications of NLP include language translation, sentiment analysis, and chatbots.

A conductor of a symphony orchestra, where each musician plays different instrument and has a different language for interpreting the music. Your job is to interpret the musical score, communicate effectively with each musician, and ensure that they all play together in harmony to produce music.

Computer Vision

Definition: Computer vision is an AI field that trains computers to interpret and understand the visual world. Using digital images from cameras and videos and deep learning models, machines can accurately identify and classify objects, and then react to what they "see".

It’s like giving a computer a pair of eyes and a brain to identify objects, people in images and videos. The more you expose the system to images, the better it gets at recognizing the objects.

Supervised Learning

Definition: Supervised learning is a type of ML where the model is trained on labeled data. Each training example includes the input data and the corresponding correct output. The model learns to map inputs to outputs based on this training data.

It’s like learning with a teacher who provides the correct answers to guide you. Provides with proper material and textbooks with predefined details with no ambiguity.

Unsupervised Learning

Definition: Unsupervised learning involves training a model on data without labels, letting the model find patterns and relationships on its own. It’s often used for clustering, dimensionality reduction, and anomaly detection.

It’s like exploring a new city without a map, discovering interesting places and routes on your own. You make sense of the city as you move around and experience different aspect of the city eventually figuring out the city by yourself.

Convolutional Neural Network (CNN)

Definition: CNNs are a type of neural network particularly effective for analyzing visual data. They use convolutional layers to automatically and adaptively learn hierarchies of features from input images.

Everyone would have solved a jigsaw puzzle. Instead of tackling the entire puzzle at once, you break it down into smaller sections. You first focus on identifying and assembling the edges and corners. Then, you look for specific patterns and colors within each small section, gradually piecing them together. As you work through each section, the complete picture starts to emerge more clearly.

Recurrent Neural Network (RNN)

Definition: RNNs are a type of neural network designed to recognize patterns in sequences of data, such as time series or natural language. They have connections that form cycles, allowing them to maintain a memory of previous inputs.

A storyteller telling a long tale to an audience. As he narrates the story, he keeps track of the plot, characters, and events that have occurred so far. This memory helps him introduce new elements that make sense within the context of the story, reference past events, and build the narrative, ensuring that your tale remains engaging.

Machine Learning Model

Definition: A machine learning model is a mathematical representation of a real-world process created through ML training. It is built to recognize patterns in data and make predictions or decisions without explicit programming.

A chef learning to perfect a new recipe. He starts by gathering various ingredients and following a basic version of the recipe. After each attempt, he adjusts the amount of spices, cooking time, and techniques based on the feedback he received about the taste, texture, and appearance of the dish. Over time, through this process of trial and error, you refine the recipe until you consistently produce a delicious meal.

Pre-Trained Model

Definition: Pre-trained models are ML models that have been previously trained on a large dataset and can be fine-tuned for specific tasks. This approach saves time and resources compared to training a model from scratch.

A experienced chef who has already mastered the basics of cooking through years of experience and practice. Now to learn a new, specific cuisine, like Italian, instead of starting from scratch, he can leverage his existing cooking skills and knowledge, quickly adapting and fine-tuning them to master Italian recipes.

Generative AI

Definition: Generative AI refers to algorithms that can generate new content, such as images, music, or text. These models learn from existing data and can create new instances that resemble the training data. Examples include GPT-3 for text generation and GANs for image creation.

An experienced artist who has spent years studying different art styles, techniques, and historical works. With this deep knowledge on art, he is now capable of creating entirely new and original artworks. He can blend different styles, invent new patterns, and produce unique pieces that have never been seen before, all inspired by the vast array of art you have studied.

Large Language Models (LLMs)

Definition: LLMs are advanced AI models trained on vast amounts of text data to understand and generate human-like text. These models have billions of parameters and can perform a variety of language-related tasks, such as translation, summarization, and question answering.

A seasoned traveler who has visited every country in the world. Throughout her travels, she has learned multiple languages, experienced diverse cultures, and gained understanding of different customs and traditions. Now, whenever someone asks her about any place, she can provide detailed information, answer questions in the local language, and even share stories or write articles about her experiences.

Prompt Engineering

Definition: Prompt engineering is the process of designing and refining prompts to efficiently extract desired outputs from AI models, particularly language models. The quality and structure of the prompt significantly influence the model’s responses.

It’s like crafting the perfect question to get a specific answer from a knowledgeable expert which otherwise would lead to ambiguous answers and plenty of follow up questions to get the requirement information.

Generative Adversarial Network (GAN)

Definition: GANs are a class of AI models where two neural networks contest each other in a game. One network (the generator) creates new data instances, while the other network (the discriminator) evaluates them for authenticity. This competition helps both networks improve.

An inexperienced sculptor trying to create a statue that looks exactly like a famous masterpiece. He works in a studio where a master sculptor evaluates the attempts. Every time he create a new sculpture, the master examines it and provides feedback. Every time it gets better and over time, the sculptures become so accurate that even the master sculptor cannot tell the difference between his sculpture and the masterpiece.

Inferencing

Definition: Inferencing is the process of making predictions using a trained ML model. It involves feeding new data into the model and generating an output based on the model’s learned patterns.

An experienced surgeon who has performed thousands of operations. When a new patient comes in with a complex medical condition, you review their medical history, symptoms, and test results. Using his experience from past surgeries, he can diagnose the condition and decide on the best surgical procedure to treat the patient.

Retrieval Augmented Generation (RAG)

Definition: RAG is a technique that combines retrieval of relevant documents with the generative capabilities of language models. This approach enhances the quality of generated responses by grounding them in factual information.

A journalist writing an in-depth article. To ensure accuracy she has to first search through an extensive archive of past articles, books, and interviews to gather relevant facts, quotes, and data. She then uses this collected information to develop a detailed article. Her final article not only reflects her own writing skills and knowledge but also includes verified information from reliable sources.

MLOps

Definition: MLOps (Machine Learning Operations) is a practice that combines ML, DevOps, and data engineering to deploy and maintain ML systems in production. It focuses on the reliability and efficiency of ML systems, ensuring they operate smoothly and can be updated regularly.

It’s like an automobile manufacturing assembly line that builds, tests, and delivers cars with efficiency.

Edge AI

Definition: Edge AI refers to running AI algorithms locally on a device rather than in the cloud or a data center. This approach reduces latency and bandwidth usage, making it ideal for real-time applications.

Instead of you going to the gym and training with a personal trainer, a personal trainer who comes to your home instead to train in your personal gym.

I hope it helps demystify some of the AI/ML jargon. For those who are curious to dive deeper, here are some additional resources:

- [Towards Data Science - AI/ML Basics](https://towardsdatascience.com)

- [Machine Learning Mastery](https://machinelearningmastery.com)

- [Google AI Blog](https://ai.googleblog.com)

- [Google ML Glossary](https://developers.google.com/machine-learning/glossary)

Happy learning!