Fundamentals of Artificial Intelligence – Part Three: Concepts
Yes, Did you like our first two posts related to Artificial Intelligence (AI) learning for Linux users?So today you'll most likely appreciate this third part, where we'll cover several of the most important concepts and terms associated with this technological field. And if you haven't read the previous two parts, we'd like to clarify that In the second part we will address the point about the difference between Symbolic Artificial Intelligence and Connectionist Artificial Intelligence and on the historical milestones of this technological field over the last 100 years. Whereas, in the In the first part we focus en What is AI and about the various types that have already been created (Narrow or Weak, Generative and Agent) and that are expected to be created (General or Strong, and Artificial Superintelligence).
And as on previous occasions, keep in mind that just like in their day Electricity, Radio, Television, Computers, Telephones, the Internet and Social Networks; Artificial Intelligence is a technology that has arrived to become deeply integrated into everythingAnd that, by saving us valuable time and costly working hours, and making us more productive, creative, and versatile; most certainly, Little by little, it will displace those who have the least idea of its existence and operationTherefore, there's nothing better than starting to get some context about everything related to it.
Fundamentals of Artificial Intelligence – Part One: Origins II
But, before we delve into this "The third part of the exciting and surprising field of Artificial Intelligence"And more specifically regarding some of the most important concepts and terms to consider, understand, and master, we recommend exploring the previous and recent related publication in this series of publications, upon finishing reading it:
Fundamentals of Artificial Intelligence – Part Three: Related Concepts and Terms
20 important concepts and terms that are always mentioned in the field of Artificial Intelligence
1. Hallucination
Hallucination refers to a situation in which an AI system, especially one handling natural language processing, begins to generate results that may be irrelevant, nonsensical, or incorrect based on the input provided. This often occurs when the AI system is unsure of the context, relies too heavily on its training data, or lacks a proper understanding of the subject matter.
2. Machine Learning
A subfield of AI that encompasses the development of algorithms and statistical models that allow machines to improve their performance with experience. An example would be a machine learning algorithm that could predict the likelihood of customer churn based on past behavior. Thanks to a machine learning program, the machine learns to solve problems from examples: it can thus compare and classify data and even recognize complex shapes.
3. Deep Learning
A subfield of machine learning that uses multi-layered neural networks to learn from data, meaning it employs different neural network structures to learn successive layers of increasingly meaningful data representations. Currently, it is the most popular field within AI. It emerged from the combination of machine learning algorithms with formal neural networks and the use of big data.
4. Big Data
Big Data is the branch of Information Technology that studies the inherent difficulties of handling large datasets. Also called Macrodata, these refer to large and complex datasets that are often too large or varied to be processed using traditional methods. Furthermore, they come from a variety of sources (social networks, sensors, etc.) and require advanced techniques, such as artificial intelligence and machine learning, to analyze them and extract useful information.
5. Data set
A dataset is a collection of data, typically structured in a table. It is an organized collection of data used to train, test, or validate artificial intelligence models. It can contain text, images, videos, or other types of information, usually labeled to allow machine learning algorithms to recognize patterns and make predictions.
6. Model training
Model training is the process of using a dataset to teach an artificial intelligence or machine learning model to perform a specific task, such as classification or prediction. The model adjusts its parameters based on the provided examples to improve its accuracy with new data.
7. Enhanced Generation by Recovery
RAG (Retrieval-Augmented Generation) is a method that allows an AI to query external databases or search the internet for new and updated information before generating a response. Instead of having to retrain the entire model with new data, RAG functions as a kind of "real-time education" for the AI. This is very useful because it forces the system to rely on specific and verified documents, which greatly reduces the risk of the AI fabricating information (hallucinations).
8. GPT (Generative Pre-trained Transformer)
They are a family of language models developed by OpenAI that use neural networks to understand and generate text naturally. These GPT (Generative Pretrained Transformer) models use an architecture called Transformer that allows them to analyze prompts and predict the most likely response based on training with massive datasets. Because of this, GPT is able to mimic human language comprehension capabilities.
9. Natural Language Generation (Natural Language Generation)
Natural language generation is a subfield of artificial intelligence that involves automatically producing understandable texts or speech in human language. It allows a machine to transform raw data into natural sentences or paragraphs, such as in automated summaries or virtual assistants.
10. Prompt Engineering
It is the art of developing prompts for requests to achieve meaningful generative outputs—that is, more accurate, productive, and innovative outputs. Or, put another way, it is the science of designing, optimizing, and refining the instructions we give to a generative artificial intelligence. Its goal is to guide the AI to obtain accurate, relevant, and high-quality answers, acting as a communication bridge to avoid generic or erroneous results.
11. Model
A model is the object that will represent the output of the learning algorithm. Therefore, it's what AI uses to make output predictions. For example, foundational models are a broad category of AI models that include large language models and other types of models such as computer vision and reinforcement learning models. They are called "foundational models" because they serve as the foundation upon which to build applications, addressing a wide range of domains and use cases.
12. Large Language Models
They are a type of deep learning model trained on a large dataset to perform natural language understanding and generation tasks. Well-known models include BERT, PaLM, GPT-2, GPT-3, GPT-3.5, and the innovative GPT-4. These models vary in size (number of parameters that can be adjusted), the range of tasks they can perform (coding, chat, scientific, etc.), and the knowledge base on which they were trained.
13. Small Language Models
They are pocket-sized versions of large language models, or LLMs. Both use machine learning techniques to help them recognize patterns and relationships so they can produce realistic natural language responses. But while LLMs are enormous and require a lot of computing power and memory, small language models like Phi-3 are trained on smaller, more curated datasets and have fewer parameters, making them more compact and even usable offline.
14. Bias
These are the assumptions made by an AI model about the data. A “bias-variance trade-off” is the balance that must be struck between the assumptions a model makes about the data and the amount a model’s predictions change, given different training data. Inductive bias is the set of assumptions a machine learning algorithm makes about the underlying distribution of the data.
15. Natural Language Processing
It is an interdisciplinary field whose central idea is to give machines the ability to read and understand the languages humans speak. Natural Language Processing research aims to answer the question of how people are able to understand the meaning of a spoken or written sentence and how they understand what happened, when and where it happened; and the differences between an assumption, a belief, and a fact. Consequently, it involves programming computers to process massive volumes of linguistic data and focuses on transforming free-form text into a standardized structure.
16.Prompt
A prompt is an instruction or set of instructions (input) given to a computer system to generate a response. In the context of Artificial Intelligence, it is the initial instruction, command, question, request, or text statement provided to the model to guide and direct the generation of a specific response or output. The prompt acts as a stimulus or trigger for the AI, indicating the topic, style, direction, or type of response expected. The way the prompt is formulated (its framing) directly determines the quality and relevance of the results received.
17. Neural Network
They are a computational model based on a large set of simple neural units (artificial neurons), roughly analogous to the behavior observed in the axons of neurons in biological brains. They are the fundamental working unit of Deep Learning. Furthermore, it is a machine learning model that allows for pattern recognition and problem-solving, based on machine learning or deep learning, seeking to model the analytical mechanisms of the human brain.
18. Transformer
Transformers represent a type of neural network architecture designed to process sequential data, such as text. Transformers learn context and, consequently, meaning. They capture the relationships between different elements of a sequence (words, sentences) in parallel, rather than sequentially like traditional models. Transformers are the foundation of successful models like GPT and BERT and are used for natural language processing tasks.
19. Token
A token in Artificial Intelligence is the basic unit of data (words, parts of words, characters, or symbols) that Language Learning Models (LLMs) like GPT or Gemini use to process, understand, and generate text. It is the "currency" or essential piece of information that AI reads and analyzes, allowing it to break down human language into manageable (computable) chunks. Consequently, AI models process tokens to understand the relationships between them and unlock capabilities such as prediction, generation, and reasoning. The faster the tokens are processed, the faster the models can learn and respond.
20. Cross-validation
It is a method used to evaluate the results of an analysis and ensure the accuracy of machine learning model predictions. Therefore, it is a technique for evaluating machine learning models that focuses on dividing a dataset into several subsets (or "folds"). The model is then trained on some subsets and tested on others. This allows for a more reliable estimation of the model's performance by reducing overlearning.
Other fundamentals of Artificial Intelligence worth investigating, understanding and mastering in this third part
These are just a few concepts and terms out of the many that are truly worth knowing, understanding, and mastering. in this modern and innovative field of Artificial Intelligence. Therefore, to expand upon the information offered here and easily guide you to more, we offer you this Interesting infographic created with AI to give you a small idea of others that are equally important and essential:
Summary
In short, we hope that this "The third part of the exciting and surprising field of Artificial Intelligence" with an emphasis on Generative Artificial Intelligence (GAI), may have contributed a little more to strengthening and increasing your knowledge and skills in this modern and innovative field. Don't forget that, Artificial intelligence and robotics are changing the world and the way we live our lives.And that in years to come, when quantum computing becomes commercially available and quantum energy (produced through the Casimir effect) is widely accessible to all of humanity, it could be completely different from how we experience it today. So, it's better to learn about Artificial Intelligence now than to refuse and have regrets tomorrow. And of course, don't forget to continue participating, contributing, and supporting our beloved and valuable Linuxverse.
Lastly, remember visit our «homepage» in SpanishOr, in any other language (by simply adding 2 letters to the end of our current URL, for example: ar, de, en, fr, ja, pt and ru, among many others) to find out more current content.