In today’s fast-paced technological world, understanding the basics of Artificial Intelligence (AI) has become crucial, for all companies, not just those in Technology. 2023 was marked as the breakout year for Generative AI, with the topic at the forefront of most conversations relating to technology as we move deeper into 2024
For most, however, despite lots of chatter, AI can be another subject only understood by experts, filled with acronyms and technical “jargon”, that’s challenging to break down and identify the real-world use cases for their organisation.
This post aims to demystify AI and its subset, Large Language Models (LLMs) and the concept of Generative AI for those new to these concepts. It’s part of a new series of posts from Techary technical teams, codenamed the “AI Genius” team. The aim is to help, through a series of blog posts, give an introduction to the key concepts of AI, and share our thoughts and experiences on how this may impact both what we do, and how our customers can plan the developments across their technology stack.
In this, the first post in this series, we’ll look at:
- What is AI?
- Neural Networks: The Building Blocks of AI
- Large Language Models (LLMs)
- Generative AI
What is Artificial Intelligence (AI)?
Artificial Intelligence, in its simplest form, refers to the ability of a computer or a machine to think, learn, and act like a human. AI is not just a single technology but a broad field encompassing various techniques and technologies, from simple automated responses to complex problem-solving capabilities.
AI in Everyday Life
You interact with AI more often than you might realise. It powers the recommendations on your favourite streaming services, the voice assistant in your smartphone, and even the spam filter in your email.
AI Evolution
When you step back to think about it, AI has been enhancing and developing across our lives for some time now. Think of when self-checkouts started arriving at supermarkets, essentially, that could be deemed an AI-driven improvement. AI as the generalist phrase has been spoken through various terms in recent times, probably more than the generalist term, but AI encompasses a wide range of terms, depending on the specific context, such as; Machine Learning, Computation Intelligence, Cognitive Computing, Deep Learning, Smart Technology, Data Science, Neural Networks, Predictive Analysis — all topics that centrally link back to the evolution of AI.
Essentially, AI has been a focus of technological advancement for longer than we think, through the various terms and developments in recent years.
Large Language Models (LLMs): The Power of Words
A significant breakthrough in AI is the development of Large Language Models like OpenAI’s GPT-4. These are advanced AI systems trained on vast amounts of text data. They can understand, generate, and interact using natural language, making them highly versatile for various applications.
How LLMs Work
Imagine a librarian who has read every book in the world and can recall any information instantly. LLM are the digital equivalent, offering responses based on the vast information they’ve been trained on.
Natural Language Processing
Natural language processing is the ability of machine to mimic human speech. In order to mimic human speech, natural language processing needs to consider things such as past, present and future tenses. Algorithms that are used for NLP also need consider linguistic idiosyncrasies such as context, sound, meaning and syntax. Algorithms then use LLMs to achieve natural language processing.
Tokenisation
Tokenisation is the process of adding values to a given variable to help to train a LLM. There are many algorithms that are used, such as byte pair encoding. Characters are assigned a numerical value, and these can be done in pairs to look up the frequency of patterns. The values are then stored in a tensor such as a matrix, which is 3 dimensional.
Neural Networks: The Building Blocks of AI
A transformative element in the field of Artificial Intelligence is the advent of neural networks. These computational models are inspired by the human brain and consist of interconnected nodes or “neurons” that work together to process and analyse data. Neural networks are capable of learning from experience, making them incredibly effective for pattern recognition, classification, and predictive modelling across numerous industries.
The Roles of Neural Networks in LLMs
For Large Language Models, neural networks are trained to handle and predict language patterns. This training involves feeding the network a vast corpus of text data which it uses to learn about everything from basic grammar to complex contextual relationships between words and phrases. As it learns, the model adjusts the weights of the connections between nodes in the network, fine-tuning its predictions to improve accuracy. The architecture of choice for many LLMs, including GPT-4, is the transformer neural network model.
Generative AI: The Creative Aspect of Artificial Intelligence
Generative AI is a fascinating subset of artificial intelligence that focuses on creating new content, be it text, images, or even music. Unlike traditional AI that is primarily about understanding or processing data, generative AI takes a step further by producing new, original content based on its training.
The Magic of Generative AI
Imagine an artist who can paint in any style, a composer who can create music in any genre, or a writer who can pen down anything from poems to technical manuals. That’s what generative AI can do in the digital world. It learns from existing data and then uses that knowledge to generate new, unique creations.
Generative AI vs. Traditional AI
While traditional AI systems are designed to understand, classify, or respond based on the data they’re fed, generative AI is about creating something new and often unexpected. Traditional AI might help you sort your emails, but generative AI could write a new email in your writing style.
This post is aimed to provide an introduction to the key concepts and terminology used within AI. In our next blog post, we’ll focus on the Technology Stack Powering AI, and delve deeper into the different technologies, hardware and software being used to power this functionality.