Understanding AI
What is Artificial Intelligence (AI)?
Welcome to the world of Artificial Intelligence (AI)! AI is a field of computer science that creates computer systems and software capable of performing tasks that typically require human intelligence. In essence, AI aims to replicate human cognitive functions using machines and algorithms. It includes a wide range of techniques and approaches, such as machine learning, neural networks, and natural language processing. AI has broad applications across various industries. It can automate repetitive tasks, enhance decision-making processes, and revolutionise how we interact with technology and the world around us.
Over the years, AI has taken on various meanings, but not all have been clear-cut. In 2017, a groundbreaking paper named “Attention is All You Need” emerged, proposing a fresh approach to AI that excels in understanding human language. This led to the birth of large language models (LLMs), like ChatGPT, Bing, Google’s BARD, and Anthropic’s Claude (2). These large language models can process and generate human-like text, opening doors to a new era of AI.
Simultaneously, AI has branched into areas like art, video, and creative tasks, collectively known as generative AI. This newfound capacity for creativity has transformed the landscape. These large language models and creative AI have reshaped what AI means today, especially in terms of generating new content.
Large Language Models (LLMs)
At their core, all AI models are in the business of prediction - trying to forecast what comes next. But when it comes to large language models, things take an intriguing turn. They are adept at predicting the very words or parts of words that will follow in a sentence.
So, how do they do it? The journey begins with an intensive training process. These models dive into the vast expanse of the internet, devouring billions of documents over months of dedicated computation. During this immersion, they develop a sophisticated understanding of language, unveiling connections between words. Some connections are as clear as "sun" and "shine," while others are more unexpected, such as "coffee" and "mornings." Beyond the surface, these models identify intricate relationships that span various dimensions, baffling even our human comprehension.
Armed with this intricate understanding, the AI embarks on its predictive mission. Imagine you start a sentence with "I like to use AI because it's very good at." In that case, the AI springs into action, attempting to guess the next word. Will it be "prediction," "lessons," or "exploration"? With its forecast in place, the AI promptly delivers the next word. Think of it as a finely tuned autocomplete, elevated to a new level of sophistication. This predictive mechanism forms the bedrock of how AI models function.
The table below is a quick reference that summarises the current state of LLMs (as of Summer 2023):
Source: https://substack-post-media.s3.amazonaws.com/public/images/158ee53f-0551-4381-988d-20f2d9b7adb2_2006x1169.png (2006×1169) (substackcdn.com)
Source: https://substack-post-media.s3.amazonaws.com/public/images/158ee53f-0551-4381-988d-20f2d9b7adb2_2006x1169.png (2006×1169) (substackcdn.com)
Until recently, large language models were intriguing but not overwhelmingly impressive. However, the debut of ChatGPT in late 2022 marked a turning point. This enhanced version unleashed an unexpected level of capability. Recent studies found that tasks involving writing and coding were done faster and better with ChatGPT's assistance. Interestingly it outperformed expectations, from attempting to complete tests to writing tasks. GPT-4, the successor, is larger, smarter, and even more efficient than its predecessor, accepting both text and image prompts. This progress was rapid and surprising, promising a future of potent AI prowess.
AI is an incredible tool that has the potential to revolutionise how we teach and learn at Queen’s. While it brings incredible potential, it also challenges us to balance its benefits with ethical considerations. Let's explore and adapt together as we navigate through a new era of AI-powered possibilities in AY23/24.
Related Information
Explore more about AI with these articles from QAA and Advance HE
QAA publishes response to Department for Education’s generative artificial intelligence consultation
Learn moreFind out more with a selection of Jisc articles
