Long read: An Intuitive Guide to How LLMs Work

Whilst Artificial Intelligence (AI) has been around for a while, it has become ubiquitous over the past two years since the launch of ChatGPT. It is no longer just the realm of tech companies finding niche applications working in the background. With Meta’s AI embedded into Whatsapp chats, Google’s into Android phones and Microsoft’s into its Office suite, AI has become a part of the common man’s daily life. What changed? The specific technology behind Generative AI, the likes of ChatGPT is known as Large Language Model (LLM). Whilst most of us use the internet without knowing exactly how it works, knowing the science behind LLM, atleast in layperson terms, is likely to help us extract better from this technology. This blog does a great job of explaining it in simple English or rather Math, without getting into the complexities.

“…at their core, LLMs are nothing more than fancy probability engines.

Probability is a tool for quantifying randomness and uncertainty. Having a probabilistic nature is the source of an LLM’s power… and its unpredictability. It’s what makes it possible to generate novel, creative, actionable responses, and also makes LLMs very difficult to train or debug.

Whether you’re a layperson, practitioner, or researcher, the entire process of working with LLMs is an exercise in forming and manipulating their latent probability distributions into giving you the outputs you want.”

The blog gives a quick refresher on probability distributions, something that most of us would have studied in high school math, before going on to explain how LLM’s use probability of the occurrence of a word to generate text which is likely to make sense and give us the outputs we seek. Given this context, unsurprisingly, the most common application of LLM so far has been Chatbots. He gives an intuitive sense of how LLMs drive chat:

“Chat interfaces have become the dominant way for most people to interact with LLMs, capturing the public imagination and showcasing these models’ capabilities. But how do we get from generating individual words to engaging in full-fledged conversations? The answer lies in cleverly applying the principles we’ve discussed so far.

Here’s how it works:

This approach allows the model to maintain consistency and context throughout a conversation. It can refer back to earlier parts of the chat, answer follow-up questions, and generally behave in a way that feels more like a coherent dialogue than isolated text generation.

However, this method also introduces some challenges:

1. Context Length Limits: LLMs have a maximum amount of text they can process at once (often referred to as the “context window”). For very long conversations, the earliest parts might get cut off when this limit is reached.

Despite these challenges, this simple yet effective approach to chat is what powers the conversational AI interfaces we interact with daily. By treating the entire conversation as a growing context for probabilistic text generation, LLMs can engage in surprisingly coherent and context-aware dialogues.”

Whether you are a user of AI or an analyst trying to assess the disruptive potential of this technology, the blog is a compelling and easy read to understand what’s tipped to be the next big thing, as seen with the tens of billions of dollars already being invested by the world’s biggest technology companies.