Why is AI a big but poorly understood topic?

Artificial intelligence (AI) has become one of the most talked concepts in business and technology today, but understanding how these tools can add value is less well understood.  Algorithms such as large language models and automating processes have enormous potential to support small business, particularly when looking to scale and increase efficiency.  That doesn’t mean that these tools are AI. 

At the simplest level, AI refers to algorithms that can learn and adapt through experiences – that is they can mimic human thinking, reasoning and learning. Techniques under this broad banner cover areas such as visual perception, speech recognition and language translation.  In this sense, many of us now use AI – but what we are actually using are large language models (think ChatGPT and Claude.ai) that have been trained on available data sets.  These are vastly different to artificial general intelligence (AGI) which can reason and learn like a human being. 

In essence what is available to everyone who can access a computer (or indeed a smart phone) is a narrow version of AI – algorithms that have been trained to undertake well-defined tasks.  This is important to remember because it explains why early versions of ChatGPT would have hallucinations: the model could only respond based on what it has been trained on.  So whilst large language models seem to be intelligent, it is because they are operating within their designed functionality.  The algorithm can’t transfer that knowledge to other contexts.   

So when the term AI is casually used, in what seems like almost every walk of life and every area of business, it is important to understand the nuances.  Current AI systems are impressive but focused on narrow applications.  In truth they are not AI, rather machine learning and deep learning algorithms.  Actually, shock horror, some of them are just old fashioned, really good statistical and maths modelling.  The algorithms don’t possess generalised intelligence, can’t operate independently and need human intervention.  You needn’t worry just yet about AI taking over the world. 

That said, using and leveraging the power of machine learning, deep learning, statistics and automation can drive operational efficiencies, provide previously unseen insights from data, and help improve customer experiences.  They are incredibly powerful tools to help businesses scale and improve the lives for everyone in society.  However, we need to be open and honest about the algorithms limitations.  It is not as simple as “just using AI tools”  When implementing them into business process all involved in developing and using the approaches need to cover the full breadth of AI which includes the ethics of using the tools and possible bias within data and results.   

Instead of getting caught up in the AI hype, why not focus on a more measured approach.  Identify if a particular practical approach shows value to your business process and then spend time learning about its recent applications, understanding its limitations, identifying where bias might enter the results and testing if the outcomes add value to decision making and customers.  Then implement – but be open and transparent if its AI, machine learning, deep learning or statistical modelling because maths, not magic, should positively impact every life.   

By Dr Sophie Carr

Leave a comment

Your email address will not be published. Required fields are marked *