Paul Berkovic is an expert in machine learning and its applications. In this article, he unpacks the differences between artificial intelligence and machine learning, and explains why businesses shouldn’t just rush to the newest and shiniest thing.
As a small-business owner, it can be understandably hard to wrap your head around the nuances of new technologies. This is particularly true when it comes to technologies that are overhyped – see Artificial Intelligence (AI).
Whilst most of us are becoming familiar with AI in our day-to-day lives – from Siri and ChatGPT to Microsoft 365 Copilot – when it comes to implementing it across our organisations, it’s critical to understand what AI is, how it intersects with other concepts, and whether it’s actually going to help you meet your goals. This clarity will help you avoid confusion and support you to make the right investment decisions for your business.
Differences and misconceptions
AI and Machine Learning (ML) are two terms that are often used interchangeably. ML is a subset of AI and, actually, when we’re generally discussing AI today, we’re really talking about Generative AI (GenAI) – another subset of the broader AI field.
GenAI mimics how humans behave via a large language model (LLM) that’s been trained on all the data available on the internet to deliver responses, whilst ML involves using specifically-designed algorithms for a particular purpose that are directed to analyse specific data sets. GenAI is (in nearly all circumstances) a generalist, whereas ML is used to draw specific statistical insights and predict outcomes from curated data sets.
The aim of GenAI is to use LLMs to solve a multitude of complex problems, whereas ML is used to learn from specific data to predict and/or improve the performance of a specific task. For example, you can ask ChatGPT to give you a meal plan for quick healthy dinners or plan a trip to Lisbon; it uses logical, human-like thinking to provide answers that fit the question.
This is in contrast to how Google Maps uses ML to analyse real-time traffic conditions to provide the best driving route and predict an estimated arrival time. It has learnt from millions of data points it has collected over time to make that specific prediction – ask for a Carbonara recipe, and it wouldn’t know where to start.
You may also have heard of the concept of ‘hallucinations’ within AI, where models will make up plausible answers to questions with no evidence to support them. This only applies to GenAI – ML is based on hard and fast data. This doesn’t mean ML predictions will be correct, but it will learn from deviations and correct future forecasts in line with them.
Avoiding the rabbit hole
The reality is that the pervasiveness of AI in business has been incredibly overhyped. A survey from Riverbed found that in the retail industry, 95 per cent of leaders consider AI a top priority at the C-Suite level, yet only 40 per cent feel fully prepared to implement AI initiatives presently. There are also reports of many high-profile AI projects being cancelled, with more than 80 per cent of AI projects failing.
These findings indicate that organisations are rushing into AI without really knowing where to start, how to implement it effectively and the tangible benefits it will bring to day-to-day operations. For example, to improve an inventory management strategy, ML-based systems can help analyse sales data in order to predict future sales trends based on a business’ own data (not general sets) that allow them to automatically adjust stock levels. This problem doesn’t need huge investments into complex GenAI technologies because, in all likelihood, they aren’t going to deliver the benefits sought.
Organisations shouldn’t simply jump on the newest and shiniest thing. It’s important to understand AI technologies and create a strategy based on realistic outcomes, rather than lofty promises, and make investments accordingly.
