AI or traditional software?
Consider the following scenario. You’re at a high-class restaurant that has two chefs. The first chef makes every dish by following tried-and-true recipes, step by step. The second chef follows his gut and creates unique and interesting meals, but sometimes it’s hard to know what to expect. Would you choose the first chef or the second?
With the first chef, you know exactly what you’re getting, and you know it will taste good—but his range is very limited, and the food can become repetitive. The second chef, however, is a little more interesting. You know that his range is much broader, and he can be more creative with his flavours and uniqueness. But at the same time, every meal presents a bit of a gamble, and it’s difficult to know if you’ll like what he makes every time.
With traditional software vs AI, it’s a similar comparison—only the stakes are much higher. Traditional software is the recipe-based chef in our example, and AI is the less predictable chef with a greater repertoire.
First, let’s look at how these work. Traditional software is built on rules. It’s about the computer following clear instructions written by a programmer. Every case, functionality, and decision the software makes has been planned in advance by the programmer. Each scenario needs to be meticulously considered, and everything is designed and tested to function in a consistent way.
Software programs of this kind are reliable and can be highly capable, but each new consideration or feature must be manually written by the programmer. So the software is only capable of doing what it has been made to do. As a result, writing many features can be expensive, and it is difficult to cover every case.
AI, however, is built on data. AI learns to understand patterns in the data it is given. It relies on statistics, which means it learns to predict patterns more and more accurately. However, this comes with a fundamental trade-off.
AI is not always 100% reliable or consistent because it makes statistical predictions, and its “thinking” is learned rather than programmed. It is much more flexible, as it can consider data it hasn’t been specifically trained on, as long as it follows similar patterns.
Take ChatGPT, for example. The data it is trained on is essentially the entire internet, which means it has a very broad understanding of patterns and skills, and is highly capable across many fields.
AI can be cheap—or very expensive—depending on the size of the models and the amount of data it needs to be trained on. Task-specific AI may only require a small amount of reference data and can be inexpensive to run. However, developing state-of-the-art intelligent models is extremely expensive, requiring many computers running simultaneously to train them.
When considering whether to use AI or traditional software, it’s important to weigh up how critical each of these four aspects is: cost, reliability, flexibility, and capability. Used correctly, AI can be an extremely powerful tool. However, in many contexts, the price and reliability of traditional software remain unbeatable.
As the world trends towards AI in the coming years, it will never truly replace traditional software. The task now is to decide what suits your business and projects best.
