Generative AI vs. Traditional AI: A Beginner’s Guide to Understanding the Differences

Traditional AI as a Specialist

Traditional AI is a specialist who is trained to perform a specific task based on algorithms programmed by human experts. For instance, imagine it as a detective who is excellent at identifying fraudulent and suspicious emails. In general, traditional AI is capable of reasoning, inferring, and predicting data and scenarios.

However, whenever new challenges arise, the company needs to find another specialist to address problems. This requires investing significant resources to collect data and train a new AI model which is very time-consuming. (According to IBM’s research, companies are spending 59% of their AI budget on collecting training data.) Traditional AI is efficient but limited in its adaptability to diverse scenarios and developing new AI specialists is costly.

Generative AI as a Generalist

In contrast, Generative AI is a generalist with a diverse skill set. It’s built on a Large Language Model (LLM) training on a vast amount of data scraped from websites such as Wikipedia and Reddit. It’s capable of generating all kinds of texts and codes, summarising documents and even analysing the sentiment of the context.

The problem with Generative AI is even though it can conduct various tasks with creativity, it often lacks depth in enterprises’ secret sauces. For example, if the enterprise wants to create product roadmaps based on its 5-year company strategies, often we need to feed it with more data that are directly relevant to the problems to enable it to create a better and more viable outcome.

Empowering Generative AI: From Generalist to T-Shaped Talent

Many businesses seek T-shaped talents who have both horizontal transferable skills and vertical specialised expertise. For instance, they understand general industry know-how and communication skills as well as deep expertise in specific areas such as the company’s procurement process or front-end engineering.

In the context of AI, T-shaped talents represent systems offering both a wide range of generative capabilities and in-depth proficiency in specific domains. Various techniques can help us transform Generative AI from a generalist into a T-shaped talent. For instance, methods such as prompt engineering, fine-tuning, and techniques like RAG (Retrieval Augmented Generation) can ensure that the outcomes are tailored to businesses’ unique requirements and/or challenges without building an AI model from scratch. This saves tremendous time and effort and most importantly it doesn’t require as many AI experts as traditional AI development does.

(Please also note that traditional AI has quite different capabilities than Generative AI so it’s not saying that Generative AI is going to replace traditional AI.)