AI Evolution, From Simple Data Collection To LLM Development

Development of AI from time to time (photo: dock. unsplash)

JAKARTA Technology encourages rapid data developments. As data accumulates, big data is formed. This is considered an introduction to the world based on Artificial Intelligence (AI).

CEO said.ai, Irfan Raditya, calls it the AI Era. This era is marked by the emergence of various AI-based chatbots. Not only ChatGPT, but also Gemini from Google, Claude from Anthropic, Copilot from Microsoft, and many more.

AI technology has been around for a long time, but the naming has been different in the past. In the early days of computing, programmers were more familiar with the Object Oriented Programming (OOP) method.

The method with this approach is very limited because the computer can only recognize objects based on rigid specifications and rules. The reason is, computers are only taught to study one specific object.

"Simply put, we only teach one object computer. Consider the object a cat and we make specifications like two ears, four legs, four hair, and like to play ball. If there is a cat that is not hairy, the computer will say it's not a cat," explained Irfan at the SMBC TechConnect 2025 event on Monday, November 24.

Around the 2010s, the world began to enter the deep learning era which is a derivative of machine learning. With deep learning, machines can learn through data, predict, and read various formats such as text, images, audio, and videos.

The main advantage of deep learning is the accuracy and very high precision in recognition. However, the drawback of this method is the limitedness of data. This method requires a lot of data so that the system can only recognize the objects that have been taught.

Along with increasingly sophisticated technology, the Large Language Model (LLM) or Large Language Model (LLM) method also emerged. LLM takes big data from various sources, ranging from official websites such as Wikipedia, social media, forums, to e-commerce.

With this training, computers are becoming more sophisticated. Users can interact directly with AI creatively and more flexibly. In fact, Irfan said that AI can 'explain cats, but in the form of Taylor Swift's song.

AI can be found in simple to complex features. For example, this feature is integrated into email filters and engine recommendations on social media. This machine will present content that users are interested in even though they have never seen or followed the account.

At the highest level, for example, like OpenAI's ChatGPT-4 model, AI has an intelligence level that is equivalent to 76 percent of the PhD level. This shows that AI can penetrate college entrance exams to become a doctor or lawyer.