AI is Not Always Correct: Why Students Need to Double-Check Their Answers
Artificial Intelligence (AI) is becoming increasingly easy for students to use to assist various needs, including the learning process. However, this ease of use does not mean that every answer provided by AI can be considered correct.
Rizki Putra Prastim, a lecturer in the Robotics and Artificial Intelligence Study Programme at the Faculty of Advanced and Multidisciplinary Technology, Universitas Airlangga, explained that AI essentially works by learning patterns from data provided by humans. This data can take the form of text, images, or sound. The machine then learns patterns from this data numerically to be applied when encountering new data.
“Much like a child receiving information or knowledge from parents, artificial intelligence works similarly. Humans provide vast amounts of data, which can be images, text, or sound. The machine will learn the patterns of that data numerically,” explained Rizki.
He provided a simple illustration through object recognition. When a computer is trained with various images of chairs, the machine learns the characteristics and patterns present in that object. When presented with an image of another chair, the computer can then recognise it based on the similarity of the patterns it has learned.
According to him, this process fundamentally involves many mathematical equations. Because it operates based on patterns from learned data, AI still possesses the possibility of producing errors.
Another lecturer, Tio, emphasised that scientifically, no system has a zero per cent error rate. AI is no exception, as the process of formation and training involves data and human decisions that can also contain errors or bias.
“If every forming factor contains an error, then the final product will also contain an error,” he stated.
In AI used as assistants, such as ChatGPT and similar services, errors can also arise when the machine fails to understand the instructions given by the user accurately. As a result, the output generated may differ from what was expected.
Tio had even tested several AI tools by providing mid-term exam questions for analysis. The majority of the AI tools he used provided incorrect answers. When informed that the answers were wrong, the machines re-analysed the problem but still produced incorrect results. It was only after Tio requested the machine to use a different method for analysing the question that the correct answer was obtained.
This experience demonstrates the importance of maintaining a critical attitude when using AI, especially in the educational process.
Students today have access to various AI services, ranging from ChatGPT, Gemini, and DeepSeek to AI technology embedded in other software, including photo editing applications. However, this convenience also brings risks if users accept AI results without performing a re-check.
When someone is learning something entirely new, for instance, they may lack the knowledge to determine whether an AI’s answer is correct or incorrect. If an erroneous answer is immediately accepted, a misunderstanding can form and subsequently be regarded as correct knowledge.
Therefore, AI is more appropriately used as a tool to support the learning and working process, rather than as the sole source of truth.
The development of AI itself is not a entirely new phenomenon; development of this technology began as far back as the 1950s. However, the limitations of computer technology at that time meant its progress was not as rapid as it is now. Advancements in computing power and supporting technologies have allowed AI to evolve much faster and integrate into various human activities.
With increasingly widespread use, ethical issues also become an important consideration. AI can provide great benefits if used correctly, but it still requires human supervision and judgement. Therefore, students and the general public need to understand that AI is a tool that can assist work, but the results must still be verified. A critical attitude is essential so that the convenience offered by AI does not lead users to accept erroneous information as truth.