ITB AI Expert Warns Against Using AI as Primary Reference

The ability of artificial intelligence (AI) to produce text, images, and videos resembling human work is expected to become increasingly indistinguishable in the future. Consequently, the public is urged not to make AI their primary reference and to remain critical of every piece of information it generates.

Ayu, an AI expert from the Bandung Institute of Technology (ITB), explained that the working mechanism of Large Language Models (LLM) like ChatGPT is called next token prediction, where each word is generated based on probability values. This process is often referred to as a stochastic parrot, akin to a parrot that can answer after being trained.

“Like a parrot taught to reply ‘waalaikumsalam’ when someone says ‘assalamualaikum’, the parrot does not understand that the utterance is a greeting or that it carries a reward,” Ayu told Republika on Tuesday (30/6/2026).

She explained that the parrot only knows to respond with a specific phrase upon hearing a trigger, without comprehending the content. Similarly, AI produces a series of words that are essentially probabilistic guesses, without understanding the meaning behind the sequence.

“This working mechanism means that AI (LLM) still has the potential to generate incorrect answers or responses. Therefore, according to UNESCO guidelines, the responsibility for using information produced by AI lies with the user,” Ayu clarified.

Ayu stressed that the primary party responsible for the use of information obtained from AI is the user. Thus, users are required to be critical of the output generated by AI. She reiterated that AI should not be used as a primary reference because it still has the potential to produce erroneous output.