Why We Shouldn’t Use AI in Learning?

Nowadays, everyone is talking a lot about Chat GPT, MidJourney, and various AIs, with many believing that AI can analyze accurately or perform tasks better than humans. However, in reality, there are some tasks that we should let AI handle, but there are still many tasks, especially related to learning, that we should not let AI do. The reasons why we shouldn’t use AI in learning are:

Reasoning != Probability of AI

Even though AI can analyze well, it relies on probabilities. Therefore, the information obtained can be both correct and incorrect, leading to what is known as AI Hallucination.

Imprint Problem

When we learn new things, we tend to believe that the information we receive is more than the information we will get in the future. This is similar to ducklings following the first thing they see. So if we get incorrect information from AI, it can cause problems in the future.

Bias in AI Training

Before we have an AI model, we need data, that can be biased, such as gender or racial discrimination. If we are unsure that our data is truly effective, we should not be overly confident in the model.

Black Box AI

Lastly, if we use black box AI, whose workings we do not understand, as part of learning, and it turns out to be wrong, there may not be anyone to correct it. The incorrect information might be passed on to others. An example we encountered directly was an application that used Gen AI to create storybooks for children. At first glance, it seemed okay, but upon reading in detail, it was horrifying. The app had to be deleted quickly because a child shouted out loud in the bedroom, “Fish, poop, spinning around.”

References:

https://www.science.org/doi/abs/10.1126/science.adi0248
https://thailand.kinokuniya.com/bw/9780241586488

If you want to know in which situations AI should be used, feel free to contact us via the contact form here [Contact].