What does ChatGPT feed on? – 02/02/2023 – Education

“— How do you rate the reliability of the information you find?

— As a machine learning-based language model, I don’t have the ability to independently assess the reliability of information. Rather, my answers are based on information present in the data I was trained with, without the ability to judge its accuracy or reliability. It is important to remember that, in some cases, the information contained in the data I was trained on may be out of date or incorrect. It’s always a good idea to check multiple sources of information and assess reliability before making any major decisions based on an answer provided by me or any other language model.”

In times of fake news, there is little care. And who warns, in the literal transcription presented above, is ChatGPT itself — a conversation robot trained from a wide database to generate original texts on the most diverse topics and in varied styles. The fame conquered in recent months is due, among other factors, to the sophistication of the narrative structure that it presents when provoked by any of us (the tool is available, until now, free of charge, at the address

There is no doubt that we are facing an important advance in the field of Artificial Intelligence (AI). And whenever technology surprises us (enchanting or frightening), discussions arise about how machines can come to replace human beings in more and more activities. ChatGPT has also raised concerns of fraud and plagiarism in the academic community (to the point where some schools and universities ban its use or announce plans for students to do more handwritten work).

Controversies aside, there is an essential role that continues (and will continue) in the hands of human beings: that of critically analyzing any information, reflecting on the origin and intention of each message we receive —whether it is produced by AI or by a person of flesh and bone. This is a debate that needs to accompany all technological innovation and, when we think about the digital education of the population (something that has been talked about so much recently), it is important not to restrict it to learning the technological tools themselves: we need to encompass the arsenal of essential strategies to deal with reflectively and responsibly with what we can create from these tools.

More than discussing whether students will leave tasks such as writing an essay to models like ChatGPT, we have to worry about their ability to analyze what was written, assessing reliability and weighing the consequences of what that message may entail .

When we look for information on the internet (through search engines like Google), we at least have access to the website or page responsible for the content found. Even so, in many cases, it is difficult to filter the reliable and quality sources amidst an almost endless list of links. Now imagine the complexity of analyzing information from different sources selected, grouped and hierarchized by an AI — it is not a simple task, but we have to develop the skills to perform it.

Also, you have to consider that the data that feeds the bot is not updated in real time. That is, more “hot” subjects and news that are still unfolding are not on your radar (for now, at least).

OpenAi (the company that created ChatGPT) started to inform, on the system’s initial page, that the dialog format “allows ChatGPT to answer subsequent questions, admit errors, challenge incorrect assumptions and reject inappropriate requests”. It will be important to monitor, in practice, how these actions will occur.

For now, we can take the advice of the robot itself who, when asked about fake news, wrote: “It is important to be careful with the information you read and share and always check the source and reliability of the information before making decisions or share them with others”.