From traditional teaching to teaching with artificial intelligence: how to take a stand against ChatGPT?

“Artificial intelligence: friend or threat?”. This was the title of the exclusive conference organized by Ticmas in the Alvear Palace Hotelattended by more than 80 school principals in Argentina. The panels discussed various approaches to thinking about the challenges and new possibilities that AI opens up in the field of education. In one of the activities of the day, Gustave Brey (ITBA professor), Mariano Yacovino (Director of Education of Microsoft Argentina), Mariano Ruani (Director of the Baikal Institute) spoke about the necessary — and inevitable — step from traditional education to education with artificial intelligence.

Coordinated by Astrid Herrerastrategic leader of ABP in Ticmas, the three experts took up the challenge of take a stand against ChatGPT: how can artificial intelligence improve teaching and learning, what are the ethical and privacy challenges, and what is the role of teachers, administrators and families.

“Thinking that ChatGPT is going to replace teachers or make kids copy each other won’t work,” Brey said. “You can no longer think that the calculator Is this a math problem or what Google there Wikipedia They are the terror of teachers”. Brey said his experience with ChatGPT led him to improve his exams because while he believed his exams had a degree of excellence that made students think and connect the answers, the artificial intelligence approved. “It raised the bar for me as a teacher,” he said. “Today artificial intelligence allows me to optimize many educational issues and it gives me more time for creativity and to maintain the emotional part with the students”.

“The discussion we are having here today,” Yacovino said, “is taking place in any educational entity in the world. I believe that Artificial intelligence is a super powerful tool. This will not only help us save lesson preparation time and generate more interactive learning experiences, but it will also help tailor education to individual needs. This tool will help us create more interactive and personalized courses. How will we adapt? We’ll learn along the way.”

Mariano Ruani also pointed out that AI offers the possibility of Be more creative, because we can help ourselves to think of more entertaining activities and consult other experiences. But it is clear that ChatGPT must be accompanied by a teacher. “There are some risks that you need to be aware of,” he said, citing, among other things, bias, lack of contextualization and “hallucination” in responses – a misconception that requires critical reception to be recognized as such. “A lot of people think boys are going to cheat on us,” he said, “and for me, that’s how we use it to enhance learning, so that children are better, more creative and learn better. Then we will see how we evaluate them”.

Currently, there are two main positions at the level of education: there are institutions that prohibit its use and there are other approaches in which its use is encouraged, almost forced. “These teachers recognize that students are much better off than if they don’t use itRuani said, “and, moreover, it will be in the whole society: not learning to use it is depriving boys of a chance”.

The limitations of ChatGPT are in the subtleties. Yacovino asked to take the test and asked him, for example, how many bananas can you eat on an empty stomach. The AI ​​replied that there was no fixed number, although the question was tricky and the answer was clear: only one. After this banana, there is no more fasting.

“We see computing as something deterministic and it’s probabilisticRuani explained. “All the answers have a certain probability of being true. For that ChatGPT passes an exam, but it does so with eight, never with ten. The problem is that when we talk to a computer, we start to believe it. If the first time is good, the second time too, and on the fifth time you don’t check anymore and that’s when he tells you nonsense. The problem is that when he is wrong, he can be very wrong.”

“AI is more about looking for patterns in data and classifying them. If the model is poorly trained, it can have errors,” said Brey, who also pointed out that the technology is not yet capable of feed Google. “But,” he said, “it’s the tip of the iceberg”.

READ MORE