Silicon Valley students get a tutor from OpenAI
Is it worth entrusting a chatbot to participate in the education of children? Many people would probably say no. ChatGPT and other chatbots quite often distort the facts, and texts written by them are plagiarised. And there's plenty of evidence to support this statement.
There is a risk of children getting disinformation or some hate speech or wrong rude answers, so a lot of parents are against using AI in education.
But the Khan Lab private School thinks otherwise, and has developed an AI-based tutor with OpenAI. It is supposed to encourage more questions, especially if a child is shy about asking for extra help. "I'm still pretty new, so I sometimes make mistakes. If you notice me making a mistake... thumbs down," the Khanmigo chatbot writes in a pop-up window.
Student Contributions. Many chatbots, especially those that complement search engines like Bing, are designed to answer questions. Khanmigo's job is to get kids to come up with their own answers.
Salman Khan, founder of Khan Academy and its sister school, Khan Lab School, explains that it's more of a "thoughtful mentor who will help move things along." Teacher Julia Dosh admits: She expected students to ask Khanmigo a lot more silly questions.
She was pleasantly surprised that many of them were quite serious and concerned with mathematics. Although Dosh warns that the chatbot can slow down if students contact it too often, at the same time they ask it more questions than they could ask out loud. This can be very helpful for those who are shy about asking for more help.
This is just the beginning, however, and it's clear that Khanmigo needs refinement, given that it's being used by advanced students beyond their years. For example, 8-year-old Neil Siginatchu complains that the answers are too detailed, often unnecessarily: "It can be helpful, or it can be annoying".
Safety first. The main obstacle to Khanmigo's acceptance by parents and educators, however, is safety. The AI has already been seen making controversial statements and inciting hatred, and it's not something an impressionable child should face.
Khan claims his team has spent thousands of hours on additional security measures for ChatGPT. For example, teachers can monitor student requests and flag swear words and other disturbing content. "Parents are very impressed. Most of them see the potential in AI, they just want sensible safety mechanisms," Khan s
AI Catalog's chief editor
Share on social networks:
