So get this, researchers have found that OpenAI’s ChatGPT is actually outperforming the average computer science student at universities. That’s right, this chatbot is straight up beating the students in subjects like political studies, engineering, and psychology. They published a paper in Scientific Reports and everything. But here’s the kicker, when it comes to math and economics, this fancy chatbot falls a bit short. Not exactly a human calculator, you feel me?
Now here’s what they did in this study, they asked faculty members at NYUAD to come up with 10 questions from their courses, and then they got three answers from actual students for each of those questions. Then, they fed those questions to ChatGPT and got three different answers for each question from the chatbot.
Now, here’s where it gets interesting. They compiled all the answers, both from the students and ChatGPT, into one document and had assessors grade them. And get this, my man, the assessors actually couldn’t tell which answers came from the chatbot and which ones came from real students. That’s pretty wild, right?
But get this, when it came to math-related questions and trick questions, the chatbot struggled. I guess humans still got the upper hand in those areas for now.
Oh, and here’s a wild find. They tried this thing called an obfuscation attack on the chatbot to see if it could fool the detection algorithms. And guess what? It worked! Around 95 percent of the answers from ChatGPT went undetected. That’s some sneaky stuff right there.
Now, they also did a survey in Brazil, India, Japan, the UK, and the US to see how people felt about using AI in education. And get this, 57 percent of students said they plan to use ChatGPT to help with their assignments. And even more, 64 percent of them expected their peers to do the same. But here’s the kicker, 69 percent of professors said they considered it plagiarism. So, there’s definitely a difference in opinions between the students and the professors on this one.
But here’s something cool. This chatbot levels the playing field for non-native English speakers. It’s like a helpful tool for those who might struggle with language barriers.
So, all in all, this ChatGPT thing is pretty impressive. It’s beating students left and right in a bunch of subjects, but when it comes to math, humans still got the edge. And hey, it’s even fooling detection algorithms like a pro. But you know what, let’s see how the education system adapts to this powerful technology. It’s definitely gonna stir some debate, that’s for sure. Keep on learnin’, folks!