Developed by US company OpenAI, ChatGPT has taken the internet by storm, winning 100 million users since it came out in November 2022. People can ask the robot to write stories and emails, create instructions for cooking a certain food, translate languages, and answer all kinds of questions. In its own words, it is "a language model trained in a large amount of internet text to help users get human-like text."
Compared with Siri or other chatbots, ChatGPT uses a much bigger information center for training. It also uses stronger software (软件) and hardware to learn things by itself. For example, if it provides a wrong answer to your question, you can tell it the right one and it will correct itself. "It's a totally different chatbot," computer scientist Liu Xiaoguang from Nankai University told Tianjin Daily. "The knowledge level ChatGPT shows is the same as a university student. That's why it shocked the world."
But one big problem with ChatGPT is that it makes mistakes or even gives false information. When Rezza, a 28-year-old from Indonesia, used the robot to write an essay (论文), it "gave out many examples which other writers actually hadn't mentioned at all", he told The Guardian. Since the robot is trained using words from the internet, it can also pick up biases (偏见) about certain groups. These are all things that need to be dealt with.
