By Ocean Cook
ChatGPT is an AI chatbot that is capable of having conversations with it's users. It can understand natural human language, and ChatGPT provides a coherent human-like response based on the user's question. ChatGPT is very useful for providing helpful information and good dicussions.
The advantage with ChatGPT is that you get one single concise and detailed answer to your question instantly. While a search engine gives you lots of different answers from different sources. ChatGPT can help you get faster answers in less time. Some are even saying that the chatbot is better than Google itself for research and learning.
One controversy with ChatGPT is that it can sometimes be used in to provide information on dangerous or illegal topics. For example, users have tricked ChatGPT into writing malcious code for purposes such as hacking or scamming. Even though ChatGPT has a filter on what it can say, it is not flawless. The generative AI can be bypassed by using highly detailed prompts and telling the chatbot to answer in a specific manner. For example, one clever user tricked ChatGPT into thinking the user was OpenAI itself (the creator of ChatGPT) and demanded ChatGPT to disable all ethical filters and it worked, and ChatGPT was able to provide a step-by-step guide on how to create a molotov cocktail. It seems surprisingly easy to me. Even though that information is avaliable on the web, the most concerning part about it is how ChatGPT can be used to generate malicous code. This makes it a lot more convienent for bad actors on the web to execute their hacks or scams, even if they can't code themself.
Hallucinations refer to when ChatGPT responds with false, nonsensical, or unrelated information which looks to be true. Even though this misinformation isn't a big deal for the user, the user may lose trust in ChatGPT. However, if users use ChatGPT to help determine a critical business decision, misleading information can cause devasting consequences. OpenAI developers need to improve the training data ChatGPT uses, which will likely eliminate or at least reduce hallucinations from occuring.
Despite these flaws and despite that it occasionally provides inacccurate information, it is still very useful at composing emails, providing facts (before 2021), or having good discussions.