According to Wikipedia, generative artificial intelligence is artificial intelligence capable of generating text, images, or other media in response to prompts. And, generative AI has potential applications across a wide range of industries, including art, writing, software development, product design, healthcare, finance, gaming, marketing, and fashion. (Wikipedia, 2023)
ChatGPT is a type of artificial intelligence model trained to generate coherent, human-like pieces of writing on a given topic. “ChatGPT was optimized for dialogue,” (ChatGPT FAQ | OpenAI Help Center) which means users interact with a chatbot in a conversational context, and the chatbot will compose text based upon the user’s prompts. The bot is capable of generating text on a variety of topics and in a variety of styles, “you simply type in a question or give it a command and it generates text for you.” (AUA Center for Teaching and Learning)
ChatGPT launched as a prototype to the public Nov. 30, 2022. Within five days, more than a million people were using it. (USA Today, Jan 2023)
ChatGPT (Generative Pre-trained Transformer) is the latest viral sensation out of San Francisco-based startup OpenAI. ChatGPT: A new AI assistant (advanced chatbot), has been released to the public (currently free) with enhanced capabilities. ChatGPT can be used to generate text on many topics, and can be customized/fine-tuned to produce text in a specific style or on a specific topic. – You simply type in a question or give it a command and it generates text for you. Available here: https://chat.openai.com/chat
Improved efficiency and accuracy in NLP-related tasks
Quickly and accurate responses to a wide range of questions
Assistance with a wide range of tasks that require the understanding and generation of natural language
What are the downsides of AI?
Information that sounds correct but is NOT: "If it is trained to be more cautious, it can end up declining questions that it can answer correctly"
Sensitive to slight tweaks in the input: The response from ChatGPT for instance, can be conflicting because phrasing can be slightly changed or the same prompt is made several times.
Bias is an issue: The ChatGPT model can be biased due to the training data. "It can also use certain phrases excessively". Bard within Google offers similar issues.
Does not ask clarifying questions: The current model does not follow up to the question. It will make guesses if the user gives it an ambiguous question.
Response to inappropriate or harmful requests: "While the model usually refuses to accept inappropriate requests, it does sometimes respond to harmful instructions. Typically, unsafe content is regulated by Moderation API, but from time to time, it demonstrates false negatives and positives. However, the good news is that appropriate user feedback is being taken into consideration to improve the current system."