ChatGPT is a language model capable of a multitude of tasks including answering questions, compiling lists, generating paragraphs of text and writing computer codes across multiple computing languages. Dr. Alan Yang, an assistant professor of information systems in the College of Business, provides insight to how the model works and what users should know before using it.
How does ChatGPT work?
GPT stands for “Generated Pre-Trained Transformer”, which describes the main features of the agent. ChatGPT 3.5 has been ‘pre-trained’ on information pulled from a comprehensive scrape of the web from 2020 to 2021. ‘Transformer’ describes the set of rules, or algorithm, that is used to generate responses to questions.
The algorithm ChatGPT uses is based on the prioritization of words within a prompt. For example, if you were asked to finish a sentence such as “I am feeling…” you may use words such as “happy” or “sad”. When you add an additional phrase to the sentence, “I just drank a cup of coffee, I am feeling…” you may be more prone to use words such as “energized”, “alert”, or “awake.” We can finish a sentence in this manner because of our ability to process the context of a conversation. What makes ChatGPT humanlike is that it can also determine context and adapt a response based on the combination of words put together in addition to prioritizing the usage of certain words and phrases commonly used in conversation.
Is ChatGPT ethical?
Attribution of work is a big concern when using ChatGPT. ChatGPT generates information based on its transformer algorithm, which creates an answer based on an interpretation of a question and the information that it was trained on. The most likely answer generated will be the most popular one based on the algorithm’s interpretation. As a result, ChatGPT’s responses are even more difficult to attribute to a single reference as the answer is a combination of many different sources. The issue of misattribution can lead to more ethical problems, such as misinformation.
Will ChatGPT take my job?
ChatGPT can provide a best-guess answer to a prompt, but it has limitations. Multiple industries will be affected by ChatGPT, but the need for individuals capable of critical thinking and problem-solving is still going to remain high.
For those concerned, I would recommend taking a historical perspective on technological development. The device that you are reading this article on can perform more calculations at a pace that you cannot match with your brain alone. This has been the case for decades now, but it does not mean that human brains are obsolete.
Is ChatGPT reliable?
From my experience, ChatGPT excels in answering direct questions, creating lists and writing summaries. ChatGPT’s ability to write and evaluate code across multiple languages is also very impressive. The algorithm is less consistent when asked to respond to a more open-ended question, provide substantive analysis or answer questions on niche topics.