January 25, 2023
Artificial intelligence is a field which has many practical applications in today’s worlds. Simplifying research, saving time, generating better predictive models; the list is almost endless. The recently released ChatGPT is making waves in this field, specifically with its powerful natural language processing (NLP) capabilities. It’s able to generate incredibly convincing responses to queries input by a user, enabling it to tell a story, explain a topic or naturally engage with a human in conversation. How does this new fangled technology work then, and what does it mean for those of us who use the written word as part of our daily grind?
ChatGPT is a large language model developed by OpenAI. It is based on the GPT (Generative Pre-training Transformer) arch tecture, which was introduced in a 2017 paper by researchers at OpenAI. The model is trained on a massive amount of text data from books, articles and websites, allowing it to generate human-like responses on a wide variety of topics. This allows it to generate text that is similar in style and content to the text it was trained on. The model can then be fine-tuned on specific tasks, such as answering questions, writing essays, and generating creative text. One of the key features of ChatGPT is its ability to generate text that is coherent and fluent. This is achieved through the use of a transformer architecture, which allows the model to attend to different parts of the input text when generating each word.
The model is also able to generate text that is appropriate for a given context. For example, if the input text is a question, the model will generate a response that is appropriate for that question. This is achieved through the use of a large number of parameters, which allow the model to learn patterns and relationships between different words and phrases. In addition to its text generation capabilities, ChatGPT can also be used for other natural language processing tasks, such as language translation, text summarization, and sentiment analysis. The model’s large number of parameters and ability to maintain context make it well suited for these tasks.
However, like any AI model ChatGPT does suffer from some limitations. One of the main limitations is that it can sometimes generate text that is not entirely coherent or grammatically correct. This is because the model is not perfect, and it can make mistakes when generating text. Additionally, because the model is trained on internet text, it may replicate biases present in the training data. Even with these limitations, its ability to generate coherent and fluent text makes it well suited for a huge array of tasks and applications. Including, as I’ve learned, writing columns for a local paper.
For the sake of research, I actually engaged ChatGPT for help writing this very article. While I had to form the questions myself and clean up some of the language, it flows remarkably well for something generated by an AI. Upon sharing this piece with a couple of friends, they (scarily) couldn’t tell which parts are my own words, and which were the parts my new robot assistant penned. ChatGPT is currently free for anyone to try, although this may change in the future due to the cost of the vast processing power required to run it. Off I go to ponder how easily, in this instance, I could be replaced by a soulless robot. Wordsmiths beware, I suppose.