In recent days, a novel language model by the name of ChatGPT went viral throughout the world. Developed by the OpenAI organization and initially released in November 2022, ChatGPT is a language model driven by artificial intelligence that automatically feeds users with requested information such as email templates, cuisine recipes, scientific data to serve as study aids for school examinations, and many more. ChatGPT can basically be used to achieve anything. Structurally, ChatGPT is a generative model, meaning that it generates new data based on any data input into the model. It is trained with supervised learning, the process of modeling from pre-labeled training data, and reinforcement learning, a process in which AI models improve by being ‘rewarded’ or by taking feedback after output). ChatGPT is also trained using Transfer Learning, a type of machine learning that involves training models with data from one domain and testing those same models with data from another domain while trying to achieve the same goal from before.
There so far have been mixed reactions towards ChatGPT, where many people are impressed by its performance, while others believe that it disrupts creative freedom and negatively impacts education. While ChatGPT is meant to be used for facilitating simple tasks such as email templates or others, ChatGPT has been abused to a huge extent. Right now, thousands of people, both students and adults, rely on ChatGPT to perform their tasks, whether that be homework, essays, developer code, automatic responses for customer services that companies offer, and many more. Hence, it is preferable to limit the use of ChatGPT enough that people don’t view it as their first solution to any problem they have to resolve. Personally, when I have tested ChatGPT to do things such as provide practice questions for a test or a study guide, I felt the urge of relying more on ChatGPT the more I started to use it. While ChatGPT is a phenomenal resource for simply obtaining useful information, it is not something that can supersede every task that people do. Additionally, ChatGPT is not mature enough (yet) to provide highly accurate information every time a query is made by a user. Training such language models to high levels of accuracy requires lots of data as well as models that are complex enough to digest a multitude of features from a given set of data. Nevertheless, it will be interesting how language models like ChatGPT shape our future, as with time, language models will increase in accuracy. Perhaps, global tech giants such as Google, Apple, and Microsoft will create their own powerful language models to overtake ChatGPT itself!
“Artificial Intelligence & AI & Machine Learning” by mikemacmarketing is licensed under CC BY 2.0.
The Roar 