A Simple Introduction to the Complexities of ChatGPT
Many things come to mind when the topic of workplace innovations arises. At the forefront, lately, seems to be Open AI’s ChatGPT software. Many myths surround this cutting-edge “bot,” including how it can be used and the potential pitfalls of its abilities. Understating the basics will ensure you know what it does, how it’s impacting your industry, and how you can utilize it.
What is ChatGPT?
ChatGPT is a free web-hosted software anyone can access, you only need to create an account. It was released by OpenAI on November 30, 2022, and is essentially an AI chatbot. Where it separates itself is the degree of deep-learning and interaction it provides. Ask any question—simple or complex—and ChatGPT sifts through a massive corpus of data to find relevant answers. It then constructs these in easy-to-digest responses that mimic human speech and can carry on answering subsequent questions conversationally.
All of this is achieved through a ground-breaking neural network architecture, called a transformer. This is the heart, mind, and soul of ChatGPT. Massive amounts of text data, called tokens, are fed into this neural network. These are taken from books, websites, articles, social media posts, and countless other sources. In its initial development and testing phase, ChatGPT was fed over 500 billion tokens, and its neural network has over 175 billion parameters built into it. These help the software consider variables related to a user’s question to then build what it deems is the best response. Ask anything, and watch in real time as the bot types out the answer.
What Else Can ChatGPT Do?
The degree of sophistication ChatGPT can achieve with its answers is what’s most jarring at first. It can answer complex math questions, perform language translations, provide solutions for computer coding or other tasks, or generate stories, poems, and blog articles based off simple user requests. While not 100% accurate, the results users receive in a matter of seconds can serve as a foundation for whatever they’re working on.
Many companies are now exploring these capabilities as full-time supplements, and in some cases, potential replacements of employees. It's a startling “glimpse” into a potential future and the reason we’re now seeing competitors like Google Bard, Einstein GPT, and Bing AI emerge.
What Are Current Limitations?
It’s important to remember that ChatGPT was just released last year and is still in its infancy. Since it’s drawing from existing data sources, there are concerns over partial or complete plagiarism. It’s such a concern that some sites are creating checkers to determine the likelihood that content was written by ChatGPT and similar services.
ChatGPT also lacks critical thinking skills, and being an AI software, it can’t detect moral, legal, or ethical issues with the content it’s creating. The software also can’t draw from real-time data, such as the weather forecast or the start time for a sporting event. This is because it can only draw from data existing before September 2021. However, it’s an ever-learning system that will only continue to grow and advance as more users interact with it.
Check back for part 2 of this series where we’ll discuss the ways ChatGPT is being used and will potentially be used in healthcare and related industries.