Mar 134 min read

ChatGPT: Where is it Taking Us?

Audience: Middle and High School Students

Fun fact: ChatGPT gained a million users in five days. Photo by Clément Hélardot from Unsplash.

Having a full philosophical conversation. Writing lines and lines of code. Composing poetry and creating stunningly vivid art. These are things only humans can do, right? Think again. Although it may seem like something out of a sci-fi novel, technology has developed to the point that chatbots like ChatGPT are able to take human creativity and synthesize it.

These are all outputs of ChatGPT. But what is it really? ChatGPT is a chatbot. It simulates conversation with users and generates amazingly humanlike responses. It processes user input using natural language processing, which is exactly what it sounds like. It takes in natural, human-spoken language (not computer language) and processes it by using patterns it has learned. That’s just the surface-level description of the popular application.

To be more specific, ChatGPT is a large language model, or LLM. LLMs are huge AI networks that are trained on massive amounts of text taken from the web. They take a block of text from a source — for example, Wikipedia — and process the initial portion of it. Then, they try to predict what comes next in the sequence. Finally, they compare their answer against the actual source. By training themselves with large amounts of data, they become very accurate. It’s a bit like doing math exercises in a textbook. You look at the problems, write down your answer, and then flip to the back of the book and correct your answer using the official key.

What kind of outputs are possible with ChatGPT? What can LLMs do? Some of the most popular uses of ChatGPT include writing articles and emails, designing websites, writing code, producing creative works, and summarizing content. It’s more than a search engine. It can pull information from the web into one cohesive block of text, which is helpful for simplifying difficult topics. ChatGPT also has powerful data analysis and collection capabilities which are much lesser-known and underused. It has uses that range from art and writing to math and science.

To increase their accuracy, ChatGPT and other LLMs use reinforcement learning from human feedback, or RLHF, a process in which humans create prompts and rate ChatGPT’s responses to those prompts. They write two-sided conversations (one side for the user, and one side for the AI) and rate how well ChatGPT’s response matched the written response. Over time, ChatGPT learns to recognize context clues and respond more and more accurately.

Although ChatGPT has been evolving to replicate human creativity, its abilities are limited. It has trouble detecting subtle hints of emotion or sarcasm in user input. It lacks common sense, and its knowledge is limited to the information that’s on the web. Moreover, it doesn’t always sound or respond like a human; instead, it’s very obviously stilted and takes a while to process and respond, very unlike a human’s lightning-quick repartee.

Aside from its lack of humanlike responses, another major concern about ChatGPT is its accuracy. Despite being trained in many different ways, it’s still a computer program and not a human. The weakness is heightened when users ask it to produce lengthy content or to perform multiple tasks at a time. Grammar issues and small mistakes are apparent and even basic facts can be wrong.

Additionally, ChatGPT is not completely updated on current happenings. For instance, if you asked it to produce sports statistics from an ongoing event, it would have some trouble. LLMs aren’t trained on very current information, since training is very time consuming and costs millions of dollars. ChatGPT may also provide wrong answers if it doesn’t know the answer. Users may not know that the information is incorrect, which is a concern. ChatGPT does not always flag users when it’s unsure of the answer.

Students have faced issues with using ChatGPT for this very reason. They may have used it to create a last-minute essay or short story and ultimately have been called out for incorrect information or cheating. Many teachers are not fond of ChatGPT and regard it as plagiarism. It’s been an increasing concern for schools, since the faculty are disturbed by the idea that students may be passing off AI-generated text as their own.

Schools are not the only place where ChatGPT affects the world. It has impacted virtually every major industry in a variety of ways. It plays a key part in robotic assembly for manufacturing. And it’s emerging in healthcare, with disease identification and acceleration of medical discoveries. It’s clear that ChatGPT and other AI are going to play a huge part in our future and change our lives. Employees - such as the factory workers who are being replaced - will lose their jobs and need to seek employment elsewhere. However, new jobs in tech will also be created. More aspects of life will turn digital, such as the adoption of driverless cars. Privacy will become more rare in an environment where Alexas and Siris are always turned on and ready to listen to sensitive and personal information.

AI is what we make it. It’s a valuable tool that can make life a lot more efficient and easy. It can even save lives. However, we need to set boundaries to limit its usage. If it’s not carefully controlled, it can spread misinformation. It can also be used to plagiarize or invade privacy. It can dramatically change the way we conduct our daily lives. ChatGPT is not a good thing nor a bad thing. It is simply a tool. Are you ready to use it?