ChatGPT, Is This The New Magic 8 Ball?

ChatGPT

The IT girl of the year is hands down ChatGPT gaining over a million users within the first 5 days of its launch which is impressive knowing it usually takes months to years to reach those numbers. By the second month, it had over 100 million users with millions of visits daily since January of 2023. Open AI, the backbone of ChatGPT, is now valued at over $29 billion after the launch of ChatGPT. Not all of the millions of users know exactly what ChatGPT is, just like how I’m not sure about the mechanics of a plane but will still book a flight, it’s for convenience. But unlike these flight prices, ChatGPT is free for all.

Simply create an account and use its “Send a message” prompt to ask a question like “How many deaths occur due to Hippo attacks?” and see what answer you will receive. Even if you misspell some words, it can somehow still properly predict what you mean. Being able to ask any question and have a detailed response catered to any audience of your choosing is what attracts most to ChatGPT. ChatGPT is basically what people thought Siri would be. Even though there are no more “I’m not sure what you mean” responses like Siri, there are still some shortcomings.

Not Perfect But Worth It?

One shortcoming is that it does not have access to real-time information since it was trained on data that stops in September 2021, I know this because it told me. It went on to state “Anything that has happened or been discovered since then may not be within my current knowledge base”, which makes it clear it has the ability to not feed you BS when it doesn’t know what it’s talking about, a trait we would like some people to attain. It can also sometimes kick you off if it is experiencing high traffic and you may be prompted with a message asking you to try again later.

Another weakness would be that it can sometimes provide a response that seems biased, and that’s because it is. Keep in mind that some of the information these AI are trained, are provided from the internet and we all know everything on the internet is real and unbiased, right? Some of the accessed sources can be completely fake and biased leading to ChatGPT potentially having biased responses.

This also means that sometimes it gets questions wrong. Besides the fact that the information it collects can be contaminated with opinions rather than pure facts, ChatGPT can also sometimes be overwhelmed, especially when providing information for multiple parts. Even though some of these descriptions may make it seem humanlike, it does not in fact know what it is saying. Though that sounds contradictory, ChatGPT does not “know” what is being said because it is in fact just a machine providing service based on the information it was given. With that being said along with its other shortcomings, it is always important to double-check the answer it provides. So don’t worry, your jobs are safe.

What Exactly is ChatGPT

By now, you might be wondering how exactly it was trained in the first place and what exactly it is. Most people may answer that ChatGPT is a smart interactive virtual robot and that’s sort of correct. ChatGPT is a language processing tool that gives a human-like response to your questions or statements. It uses a language model which is a tool that is able to calculate the likelihood of sentences or order of words because it was trained using a large amount of text in a specific language. This training process can take weeks to complete and has several stages to go through.

Some of the stages include data processing which involves dividing and converting words into numerical vectors, selecting the model architecture because there are different models that have different layers and learning rates, and the stage of training and validation itself. There are different types of language models such as XLNet, RoBERTa, and GPT. GPT, standing for generative pre-trained transformer, is a model that’s considered unidirectional, meaning it processes text in one direction, from left to right, and can generate new text along with understanding the relationship between them.

More on GPT

This is due to the diverse range of the tasks it was trained on. GPT is preferred for a chatbot over another language model such as RoBERTa because of its ability to respond human-like. If a language model such as RoBERTa was used for a chatbot application, it’s likely to have responses that are not forth-telling and will give low-quality conversation.  If you ask ChatGPT, for example, “What’s the best redaction software for video, audio, and documents?”, it will list suggestions with a deeper explanation as to why each of them helps and give a disclaimer at the end.

A chatbot using the language model RoBERTa wouldn’t respond in the same preferred way. Now that you know about GPT, let’s talk about GPT 4. GPT 4 was released in March of this year and everyone is waiting to see how it measures up to the GPT 3 which is already impressing millions. ChatGPT (using GPT 3) if you have not noticed, is restricted to language, it cannot use audio, process images, or produce drawings.

GPT 4 is multimodal, it is expected to be able to allow users to upload images, allow more input text that is eight times the current limit, and overall perform more creatively and easily handle more requests. This version, however, is so far only available with ChatGPT Pro which can be accessed with a monthly subscription of $20, half the price it was originally. Some have already stated that the traffic does not improve as much compared to the free version, but GPT 4 is a huge selling point.

“Just Ask ChatGPT”

Chat GPT has opened doors to a lot of possibilities, including a safe haven for laziness. Students and professionals alike have utilized ChatGPT for life-altering or unseemly tasks. We can all agree that at some point we’ve looked at our school work and thought… “How is this going to help me in any way?”, or maybe simply didn’t want to do it, and with a tool like ChatGPT, you technically don’t have to. Social media is already collecting evidence of students using ChatGPT in the worst way possible.

They are simply plugging in the homework question and pasting everything exactly, including the prompt ChatGPT uses that you may be familiar with stating “As an AI language model …”. Professionals are making use of it by asking ChatGPT to write cover letters after pasting the job requirements and their resume all at once. Some are asking it coding questions, and marketing strategies and people are taking to social media to follow ChatGPTs step-by-step recommendations on how to start a business and are updating their viewers.

Claims of becoming rich with ChatGPT, having specific prompts to use it more successfully and more might make you question the morals of those around us. Questions to help decide everyday decisions such as the best meal to eat while trying to bulk up, or, the meaning of life, and even asking ChatGPT if it’s alive can be helpful, fulfilling, and entertaining, but there are always correct ways to use these powerful tools. Using it as a reference point or the same way you would use Google and rewording it using your own words is the best approach.

Remember, it is using information that already exists, simply copying and pasting can be considered plagiarism. While ChatGPT is fascinating, the attachment and reliance look similar to that of the Magic 8 ball, alleviating the pressure of misunderstanding and decision-making. But unlike the Magic 8 ball, your boss may ridicule you for using it.