A Deep dive into chatGPT by OpenAI

“ChatGPT is an advanced language model developed by OpenAI. It is based on the GPT (Generative Pre-trained Transformer) architecture and is designed for natural language understanding and generation. With 175 billion parameters, it excels in various language tasks, including conversation, content generation, and code assistance.”

This is a legit definition for what ChatGPT is. Did I write this? Well… not really. This is a response generated by ChatGPT itself for the prompt “What is ChatGPT?”. ChatGPT basically accepts prompts from users and generate descriptive responses in real time. Though it seems fairly simple on the surface, a lot goes on underneath the hood of the most used Artificial Intelligence language model in the world.

ChatGPT is entirely and completely based on artificial intelligence and machine learning. It is the spear tip in the field of natural language processing which is a field in AI that focus on interactions between computers and human languages. Natural language processing uses algorithms and models to analyze, understand and recreate languages we use as humans. Not just that but ChatGPT is also a product of generative AI. It refers to the ChatGPT’s ability to generate new content or data that has never existed before even in the training data provided for the training of the model. ??

But how does it really work? Answer to that is… a little complicated. ChatGPT, like every other AI model, is initially pre-trained on a vast dataset containing texts from the internet where the model has learned grammar, facts and how words are related to each other. It is based on the Generative Pre-trained Transformer (GPT) architecture. When you ask something from ChatGPT, it looks at the words you gave and breaks them down into smaller pieces know as tokens which carries specific information that help the model to understand what you said. Meanwhile, it keeps a general awareness about the overall context by considering the entire conversation history you had with it to understand relationships and dependencies between words. It helps to generate responses aligning with the flow of the conversation. ChatGPT also uses special techniques such as attention mechanisms to nit pic the important parts of your input prompt and weigh the significant words to generate the response.

Then comes the most important part which makes ChatGPT so unique, generating real time responses. It predicts the next word in the sentence based on what it has learned in the pre-training to answer your questions. During the process of answering your… sometimes weird questions like “How to make girl fall in love with you in 2 seconds” … ChatGPT uses a technique called top-k sampling where it considers only the most likely word or token to use next so it can improve the quality and the coherence of the responses. In addition, ChatGPT monitors special parameter called temperature parameter to control the randomness of the output. So, the result you get is never repeated and every time a new and unique response is created. A bit of post processing is also done on the response to make sure ChatGPT is actually answering what you asked it.

?

All that is really amazing in the world on information technology. But… Is ChatGPT something that you can really depend on to help in your academics? ChatGPT generates responses based on patters it has learned. So even when it is generating rather human like responses, it may not always understand what its saying and create absolute gibberish as the response. And there is the risk of generating inaccurate results. ChatGPT lacks the ability to fact check or verify the responses it generates, therefor the generation of plausible-sounding but incorrect or nonsensical information is always possible. ?As ChatGPT is learning through patters from training data, the results it generate is based on the data it is fed during pre-training. So, the outputs can also be biased if the input is biased. Considering all that, it is wise if you use ChatGPT with an experimental mindset. It may require iterations and experimentation to get outputs you are looking for.

Let’s see what ChatGPT as to say about it,

“While ChatGPT is a remarkable tool for natural language understanding and generation, it should be used with awareness of its limitations. It’s suitable for a wide range of applications, from creative writing to content generation, but users should exercise caution, critically evaluate outputs, and be mindful of ethical considerations when incorporating ChatGPT into their workflows.”

So, the next time when you ask ChatGPT to do something for you, you may need to double check it.?

ChatGPT by open AI and all other advance language models is significant breakthrough in the application in the field of artificial intelligence. If used wisely, AI can make your life so much easier. Now you know how the most popular AI tool in the world works. Next time you type in a weird question… remember all the hard work that goes into that swift answer ChatGPT gives you.