Home / Articles / Web Tools / Get to Know ChatGPT: A Guide to Using Large Language Models for Fun and (Maybe) Profit

Get to Know ChatGPT: A Guide to Using Large Language Models for Fun and (Maybe) Profit

ChatGPT is one of the most, if not the most, well-known chatbots in the world today. As someone who’s been actively reading tweets from SEOs and Artificial Intelligence (AI) enthusiasts, ChatGPT is a name I see every time I pull up my Twitter. 

ChatGPT on Twitter
See these tweets here, here, and here.

What is ChatGPT?

ChatGPT is a chatbot built by Open.AI with machine learning technology, known as “Generative Pre-trained Transformer 3 (GPT-3)”. In case this is your first time hearing it – GPT-3 is a Large Language Model (LLM). It uses machine learning to generate natural language text that is difficult to distinguish from text written by humans.

Boom! That's a lot of jargon in one paragraph.

Three Names to Remember

The term “Large Language Model“, or LLM, refers to a machine learning model that is trained on a large dataset of human language data and is able to generate natural-sounding text. 

Generative Pre-trained Transformer 3, or GPT-3, is a specific LLM, developed by a company named OpenAI, that has achieved impressive results on a wide range of language tasks such as translation and text summarizations. 

ChatGPT is a variant of the GPT-3 language model that is specifically designed to generate human-like text in a chatbot context.

So, think: 

  • LLM as the technology that helps robots read and write human language
  • GPT-3 is a type of LLM that is developed by a private company (OpenAI)
  • ChatGPT is the robot that uses GPT-3 to interact with humans in live chat

Why ChatGPT is Special?

What sets ChatGPT apart from other language models (there are many of them! more about this later) is its ability to maintain context across multiple turns in a conversation. When you chat with the bot, it has “memory” and can generate human-like responses that are relevant to the topic being discussed, and that build upon previous statements made by the user or the chatbot itself.

But that’s not all – the tool can do more than just carry on a conversation. It's also able to perform a wide range of language tasks, such as translation, summarization, coding (yes, programming languages are human language too), and even writing poems. 

What Can ChatGPT Do for You?

From generating ideas and content to providing research and support – ChatGPT is a valuable resource that can help individuals and organizations increase productivity and efficiency. Whether you are a hobby blogger, lawyer, school teacher, or even company CEO – ChatGPT can make a significant impact on your work.

Below are some specific tasks you can outsource to ChatGPT. 

For Bloggers

Example – Comparing two different tools and showing the content in table format using ChatGPT-3.
  • Generating ideas and writing drafts for your next blog posts
  • Researching and fact-checking blog content
  • Converting content (ie. reviews, development timeline) into tables or bullet lists

For Lawyers

Example – Summarizing complex legal concepts in layman's terms using ChatGPT-3.
  • Drafting legal documents and contracts
  • Summarizing complex legal concepts for lay audiences
  • Organizing and managing legal documents and case files

For School Teachers

Example – Generating quiz questions with ChatGPT-3.
  • Developing lesson plans and teaching materials
  • Creating assessments and quizzes
  • Generating different scripts for different levels of students

For Programmers

Example – Writing JavaScript with ChatGPT-3.
  • Generating code snippets and debugging coding errors
  • Explaining your codes in written language 
  • Switching functions between different programming languages

For Musicians

Example – Writing new lyrics with ChatGPT-3.
  • Generating song ideas and melodies
  • Writing lyrics
  • Researching and staying up-to-date on industry trends and techniques

For Company CEOs

Example – Quick news bite directly from ChatGPT-3.
  • Organizing and managing company projects and tasks
  • Summarizing business news, staff reports, and presentations
  • Collaborating with team members and other stakeholders (writing emails, texting, etc)

Prompts You Can Use for ChatGPT

I use ChatGPT mainly as my educator and writing assistant. Here are a few prompts I frequently use during work. Fatih Kadir Akın has curated a greater list of awesome prompts at Prompts.chat and in this free e-Book – be sure to check out if you want more.

I want you to act as an educational content creator. You will need to create engaging and informative lessons for students seeking guidance or advice in a particular topic.
I want you to act as a stand-up comedian. I will provide you with some statements and you will use your wit and creativity to create scripts based on those statements. You are allowed to incorporate personal anecdotes or experiences into the routine in order to make it more relatable and engaging for the audience. 

This is the first statements: 
Translate the following statements to English for me:

A Deeper Dive Into Machine Learning in Human Language 

ChatGPT is Not Alone

While GPT-3 is all powerful and popular, there are many other large language models available in the market (see table below).

Most of these LLMs come with their own unique capabilities and have achieved strong results on various language tasks. Language translation, sentiment analysis, and question answering are some common language work performed by LLM tools.

Other Notable LLMs Available in Market

ModelDescriptionSizeKey Features
LLMLong-form language model developed by OpenAISmaller and simpler than GPT-3Focused on generating coherent text within a specific domain or topic
BERTPre-trained language model developed by GoogleModerateCan be fine-tuned for a wide range of natural language processing tasks
XLNetPre-trained language model developed by Google and Carnegie Mellon UniversityLargeUses a novel approach called "permutation language modeling" to achieve strong results on a wide range of tasks
RoBERTaPre-trained language model developed by Facebook AILargeVariant of BERT that was trained on a larger dataset and with several modifications to the training procedure
ALBERTPre-trained language model developed by GoogleModerateLightweight version of BERT that uses techniques to reduce size and computational requirements
ELMoPre-trained language model developed by Allen Institute for Artificial IntelligenceLargeUses a deep contextualized word representation that is able to capture meaning from the surrounding text
T5Pre-trained language model developed by GoogleLargeCapable of performing a wide range of natural language processing tasks without task-specific training
CTRLPre-trained language model developed by OpenAILargeUses a novel approach called "conditional transformers" to generate human-like text
Transformer-XLPre-trained language model developed by GoogleLargeUses a novel approach called "relative positional encoding" to model long-range dependencies in text

LLMs Do Not Understand

While LLMs like ChatGPT are very good at generating human-like text, they do not have a true understanding, opinions, or beliefs of the topic (* see sidenote below) they are discussing in the same way that a person would.

* Sidenote: Though… not everyone agrees to this. Blake Lemoine, ex-Google senior software engineer, said that Google’s language model has a soul. He was suspended by the company in June 2022 and got fired later.

Instead, LLMs are able to generate text by using machine learning algorithms to analyze a large dataset of human language data and learn the patterns and relationships that exist within the data. Once these patterns have been learned, LLMs use them to generate new text that is similar to the text in the training data.

In other words, LLMs are not intelligent. They are simply very good at guessing the right words to use in their “writings”.

LLMs (ChatGPT, especially) Are Very Good at Guessing

What LLMs do is analyze whatever texts they have “read” in the past and guess what are the best words to write next. 

For example, to determine the next words to complete a sentence like “Wayne Rooney has dribbled pass…” ChatGPT, the LLM I am currently using, would analyze the patterns and relationships in the language data it has been trained on to determine the words and phrases that are most likely to come next in this context. 

Wayne Rooney has dribbled past? through? around? the defender.

The model might have learned from its training data that words such as “past,” “around,” or “through” are commonly used after the phrase “dribbled pass.” It could then use this information to generate a new sentence that reads:

  • “Wayne Rooney has dribbled past the defender”, or 
  • “Wayne Rooney has dribbled around the defender”, or 
  • “Wayne Rooney has dribbled through the defender.”

Sentences are just strings of words

Here’s another example to further demonstrate the power of ChatGPT prediction. The following snippet is picked from the first chapter of Percy Jackson and the Lightning Thief (Book 1) by Rick Riordan (my son’s favorite novel). 

"Being a half-blood is dangerous. It's scary. Most of the time, it gets you killed in painful, nasty ways." 

I asked ChatGPT to complete the paragraph. Here are two variations ChatGPT gave me:

Version #1

"But I've always known that I was different. I've always known that I was a target. That's why I've trained myself to be strong. That's why I've learned how to defend myself. I won't let anyone hurt me. I won't let anyone take away the person I am."

Version #2

"I've tried to stay hidden, to keep a low profile. But it's not easy. There are always people watching, waiting for me to slip up. I can't let them catch me. I can't let them hurt me or my family."

In both of these examples, the next three sentences continue to explore the theme of being a half-blood and the dangers and challenges that come with it. 

The first option focuses on the narrator's own efforts to be strong and defend themselves, while the second option focuses on the narrator's efforts to stay hidden and avoid detection. Both options build on the idea that being a half-blood is dangerous and can lead to harm, but they approach this theme from different angles and tones.

Impressive, right?

The Disadvantages & Issues with LLMs 

Like any machine learning model, ChatGPT is not all bright and shiny. Here are a few examples of some possible problems that may arise with ChatGPT:

1. Bias in the Training Data

Like most machine learning models, ChatGPT is trained on a large dataset. It may incorporate biases present in the data. Monkeys see, monkeys do. If the training data contains a disproportionate number of examples written by a particular group of people, ChatGPT may generate text that is biased toward that group.

Read Melissa's article The viral AI avatar app Lensa undressed me—without my consent. Not related to ChatGPT but it demonstrates what might go wrong with A.I.

2. Difficulty Understanding Context

ChatGPT is designed to generate responses based on the context and meaning of the input text. However, it may sometimes have difficulty understanding the context of a particular prompt or conversation, which could lead to inaccurate or nonsensical responses.

3. Limited Ability to Handle Open-ended Tasks

ChatGPT is designed to generate responses to specific prompts or questions. It may have difficulty handling open-ended tasks or situations where the desired response is not clearly defined.

4. Limited Ability to Handle Uncommon or Domain-specific Language

ChatGPT is trained on a wide range of text, but it may struggle to understand or generate uncommon or domain-specific language. This could be a problem if you are using ChatGPT for tasks that require a deep understanding of a particular field or subject matter.

Overall, ChatGPT is a powerful and useful tool, but it is not without its limitations. It is important to keep these limitations in mind when using ChatGPT and to carefully consider the suitability of the model for a particular task.

Wrapping Up

Perhaps the biggest reason why ChatGPT has taken the world by storm is because of its simplicity. Don’t take my word for it – You can try it out yourself here, for free. You don't have to be a tech wizard – just feed it a few words and see if it does its magic. Just be sure to double-check its output before sending it off to your boss, your grandma, or the President. 

And as always, remember: with great language modeling power comes great responsibility (and hopefully a few chuckles along the way).

Additional Notes

The Development of GPT-3

The development of GPT-3 can be traced back to the original research paper “Improving Language Understanding by Generative Pre-Training”, written by Alec Radford and his colleagues, and published in a preprint on OpenAI's website in June 2018. 

In November 2019, OpenAI released GPT-2, which was a significant improvement over the original GPT model. GPT-2 was trained on a dataset of more than 8 billion web pages and had a much larger size (1.5 trillion parameters). 

In May 2020, OpenAI released GPT-3, which was an even larger and more powerful version of the GPT model. GPT-3 was trained on a dataset of billions of web pages and had a size of 175 billion parameters, making it one of the largest language models ever developed. It was able to perform a wide range of language tasks, including translation, summarization, and question answering, and was able to generate human-like text that was difficult to distinguish from text written by a person.

About Open.AI the Company

OpenAI is a research laboratory based in San Francisco, California. The company is dedicated to advancing artificial intelligence in a responsible and safe manner. It was founded in 2015 and funded by big familiar names in tech industries including Microsoft, Elon Musk, and Peter Thiel

Notable leaders in the company include:

  • Sam Altman – CEO of OpenAI and a well-known tech industry leader. He has been involved with the company since its founding and has played a key role in its growth and development (read his blog here).
  • Greg Brockman – Co-founder and CTO of OpenAI, former CTO of Stripe.. He has played a key role in the development of the company's large language models, including GPT-3 (learn more).
  • Ilya Sutskever – Chief Scientist at OpenAI and a co-founder of the company. He is a leading expert in the field of machine learning and has played a key role in the development of the company's research and development efforts (read his tweets here).
  • Wojciech Zaremba – Co-founder and Research Scientist at OpenAI and has played a key role in the development of the company's large language models, including GPT-3 (Twitter profile).
  • John Schulman – Research Scientist at OpenAI and has played a key role in the development of the company's reinforcement learning algorithms and technologies (Twitter profile).

Also Read

Photo of author

Article by Jerry Low

Keep Reading