You’ve probably heard a lot about ChatGPT recently. Maybe you’ve read one of our blog posts, come across an article in a major newspaper, or even seen a post on one of our social media pages. Lately, it seems like ChatGPT is all everyone is talking about. And with all the hype around ChatGPT, there’s never been a better time to learn how to use it effectively.
But before we can do that, let’s talk about ChatGPT: what it is and (most importantly!) what it isn’t.
What is ChatGPT?
In ChatGPT’s own words,
“ChatGPT is a large language model based on the GPT (Generative Pre-trained Transformer) architecture. As a language model, ChatGPT is designed to generate human-like responses to natural language input, allowing it to converse with users in a way that resembles a human conversation.”
Let’s break that down a little. What we’ve just read is that ChatGPT is a large language model that accepts inputs in plain English (or any number of languages, for that matter) and generates responses similarly to how a human would.
But–what exactly is a “large language model?”
In brief, it’s a type of artificial intelligence that is trained on human-generated text data in order to be able to understand natural language patterns. In simpler terms, it’s a machine learning model that is fed massive amounts of text so that it’s able to (1) understand text inputs and (2) generate outputs that it calculates are likely to make sense given a specific input.
Here, “generate” is a very carefully chosen word. ChatGPT is generative, which means that it can create outputs that it’s never seen before. ChatGPT does not just repeat text that was in its training data; it is able to extrapolate from its training data to create brand new words, sentences, and ideas.
So, great! Now we know exactly what ChatGPT is. However, we also need to know what it isn’t before we are able to effectively use it.
What are ChatGPT’s limitations?
It might seem like an all-powerful beast, capable of absolutely everything and on the verge of taking over the world. But don’t get too ahead of yourself: ChatGPT is not able to do everything.
Yes, it’s a powerful tool when used effectively, but it can’t quite replace a human being. Like many other great tools, it can be used to tremendously enhance productivity by helping its user brainstorm ideas or automate tasks, but it’s bounded by these four factors:
A dependence on data quality
A lack of common sense
An inability to handle complex tasks
Limited domain expertise
What Can I Do With ChatGPT?
Now that we have a better understanding of what exactly ChatGPT is, it’s time to learn about what we can actually do with it. In short, you can ask ChatGPT to do anything that can be answered as text and we can list them under three broad categories: explanatory, conversational, and generative.
Explanatory: you can ask ChatGPT to explain a word, a concept, or an idea. You can also ask ChatGPT to summarize, re-word, translate, or otherwise clarify a body of text.
Example: What does the word “engage” mean? Why is the sky blue? Why do dogs sleep on their backs?
Conversational: ChatGPT can engage in a conversation with you, answering questions, providing suggestions, and helping with decision-making.
Example: Can you give me some ideas for a child’s birthday party? What should I get my partner for our 10 year anniversary? What is the best way to arrive at my destination?
Generative: ChatGPT can generate content such as articles, stories, or code. This can be useful for brainstorming or content creation, but remember to always review and edit content before using it, as the quality of the generated content may vary.
Example: Create a blog post about automated processes. Write code that fits into this project. Can you give me content ideas for my startup?
How Can I Effectively Use ChatGPT?
Whew! By now, you’ve already learned a lot about ChatGPT and are probably itching to start using it. In this next section, we’re going to quickly go over how to use ChatGPT effectively so that you can get the most out of this powerful tool.
We think that it only takes 6 steps to use ChatGPT to its full potential; you should clearly define your task, use appropriate prompts, verify the response from ChatGPT, check the response for biases, edit the content in the response as needed, and then start to experiment!
Clearly define your task
Before ChatGPT can help you, you need to know exactly what you’re looking for. In other words, you have to know what you don’t know. Once you have a clear understanding of what you need, you’ll be able to generate a prompt that accurately conveys what your task is to ChatGPT. Remember, ChatGPT can’t read your mind! You need to tell it, in writing, exactly what you want.
Use appropriate prompts
Once you’ve defined your task, you are then able to write a prompt that can enable ChatGPT to generate a relevant response. In your prompt, you should use keywords that help ChatGPT understand what kind of response you’re looking for. Some examples of keywords are:
Topic: this one’s pretty obvious, if you’re looking for information on a certain topic, be sure to include keywords related to the topic in your prompt. For example, if you’re looking for information on renewable energy, you should use the keyword “renewable energy” but also related keywords such as “zero-emissions” and “solar power.”
Question: if you’re looking to answer a specific question, it’s good to give ChatGPT more context on the nature of your question: are you asking about a who, what, why, when or how?
Action: ChatGPT can perform several different kinds of actions, but you need to tell it what to do. Some keywords that may be useful are “summarize,” “define,” “explain,” “analyze,” or “generate.”
Comparison: if you want to know more about how two things compare, you should include keywords such as “advantages,” “disadvantages,” “benefits,” or “contrast” so ChatGPT knows how best to frame the comparison.
Location: if the information you’re looking for is specific to a region or a timeframe, be sure to include that in the prompt. For example, if you want to learn more about the history of renewable energy in the US, you can ask something like, “In the last 50 years in the United States, …”
For best results, it’s important that your prompt be as specific as possible. When you leave room for interpretation, you also leave room for the generated response to be less accurate or relevant to your needs.
Verify the response
Your job isn’t done as soon as ChatGPT generates a response, however. Once you have the response, you should go through the steps necessary to validate its veracity. It’s not always easy to check if information is true, but it’s incredibly important that you do so before using it or else you risk damaging your own credibility. We recommend you take the following steps:
Fact-check the information to quickly determine if it’s inaccurate. Websites like FactCheck.org, Snopes, or the Washington Post Fact Checker are all credible.
Cross-reference the information you received from ChatGPT with information you’ve found in multiple other sources. Two sources is the minimum for verifying information is true and you should use more if possible.
After you’ve found multiple sources confirming the information is true, you still have to confirm that those sources are credible. Plus, you should also make sure that your sources don’t cite each other. If they do, then they only count as one source.
If you’re still unsure about the information, consult an expert in the field---they’ll be able to provide additional insight and give credible sources to back up their claims.
Check for biases
This next step may be harder to do without context or expertise, but is nonetheless important. Once you’ve determined a response is factually correct, you should take the time to examine if it contains any potential bias. Using information with external motives or incentives without disclosing them may be just as damaging to your reputation as using false information.
Edit the content
We’re almost done! At this point, we’ve already determined what we’re looking for, crafted an appropriate prompt, received a response, and checked that response for biases and errors. Now, it’s time to (1) correct those details and (2) edit the content to fit your particular needs.
Depending on what kind of content you asked for, you may want to copy the response to a text editor such as Google Docs or a code editor such as Visual Studio Code. Next, you may want to use a tool such as autocorrect for written content or a linter for code to help you identify obvious grammatically and structural errors. At this point, you should choose whether you want to do a full re-write or whether you’ll just be fixing the errors in the generated response. Finally, once you’re done, make sure to do one last read-through before publishing the content.
As with most things, you’ll get better at using ChatGPT with time and practice. The more things you attempt and experiment, the more you’ll learn and improve! You’ll be surprised at how quickly you can improve your prompt engineering skills by just trying different prompts and tasks. Over time, you’ll develop a better understanding of how to use ChatGPT effectively for your specific needs.
Where do we go from here?
If you’re looking to not just become a more proficient ChatGPT user but also actually break into the world of tech, check out our bootcamps. There are bootcamps available for just about anybody, and we think that you’ll surely find something that’ll help you grow in your career. Poke around our website and see which bootcamp is the best fit for you. We can’t wait to see you in class!