By now you’ve heard of ChatGPT, a service that uses generative AI to take up writing tasks, edit anything you ask it to, or create snippets of code in programming languages. It’s a powerful tool that can take on some tasks that you may find tedious or boring. It has its limits, but users are continually finding more and more things that it’s good for. In today’s post, we’ll take a look at some things you should know about it before using it, but also three ways to use it as a productivity tool.
A brief intro to ChatGPT and generative AI
These tools became very popular around December of 2022, just days after OpenAI released ChatGPT to the public. It promised to be a really good chatbot—a program that can talk to you like a friend—but it was pretty obvious to students and writers that it could be used to research and write on just about any publicly available topic one can think of. In other words, it was perfect for generating mediocre essays to turn in for writing-intensive university classes like the one that I was teaching at the time.
Of course, knowing a little bit about how the original, widely available version was trained could tell you that this is kind of a bad idea. For one, it gets factual things wrong a lot of the time. This is because ChatGPT was trained on the public internet to establish the meanings of words. This makes ChatGPT as unreliable as any other non-expert, non-peer-reviewed source on the internet, or a combination of expert and not-so-expert opinions.
The concepts that drive Large Language Models (LLM) are heavily reliant on taking averages of how close words are to each other in a typical sentence, then gleaning some kind of meaning of the word from that web. This gives ChatGPT a writing style that is typically flat and straightforward, which can be fine for technical projects, but is a cheap imitation of a good writer. You wouldn’t want to use it to write a voiceover for your video projects, for instance, without editing it to reflect how you talk.
With that in mind, there are plenty of writing tasks that don’t require any kind of imagination or connecting with your audience. These are some perfect candidates for automation with a tool like ChatGPT.
1) Generate starters and templates for any project
I’m beginning the list with something very general here, but it’s because ChatGPT can be used as a writing template generator for anything. I recently asked it to put together a business plan for an idea that I had, and ChatGPT did a fairly good job at creating something in the format of a business plan, but the data in it would have been based off outdated stats, and, more importantly, it generated a budget that proved my idea wasn’t able to turn a profit.
That sounds like a failure, but ChatGPT’s prompt system is so flexible that I simply asked the machine if it saw any problems with what it wrote. It then gave me a list of seven things to check the accuracy of, pointing out all of the flaws in what it had just written. If you have a big project to undertake, like a business plan, you will need to update all of the data anyway during an editing and research phase, but having the skeleton can get you way ahead on your timeline.
Think of how helpful it could be for you to have it generate many questions for a survey or other communication—just make sure the questions are appropriate before copying them into your form.
2) Use it to edit and summarize
Likewise, you can ask ChatGPT to edit specific passages of any text by giving it that text in a prompt, like “Please help me clarify this text: [pasted text].” Any part of the writing that isn’t clear can be cleaned up, you just have to know how to ask ChatGPT for what you want. Few people have the advanced editing skills to read deeply and quickly for that kind of project, but ChatGPT is an effortless way to edit pieces of writing. Give it a style or give it more information about how you want the writing to sound, and it will give it its best shot.
It’s important to remember that it can work with any text that you supply to it, meaning that you can take down notes, give them to ChatGPT, and ask for a summary. It’s even more important to keep in mind that you’re giving the content to OpenAI when you do so, so make sure that you’re not giving away proprietary or personal data in the process.
3) Learn to code without really learning to code
Since ChatGPT was trained on basically the whole internet, it “knows” a lot about text-based activities like writing programs, websites, and scripts. This means that you can ask it for snippets of code, or ask it to help you debug the code you’ve already written. If you ask it to build something, it might or might not work, but remember, you can also ask ChatGPT about the code and what it’s doing.
Here’s a problem that I run into fairly often when dealing with audio files, and how ChatGPT would solve it:
Notice that the code that it generated includes comments, which tell me what each part of the script does so I can inspect it. If it messes anything up, I can simply ask ChatGPT about it by asking it to debug the code for me as well. Not included in my screenshot are a couple of paragraphs that ChatGPT included to clarify which of the variables need to be replaced with the actual locations on my system.
Of course, it takes a few more steps to actually use the script, but if we’ve gone this far, why not just ask ChatGPT how to do that, too?
-Written by Derek Jeppsen on Behalf of Sean Goss and Crown Computers Team