Fast. Rapid. Lightning. Accelerating. This post does not need the help of ChatGPT to express to you how quickly the service is catching attention. It is gaining acceptance as well as gathering its fair share of hatred and bans. Here are a few ideas for how to best make use of this powerful tool.
ChatGPT is easy to use. With a simple sentence or question, users can generate hundreds of words on a topic. It can provide an amazing answer or complete rubbish, all while acting completely confident in its answer. There is the danger and warning, however, if you are not careful, you could publish something false and foolish. Caveat Emptor applies here. Triple check your results and, my opinion and advice is to use it as a starting point only.
ChatGPT Hack #1
All that said, the beauty of ChatGPT is in the name of the service — chat. It is capable of refining as it goes, of having a chat or conversation, allowing you to keep asking questions and getting the tool to focus in on your question or topic. Simple enough, right? Engage in some back and forth with the chatbot. Here are two more hacks and where I think it gets super interesting.
ChatGPT “Prompt Engineering” (Hack #2)
Our second hack is to get out of basic prompt mode. This is known as “Prompt Engineering” in a variety of tech circles. Since ChatGPT launched late last year, I have been working to fine-tune my prompts and conversations with it, and also with Midjourney, an image generating AI service (which I will post about shortly). All of these tools demand a certain mindset and language to get the most out of them.
One of the best sources I have found for this idea of engineering a prompt, or creating an advanced prompt, is via Rob Lennon on Twitter. He created a tweet thread about how everyone is stuck in beginner mode and then provided 10 ChatGPT Advanced techniques that went viral. He also offers a paid course for those who want to create productive and useful ChatGPT results.
As you can see above in the screenshot, Rob goes far beyond a simple question and combines a lot of instruction in his prompts so that the ChatGPT tool can understand what you want to achieve. (Note: this is my ChatGPT where I insert the suggested prompt.)
Some of his prompts are hundreds of words, so if you thought you had to be brief in your ChatGPT request, that is not the case. In the screenshot, you can see he instructs ChatGPT with the topic, audience, goal, and even suggests a style and you can see that it responds in a very specific way, versus generic responses based on simpler prompts (what I had been doing in my early efforts). I then continued the conversation by telling ChatGPT to please expand and it goes on for another few hundred words.
ChatGPT Hack #3
Instruct ChatGPT to take on a specific role, such as, a motivational coach, a screenwriter, or as a rapper, to name just a few. This guides ChatGPT to think as this type of person, or voice, and it often leads to more sophisticated results.
I found a GitHub page from Fatih Kadir Akın, a software developer in Istanbul, Turkey, who compiled “Awesome ChatGPT Prompts” that list out a variety of advanced prompts that you can copy and paste (and adapt) to your needs. He published a free ebook on the topic as well (although the download page asks for a contribution of your choosing, I believe you can just leave it at zero and still receive the document. It is listed on Gumroad, a reputable site). Some of the GitHub prompts listed do not work as intended, but they can give you ideas for expanding how you write your prompts.
Here is one of the examples Fatih’s site provided, which you can paste into the prompt area directly:
I want you to act as a philosopher. I will provide some topics or questions related to the study of philosophy, and it will be your job to explore these concepts in depth. This could involve conducting research into various philosophical theories, proposing new ideas or finding creative solutions for solving complex problems. My first request is “I need help developing an ethical framework for decision making.”
You can imagine that by asking a smart chatbot to take on a certain level of expertise, like the philosopher prompt above, it would provide better answers. If it does not, you can just do as I suggested in Hack #1 and keep the conversation going. I have gone as far as 10 or 11 times telling ChatGPT that the answer is not what I’m looking for and to try again. In one case, I told it that the answer was incorrect and it then apologized, and found the correct answer.
One more resource that I have found helpful in taking my artificial intelligence (AI) pursuits to the next level is this big list from Marcus Ramsey, entitled Best Chat GPT Resources. He also publishes a daily email newsletter called The Brink, which is a good way to keep up with the news around AI.