Overview
Its is no doubt that ChatGPT has transformed the educational and professional world in terms of efficiency, but with this efficiency, it comes with drawbacks. Although it has the ability to instantly produce content, sometimes it lacks the ability to give its users exactly what they want. There are ways to combat this that deal with how users input prompts to chatGPT, DALL-E and other AI software, also referred to as “prompt engineering”. In this article, I will discuss how you can optimize your results from chatGPT and Dall-E via prompt engineering and the future implications of this new skill.
Image Generation
For AI image generation through DALL-E, there are specific ways to engineer prompts to create an image similar to the one you have in your head. An example is included below:
The first image was created with the prompt “a pretty lake”, the second was created with “a cabin on a lake with mountains in the distance”, and the third prompt was “a cozy cabin in Vermont in summer photorealistic”. Each of these prompts became increasingly specific but what was unexpectedly surprising was the word Vermont produced a result that was extremely specific.
When dealing with AI image creation, try using real places, and be as specific as possible to generate results that are most similar to the picture you have in your mind. Using little details such as the proximity of objects can help too. This would look like specifying how close the mountains are to the cabin. Another way to get specific about image creation is by mentioning specific artists, artistic eras, and mediums of art. It is an art to master the prompt of AI images and because of this, people are being paid to generate specific prompts in AI software.
Chat GPT and Prompt Engineering
Aside from engineering AI image rendering, you can also craft specific prompts to elicit a targeted response from chat GPT. Some of these methods include:
- Asking chat GPT to take the stance of a certain person. Eg “write an argument for the … and take the stance of a lawyer”, “act as a travel guide and create an itinerary for a weekend in Chattanooga, TN”, “you are a professional career advisor for college students who are graduating and want to find their dream job”
- Using as many details as possible related to the topic you are discussing. Providing context helps AI create relevant responses catered specifically to the context of your question.
- Adding constraints such as the length or format of the response. Eg. “Write a 500 word summary of the book”, “create a twitter post about ..”,
- Targeting a specific audience
- Making conversation with the chat bot to provide additional input and context
These are just some of the few ways you can ask chat-gpt to generate a specific response. It will take many tries, edits, and details to craft the perfect prompt but over time with practice it’s easier to learn how the software works and adapts to different asks.
Applications of Prompt Engineering
The field of prompt engineering is relatively new and it’s quickly becoming an essential area of focus for many businesses and organizations that utilize AI technologies. Because of this, there is an increasing demand for professionals who specialize in prompt engineering. Some of the most common jobs in prompt engineering include AI engineers, machine learning engineers, data scientists, and software developers. These roles are given responsibility to design, build, and test AI models and develop effective prompts that guide these models to produce specific outputs. In addition to their technical expertise, successful prompt engineers have strong critical thinking and problem-solving skills, and the ability to think creatively and adapt to new technologies and tools. The prompt professionals must know the application of the technologies they are using and specifics regarding the algorithms of each technology. As the field of AI continues to grow, these jobs related to prompt engineering will only continue to increase in demand.
Summary
Prompt engineering has revolutionized the field of artificial intelligence and has enabled machines like DALL-E and Chat GPT to generate more accurate and creative outputs based on specific instructions. By providing specific prompts and constraints, we can guide these AI models to produce more targeted and useful results that can target various industries, including marketing, e-commerce, and content creation. These advancements in prompt engineering have also opened the door to other innovations dealing with AI, opening up new possibilities for technology to interact with humans in more meaningful and natural ways. As AI continues to evolve and improve over the years, prompt engineering will remain a crucial tool in enhancing the capabilities of these intelligent systems. As much practice and information we can learn on this subject now, it will only help us in the future.
Sources
I think it is really interesting that you pointed out the importance that these prompt engineers know the context of the technology they are working with or the topic they are using the chatbot for. That is most likely where new jobs will be created and, for the time being, a fairly safe career choice for people worried about losing their jobs.
This was a super cool post and I think it will be a super important skill in the future. Im glad you are informing us about how to be good prompt writers because it will truly unlock a new way of using AI tools as they come out.
Hey Mary! I like that you included examples of how specifications in prompts can elicit various responses. I’m excited to see how the roles of machine learning and AI engineers and software developers will change as AI changes and we get more information.
This is actually a very useful blog. Writing the correct promo that makes it clear what you are expecting to get in return from these programs is very important and probably will be a useful skill while AI is becoming more and more used.
Love it! I have been thinking lately that AI prompts are a bit like me going to training for my dog. Dog training is actually less about training the dog as it is about training the people to become better communicators with the dog. I think the same is going to be true of AI, we are going to need to teach people to be better users of it.
I love this blog post! After learning how to optimize my prompts, I have gotten a lot more meaningful outputs from ChatGPT, specifically. I love the analogy that Professor Kane brought up with training the dog and how it is more about training people to be better communicators. With ChatGPT, I use the brainstorm keyword to come up with ideas for me to use later.
This is something that definitely improves daily. I always see on Twitter a new thread of great ways to prompt for ChatGPT. It’s bound to only get better over the years.