Conversational AI can produce amazing results. You just have to know how to ask.
You’ve probably seen all the cheat sheets, user’s manuals, and online courses for AI prompt writing that have sprung up online recently. Maybe they’ve convinced you that using conversational AI is an advanced skill gained only through long study.
If so, this article is for you. I will use examples to show that productively using Large Language Model (LLM) conversational AI is entirely about successful communication.
Why? The AI engine understands how to communicate coherently about every subject within the scope of all human knowledge. Therefore, the best way to get the desired result is to communicate your request as clearly as possible. In other words, communicate with intention.
Here’s an example from a session with Microsoft Bing, a version of OpenAI's ChatGPT integrated with the related DAL-E module.
I could ask "Please show me a picture of a dog." From that request, I might get a response listing the names of some websites that could have pictures of dogs, and maybe even a definition, as in the first example below.
But let’s say I wanted Bing to generate an image of a dog. I could revise my request to, "Please create a picture of a dog." When presented with only a subtly different statement, Bing knows that I expect it to render a dog picture. In response, it could probably return a very good resemblance of a dog like the ones shown below.
It's important to note that the above images are not those of real dogs. They are what ChatGPT understands a random dog to look like. Still, it’s pretty realistic. I don’t know about you, but if I saw any of these on a dog grooming website, I wouldn't even blink.
Not quite what I was looking for, though. I might then supply an added prompt like "Please make it look happy," and ChatGPT would respond by rendering images of a happy looking dog — something like those shown in the following samples.
ChatGPT understands the face of a happy dog enough to make some various pictures of that condition.
However, let's say that is still not really what I wanted. I was trying to accomplish something much more elaborate, maybe even something I might not otherwise dare ask of someone else. How about this: “Please create a picture of a smiling dog wearing sunglasses and a straw hat, sitting on a blow-up pool toy in a swimming pool at night with backlit palm trees in the background”? You know, just as an example.
This kind of specific request might take longer to process — roughly 10 seconds compared to the 3 to 5 seconds to generate the previous examples. However, when you see the results generated from such a specific request, you can see the level of detail you can achieve with systems like Bing and the base engine of ChatGPT.
So, here is my argument in favor of intentional communication. When you are using the newest forms of AI, you will often receive results specifically tuned to what you are asking for. Therefore, if you focus first on communicating your intentions effectively in the request, there is a very good chance the responses will generate exactly what you want.
Do you use conversational AI in your workflow? If so, what do you think are its strengths and weaknesses? Share your thoughts on our company LinkedIn page. While you're there, be sure to follow us to stay updated on new blog posts.