The Problem Might Be You, Not AI:
How to Use Generative AI Effectively
If you're getting bad results from generative AI (which I'll just call AI in this article), you might be using it wrong. The problem may be you, not ChatGPT. More specifically, you may be using AI inappropriately. In this article, I'll explain why and offer some insights on how you can use AI more appropriately.
Recently, I heard someone on a podcast talk about how bad AI output is. They said that AI-generated content was bland, uninteresting, and boring. They were specifically talking about ChatGPT, but I think their comments could apply to virtually any generative AI chatbot. The person's comments were reasonable; they even went so far as to say that some of the poor results might be due to a lack of prompting skills.
They had a point. A lot of AI-generated content is garbage. You've probably seen poor quality output that was created with AI. The writing is bland or over the top, using obvious AI "tells" with respect to word choice and pacing. (How often do you write "nonetheless" or "delve?" The writing is often pretty generic, the pacing repetitive and exhibits a lack of nuance. AI simply isn't great at generating quality content. Does that mean we should abandon it for creative work? Nope, not at all. Just because lazy AI use results in bad output doesn't mean that thoughtful use can't result in great output.
Why you might be the problem (well, not YOU, but other users)
Poor quality AI typically results from using AI poorly. That sounds trite, but it's true. Part of the problem may be a lack of skill and experience, especially about AI's limitations. Generative AI chatbots are simple to use, they'll give you something even if your prompts aren't great. But the most skilled you are, the better your results will be. So, to an extent the problem is the user, not the tool. It's like me with a jigsaw. The poor quality cut isn't the fault of the saw. (I am not a good woodworker.)
A second, more interesting problem, is thinking about AI the wrong way. There absolutely are AI-based tools that can do work for you. Some of these are quite good and they're getting better. That's not what I'm talking about here though. Here I'm discussing the use of chatbots for creation. The key to producing great content is two fold, understanding the limitations of the tool, and having the right mental model. The latter is my focus here.
If you use ChatGPT or some other AI chatbot to help you create, get the idea of AI doing the work FOR you out of your head. AI should supplement you, not replace you. Ultimately, YOU are the creator, AI just helps you. And AI can be a massive help. It doesn't matter what you create, even if it's just email messages. Used as an assistant AI can be be helpful in ways unimaginable a few years ago.
How? Well, I see two main ways. AI can create content that you edit, or AI can assist you throughout the creation process. The former is fine for less important creations such as routine emails or letters, but for anything important, you're better off to think of AI as an assistant not a drone doing your work.
Why? Well, there are a few reasons. At a macro level, we're in danger of drowning in a proverbial sea of AI generated content. (We're already seeing this.) Do you really want to contribute to this disturbing trend?
At a more personal level, if you're just letting AI do your work, anyone with a bit of skill in AI and knowledge of your area can do the same thing. If you're churning out AI-generated garbage, you may soon find yourself redundant.
The next reason is a bit esoteric, but it's important. If AI is just cranking out work for you, you're robbing the world of your unique voice and your unique way of viewing the world. Yeah, I know, that sounds a bit new-agey, but I 100% believe that it's true. Again, I'm not talking about emails or cut-and-dry reports (that nobody will ever read) here, I'm talking about more creative work. Work that, in part, defines who you are and how people see you.
The final reason is that having AI generate output for you deprives you of an important part of the intricate process of thinking and communicating. For example, writing isn't just typing out words and sentences, it's a process of thinking. A subtle corollary to this is that when AI does the work for you, you rob yourself of the growth and learning that comes from the struggles of creating content. If you don't do the work, you don't reap the learning.
Changing your mental model of AI
Think about it this way. If you're working on something important, would you hand it off to a subordinate or work you work on it yourself? Hopefully, you'd tackle the task yourself. But while working on that important thing, it would be handy to have a colleague to help you develop and clarify your ideas and thoughts, especially if that colleague had infinite patience and a lot of knowledge. It's even better if that colleague is willing to critique your creation like a good editor, offering insights on how you can improve your output. AI can serve as that colleague. I've written about this before. My personal mental model for AI is as an infinitely patient, extremely versatile and knowledgeable colleague. Such a colleague can be invaluable.It's even better if that colleague is willing to critique your creation like a good editor, offering insights on how you can improve your output. AI can serve as that colleague. I've written about this before. My personal mental model for AI is as an infinitely patient, extremely versatile and knowledgeable colleague. Such a colleague can be invaluable.
Using AI as an effective assistant
Here's how I use AI to help me create articles for AI Goes to College. Not every article gets created this way, but here's the general process.
Critiquing the idea: Often, I start off by asking AI what it thinks about a certain idea. The prompt usually goes something like this: "I'm thinking about writing an article for my newsletter AI Goes to College about [topic]. What do you think?" I like to ask a pretty general question at this point, just to see how AI will respond. Depending on the response, I may continue the chat, refining the idea along the way.
Developing the idea: This step takes different forms. If I already have a good idea of the points I want to cover, I might develop an outline, then ask AI to critique the outline. (One of my favorite prompts is "Please critique this passage/outline/article.) If the idea is still a little fuzzy, I may ask AI to give me its idea of the main points I should cover in the article. Sometimes I'll ask AI for an outline for the article. Note that these primarily help me overcome the blank page problem. I never use the outline as is. But, AI is incredibly helpful in getting me started.
Developing the content: Once I start actually writing the article, I ask AI to critique it. Depending on the article, I might do this several times as I develop the article. AI often makes useful suggestions. For example, when critiquing an earlier version of this article, ChatGPT suggested adding some specific advice, which resulted in this section. It also caught a repeated sentence that I missed. My typical prompt is something like "Here's an article for my newsletter AI Goes to College, which helps helps higher ed professionals navigate the world of generative AI. What do you think?" Then, I paste in the article (or portion). This simple approach works quite well for me, but your mileage may vary, so you may need to adapt my method to your own purpose.
Here's the main idea. If you want to use AI to produce good output, don't ask it to do the work for you, ask it to help YOU do the work. Focus not just on efficiency, but on effectiveness as well.