Article's Content
Article's Content
Using ChatGPT well makes it feel like you have a superpower.
With just one well-worded prompt, you can generate content that would otherwise take hours.
But that’s the key: You have to write your prompts correctly.
Here at Foundation, we’re excited about the opportunities that ChatGPT offers. But those opportunities are only as good as your ability to take advantage of them.
So, let’s walk through the most effective ways to get the responses you want out of one of the most powerful content-generating tools available. It’s not all automated yet, and it’s a little trickier than you might think!
Don’t Ask for What It Can’t Do
If you’re trying to hammer nails, don’t use a screwdriver. That seems obvious, but I’ve seen plenty of people showing how they use ChatGPT, and frankly, they’re misusing it to do things it wasn’t designed for. Some of these can change depending on what plugins you’re using (we’ll get to that), but for now, we’ll assume you’re using GPT-4 with no plugins enabled.
Fair enough? Then, let’s get to it.
It Can’t Make Up Your Mind for You
While ChatGPT is an impressive tool that can generate content, brainstorm ideas, and even engage in interactive dialogues, it can’t make decisions for you.
It lacks personal context and, therefore, can’t align its suggestions with your unique situation or preferences. Additionally, it doesn’t possess the ability to process emotions or perform ethical judgments.
It’s designed to help you generate options and stimulate thought processes, but the ultimate decision must always be made by a human. Always remember that ChatGPT is a tool, not a substitute for human judgment.
It Can’t Do Complex Math
That’s right, ChatGPT is an advanced AI, but it can’t do the math that a TI-84 can. That’s because ChatGPT isn’t designed to handle complex mathematical problems. While it has an internal “calculator” that can handle simple arithmetic, it doesn’t have the capability to solve advanced algebraic equations, calculus problems, or statistical analyses.
The reason for this is twofold. First, GPT models learn from text, and the underlying structure of the model isn’t built for numerical computation. Second, the training data itself is primarily textual, not numeric, so the model isn’t exposed to enough mathematical content to learn advanced math problem-solving skills.
This is more true for large numbers and complex equations. Take the example below:
The cube root of 82,058 is actually about 43.455. That might not seem like a huge difference, but it’s a relatively simple operation. You shouldn’t rely on ChatGPT for any complicated calculations—if you do, you might not even know how wrong it is.
It Can’t Do Good Research
ChatGPT cannot conduct real-time research or provide the most up-to-date information. This is because it is trained on a fixed dataset that only includes information up to the time of its last training session.
Any developments or new data published after that point will not be included in its knowledge base. Although it possesses vast amounts of information, it doesn’t actively update its knowledge or fact-check in real time against current data or sources.
This means that while it can provide a general understanding of a topic based on its pre-existing knowledge, it cannot offer the latest insights, statistics, or developments on a subject. Always cross-verify any information provided by ChatGPT with reliable, current sources.
Just check out what happens when I want to know who the NBA champions are!
It Struggles with Tone and Humor
This one is harder to prove, but ChatGPT can struggle with humor and tone. Maybe I’m just not a fan, but I don’t find its jokes especially humorous. You can decide for yourself how successful it is at this prompt, for example:
Depending on the request, it may struggle to hit the exact tone you want as well, although this improved dramatically with GPT-4.
It Can’t Multitask
ChatGPT is an incredible tool, but it does stumble when it comes to handling complex requests. This is primarily due to the way it’s programmed. The model is trained on a vast array of text data and makes predictions based on patterns it identifies in that data.
However, it doesn’t possess a deep, semantic understanding of language or the ability to follow long, complex instructions. Instead, it generates responses one token (a single word or character) at a time, without the ability to plan ahead or consider the overall direction of the query.
This design is why it sometimes produces incorrect or nonsensical results when faced with complicated tasks. In the example below, it did an admirable job but did use the letter “e” despite the instructions saying not to:
To get around this, consider breaking your complex requests into smaller, simpler queries. Instead of asking ChatGPT to write an entire article in one go, guide it section by section.
For example, you might first ask for an introduction, then a paragraph about the main point, and so forth. Alternatively, try to clarify and simplify your instructions as much as possible, focusing on one specific aspect at a time.
Remember, clarity and precision are your best allies when it comes to getting the most out of ChatGPT.
It Can’t Sniff Out BS
ChatGPT can’t evaluate whether to trust a source. It can tell you whether what you’ve submitted conforms with the general consensus as of its last update (January 2022), but it can’t evaluate new information.
For example, I inputted the contents of the LK-99 paper claiming to describe the creation of a room-temperature superconductor. The 2023 paper claimed that it would change the world, but later replication efforts found that it didn’t contain the secrets of superconductivity at all.
Anyways, ChatGPT’s response here is correct: The paper gives the impression that they’ve succeeded, but it’s something scientists would have to verify first.
That’s because ChatGPT can’t do that evaluation all on its own. It can’t confirm or debunk claims for which it doesn’t have the appropriate evidence. Additionally, it will accept most information you provide to it as true. For example, you can quickly convince it about events that happened since its last update:
That instinct to trust user-provided information is dangerous, especially if you’re relying on the program to sniff out BS for you. Be careful if you rely on ChatGPT in this way because it might just take your word as true.
Learn Expert Prompt Engineering
The above are all important things to know not to expect ChatGPT to do. But what can it do? How can you engineer your prompts for the best possible responses?
Be Specific!
One of the most crucial elements of a well-written prompt is specificity. The more specific you are, the better the chance ChatGPT has at generating relevant and accurate responses. When asking for a response on a particular topic or subject, provide as much context and as many details as possible. This will help narrow down the scope of GPT’s suggestions and increase its chances of producing high-quality responses.
For example, instead of asking for “content on climate change,” specify a particular aspect or angle you want to explore in your content, such as “the impact of deforestation on global warming.” The more details you provide, the better ChatGPT can assist you.
Guide Tone and Focus
Another way to improve your prompts is by guiding the tone and focus of ChatGPT’s responses. If you’re looking for a formal, professional tone, make sure to use formal language in your prompt. If you want a more casual or humorous tone, provide examples of jokes or puns that reflect the style you’re seeking.
Additionally, guide its focus by including keywords, phrases, or specific information in your prompt. This will help ChatGPT understand what aspects of the topic you want it to prioritize and incorporate into its responses.
Break Down Complex Requests
As mentioned earlier, ChatGPT struggles with complex requests. To ensure the best results, try breaking down your prompt into smaller, more manageable parts. This will help ChatGPT understand each section better and produce more coherent responses.
For example, if you want to generate an entire article on a particular topic using ChatGPT, consider breaking it down into sections such as an introduction, main points, and concluding statements. This will help ChatGPT understand the structure of the article and generate responses that fit into each section seamlessly.
Provide the Information It Needs
Earlier, I described ChatGPT’s trust in the user as a downside. However, it can be very useful, too, allowing you to provide it with updated information and keywords throughout your prompts. This will ensure that the AI is able to properly form its responses.
For example, if you’re trying to generate content about a certain subject, use related words and phrases during your prompt to help ChatGPT understand what it should be focusing on. Additionally, try providing it with extra information as you go along so that it can keep up with changes in the conversation or topic. Doing so will allow it to process the data more quickly and accurately.
Evaluate Its Outputs
Finally, it’s critical to evaluate ChatGPT’s responses and provide feedback. This will help the AI better understand what types of responses you’re looking for and improve its performance in future prompts.
Additionally, providing feedback can also help correct any potential errors or biases that may arise in its responses. It’s important to remember that ChatGPT is constantly learning from the data it receives, and your feedback can help shape its future outputs. If you’re struggling to get the answer you want, don’t be afraid to hold its hand a little to guide it to the answer you’re looking for!
The example below is a little silly, but it illustrates the kind of hand-holding the AI sometimes needs:
Expand Its Abilities with Plugins
We explained ChatGPT’s limitations earlier, but it’s important to know that with plugins, it can expand its capabilities considerably. Right now, there are only a few in existence, but you can expect that to change as GPT models continue to advance and more third-party developers jump on board.
A quick tip here: If you aren’t sure whether you need a plugin, or what plugin to use, there’s a plugin for that! The plugin “There’s An AI For That” helps ChatGPT recommend what plugins would be best for a given task.
Mathematics
Remember how ChatGPT’s bad at math? It doesn’t have to be!
There are specific plugins to help you with math. One of the best is Wolfram, which uses WolframAlpha’s mathematics engine to accurately process complex math questions. Remember that math problem from earlier? Here’s how ChatGPT answers it with the Wolfram plugin installed:
Not only does it give me the correct answer, but it also visualizes it and gives me a more complex answer that, to be honest, I don’t really understand. But that’s okay because I had ChatGPT explain it all to me!
Pulling Information from Links
Another major drawback of ChatGPT is its inability to refresh its database with new information. But plugins that allow it to follow links, such as Web Requests, help overcome that challenge by allowing ChatGPT to pull and analyze information from websites.
Now, you can ask it questions about current events, and it will give a detailed answer with stats pulled straight from the internet. This opens up a whole new realm of possibilities for research, although you do need to be careful to check the sources. Here, you can see how it successfully answered a question it failed to get right earlier:
SEO
There are plenty of separate, AI-driven SEO tools out there, including the excellent Jasper.ai. But if you’re determined to stick with ChatGPT, there are numerous plugins to make your SEO journey even better!
First, SEO.app writes articles optimized for certain keywords. It can also edit articles for relevancy and performance.
Quick Creator promises similar results but has a landing page tool that it says will guarantee high ranking in Google searches and make landing page creation a much faster process.
Finally, Linkhouse is focused on helping you build out your backlinks. It identifies the best websites for you to pursue for backlinks based on relevancy, domain rating, and your budget.
Prompt Creation
Ever since Prompt Engineer jobs with six-figure salaries started popping up, people have been increasingly interested in the new skill of prompt engineering. But did you know that you can even have AI help with that?
Plugins like Prompt Enhancer make it easy to write precise, detailed prompts that get you the results you want. Just make sure that the prompt that it creates actually meets your requirements—it might “hallucinate” specifications that you hadn’t intended, so you still need to follow that “evaluate its outputs” advice from earlier!
App Coordination
Orchestrating multiple applications to do what you want is a difficult but powerful ability—that’s why companies like Zapier make hundreds of millions of dollars doing it. Well, Zapier actually has its own ChatGPT plugin to make task creation and execution even easier.
With the app, you can tell ChatGPT to execute any of Zapier’s 50,000 actions with its over 5,000 supported apps. You can send an email, update a database, and message your team all from the same app window.
Reading, Writing, Running, and Compiling Code
In ChatGPT’s default state, you can ask it to write code for you, as in the example below. But you can’t compile, test, or interpret that code. That’s where plugins come in.
Plugins like AskTheCode allow you to analyze and ask questions about GitHub repositories. That means you can access code from any repository you own (or are given access to) and test it.
If you’d like to compile or run code, there are other plugins for that, too! Code Runner, as the name implies, is built for exactly this. It supports over 70 languages, so you can easily execute your code straight from the editor.
Specific Product Use
There are tons of other ways you can use ChatGPT outside of these! Plenty of apps and websites have created their own plugins with the platform, so you should explore the plugin store for yourself and see what appeals to you.
One interesting application of this is in travel planning. Plugins like Travelmyth, Trip.com, Staypia, and Kayak promise to help you find the best flights, hotels, and attractions wherever you want to travel.
Others, like Klarna and Shop, are built to save you money and make searching for what you want easier. They scour the web for the best deals and give you the link directly, so you don’t waste your time browsing through product pages.
How AI Continues to Shape Our Industry
ChatGPT’s capabilities are expansive and continually evolving. From enhancing the specificity of your prompts, guiding tone and focus, and breaking down complex requests to using plugins for various applications such as mathematics, pulling information from links, SEO, prompt creation, app coordination, code execution, and specific product use, there’s enormous potential to leverage this tool for a myriad of tasks.
While its limitations are real, the technology is constantly advancing, with new plugins being developed that augment its abilities significantly.
As end users, it’s essential to keep track of these developments, continually evaluate outputs, and provide feedback to guide its progress. With strategic prompt engineering and the judicious use of plugins, ChatGPT can become an invaluable tool in your repertoire.
Access Hundreds of Prompts [2024]
Ready to kick your AI mastery up a notch? Check out the AI Marketing Console.
Trust us, it’s the real deal for marketers. Better your ChatGPT & Midjourney skills and prompts to get more done, faster. This isn’t just about using AI—it’s about leveling up your game.
