AI Wants To Help You: 5 Creative Ways To Use LLMs

Over the past 18 months, LLMs (Large Language Models) have been making dramatic strides forward, and better yet are improving all the time. These improvements have sometimes been locked behind paywalls, like GPT-4 and DALL-E, but new mediums have also become available. LLMs are now available for everything from AI content generation, image processing, video editing, and much more, as well as a host of software generation and debugging tools. The rate of change within AI at the moment is stunning, so it’s easy to miss opportunities where AI could be saving you time, money, and hassle right now. Let’s take a look at how you can make the most of these tools.

1 Use LLMs to prototype your designs faster

Image generators like DALL-E are scarily impressive

Text-to-image generators have come along significantly and are getting better all the time, so there are many ways to use them. You could generate an entire image from scratch (though it may look a little ‘odd’) or use it to generate a specific part of an image which you can then modify further. They’re also great for generating mockups, wire frames, and prototypes, which you can then interactively tweak using further prompts.

While there are free image generators out there, if you’re a creative professional you might already be paying or have access to one with Adobe Firefly. Free generators can be great, like Microsoft Designer’s Copilot-supported generator, but they can sometimes be poor quality or have a distinctive ‘AI’ look. DALL-E is included with a ChatGPT Plus subscription, and is one of the best out there. Google Gemini is able to generate images, albeit with restrictions put in place after some recent controversy, but this isn’t available in the EEA, UK, or some other countries at the time of writing. Different generators will often have a distinct ‘feel’ or aesthetic to them, and are capable of different things, so we’d recommend testing a few out or looking at some samples online before committing to a subscription (or sticking with a free option you like!)

2 Use LLMs to learn about new things

You can ask an LLM even your most embarrassing questions

In my view, the educational aspect of LLMs is an underrated advantage which they offer to anyone who is curious enough to ask questions. They’re seriously good study-buddies (although not to write that essay for you, of course). You can ask an LLM questions on any topic, either for a brief summary or a deep-dive into the specifics – just ask them why, how, or what causes something that you’re learning about. They are also great for generating flashcards, practice questions, and quizzes on a subject. For example, you could ask ChatGPT to generate a short quiz on a topic you’re studying but ask it to leave out the answers. Next, write out your answers in the prompt box, then ask ChatGPT to grade it for you.

Obviously, LLMs aren’t perfect; they have been known to hallucinate and can get some complex topics confused. For anything important to your education or career, we’d recommend that you still lean heavily on more academically sound reference materials like textbooks or class slides.

3 Use LLMs to generate logos or branding

LLMs can be great for a finished product or inspiration

LLMs can offer inspiration as they generate custom images to represent your brand. Logos and brands inherently follow patterns, because while every brand wants to look unique, they also want to be recognizable in their sector alongside other similar brands, so they tend to follow design trends. By describing what you’re looking for (e.g. logo, business card) and what your brand represents (e.g. mission statement, value proposition), you can at least gather some novel suggestions for consideration!

It helps to give as much detail as possible here, being clear about what you’re looking for. This can include asking for specific colors, text, or stylistic choices in your prompt. Giving more context on your business can also help. You could generate several different versions of an image for inspiration, then have your graphic designer combine the elements you like best into a final product. Some of these image generators really seem to struggle with outputs that include text though (with wacky spelling and hallucinated phrases), so we’d suggest generating graphics for your logo or brand, then using a conventional image editor to add text, additional details, or to combine multiple elements.

4 Use LLMs to generate templates for other software

ChatGPT Plus generated a Word document template for me to download

This one actually surprised me! ChatGPT Plus (on GPT-4) was able to generate an Excel document budget template, outputting both the document file and the python that it used to generate it. While it wasn’t a super complex format, the output could easily be enhanced by adding some smaller changes or more specific instructions in the prompt.

ChatGPT has already been great for generating cover letters or part of a resume, but this definitely takes it to the next level. We’ve already covered how to use ChatGPT to get good at Excel, but generating full templates makes this even more powerful and handy. Obviously, these templates can be limited, but they’re a great starting point if you’re facing writer’s block or not sure how to create something quickly, it will launch you in the right direction (or even just something you might hate but are now motivated to adjust)!

This can extend past the Office suite too, of course. LLMs like Gemini or ChatGPT are great at generating boilerplate items for just about any app that accepts a template. Because of how it generates templates using scripts, it isn’t limited by file types. As long as it can write some code to build a template, it should do the trick.

5 Use LLMs to translate, debug and generate code

LLMs are surprisingly good at debugging problems in your code, and even better at generating it

This one should be obvious: LLMs are great at debugging code. They’re still prone to confusing themselves on larger projects, and can pretty easily go off the rails hallucinating a feature or function that doesn’t actually exist, but for debugging they’re great. To get the best results when debugging code with an LLM, give it as much detail as possible, including any specific error messages. Be careful here though, because LLMs generally store all the prompts and data submitted to potentially train other models. So if you’re in a professional setting or using proprietary information, be sure not to give it to an LLM.

Another less obvious use of LLMs for code is to translate code. Say you’ve written an API specification in Python, but want to support a Java SDK at the same time. LLMs are great at this, and can get you most of the way there by rewriting your app endpoints in another language. This use case also seems to be less error-prone than other more novel coding purposes. You can also use LLMs to generate boilerplate code though. No more searching around GitHub for a nice bit to start a project from: LLMs can do that for you, including setting up package managers or dependencies, even basic CI.

Imaginative uses of LLMs can save you time and hassle

LLMs are becoming more powerful all the time, and like any powerful tool, it’ll take us some time to learn how to use them properly. We’ve focused here on what you can do with some of the big LLMs available easily today (ChatGPT and Gemini) but as more specific tools are released there will be even more capacity for creative uses of LLMs. In the meantime, we can still have fun trying to maximize the clear potential of these incredible tools.