Generative AI, Beyond Content Creation

Or leveraging the power of AI for good through analysis and reasoning.

Have you heard about that Generative AI stuff? No? Well, good news! At last, you can create that picture of a donkey water skiing on spaghetti that you always wanted¹.

The discussion around AI at the moment tends to be very focused on its use for content generation (i.e., images, text, or even computer code). While this can be fun, handy, or even an outright skive - for example I’ve heard of programmers who are now able to execute their entire workload with it in an hour or two at the start of each day.

But this discourse often ignores one of the most intriguing and potentially useful characteristics of generative AI systems—its ability to conduct analysis and execute reasoning against the information with which it is supplied.

These abilities first started to interest me when I realised that I might be able to utilise them to overcome my dyslexic traits and assist with some of the tasks that I find challenging on occasion—such as parsing or writing very long or dry text.

For example, I receive a great many newsletters every week, which are important for me to read, but some of which are written in a very long-form manner. Therefore, it was logical to start by trying to get ChatGPT to summarise the content for me.

Unsurprisingly, my first attempts weren’t that useful—the system managed to shorten the text but spat it back in a similar tone, which I still found quite difficult to read. So, I started experimenting with alternative prompts like “In bullet-point form” or “in 100 words or less.” The results started to improve, and over a period of days or weeks, I gradually refined the commands into a Custom GPT which I now use on a daily basis: “Summary Assistant,” which is now available for everyone to use here.

The revolutionary aspect of Summary Assistant was that I managed to get ChatGPT to behave more like a software application or wizard. In summarising the text, it goes through a few simple steps: it presents a numbered list of the key points in the text, asks if you want to know more about them (which you can answer just by giving the number), and can even supply you with critical analysis of the points presented.

Buoyed by this success, I started work on my next project, a Custom GPT designed to aid a user in the creation of a customised CV and job application for a specific role—something that, I’m told, is crucial in the current job market and that I find extremely challenging, again due to my dyslexic traits.

This GPT was designed to:

• Compare a job advert/description against a supplied CV

• Provide an initial assessment of the candidate’s suitability for the role

• Suggest areas of the candidate’s CV that could be improved and ask for examples

• Provide a redrafted CV and covering letter to support the application

After a few weeks of finessing the application, I released it publicly this week, and I have to say I’m still staggered by what it can do—both in dealing with the instructions I’ve given it and in using its “intelligence” to go beyond what it’s asked for.

Over hundreds of test runs and job roles varying from a waiter in a Greek restaurant to the Prime Minister of the UK (I hear we might need a new one soon?), the system provides sensible, consistent, and solid advice, does a good job of consolidating information, and (with a Plus subscription) supplies the result in a well-formatted Word document to download.

Now, I know that there are comparable services out there that do this, but the thing is:

• This GPT behaves more like software than a chatbot. That’s interesting for a start.

• Comparable services are often expensive, e.g., TealHQ is ~£40 a month for custom CVs, whereas this is effectively free.

• I’ve only been playing with this tech casually for a couple of months, and I feel like I’ve barely scratched the surface at this point.

So, despite all the controversy about AI, which I do understand, I have to say—I’m also very optimistic about what this technology can help us achieve and the benefits we’ll derive from it.

I encourage you to have a play, see what you can do, and please do share your findings with me—I’ll be fascinated.

Links:

• Job Application Assistant: https://chatgpt.com/g/g-XiQJZE8Gj-job-application-assistant

• Release Notes: https://www.kemptand.co/jobs-GPT

• Discussion thread on Linked In: https://www.linkedin.com/posts/chriskempt_chatgpt-job-application-assistant-activity-7211488418025848832-Ci1b

1. I tried various animals before settling on this so, because “fun” is actually also an important AI use case, here are some of the others I considered - enjoy!