In today’s column, I explore the ongoing and expanding use of prompt-a-thon hackathons that serve to enhance both the core fundamentals and advanced techniques underlying prompt engineering and aid users in enhancing said vital skills. I also directly share some of my personal experiences organizing and facilitating prompt-a-thons. In brief, a well-devised and well-executed prompt-a-thon can be a highly effective means of helping users become more proficient at the best practices of prompting and suitably leveraging generative AI and large language models (LLMs).
Let’s talk about it.
This analysis of an innovative AI breakthrough is part of my ongoing Forbes column coverage on the latest in AI including identifying and explaining various impactful AI complexities (see the link here).
Prompt Engineering As A Learned Skill
Readers might recall that I previously posted an in-depth depiction of over fifty prompt engineering techniques and methods, see the link here. There is no doubt that undertaking proficient prompt engineering is a learned skill. That being said, sure, you can always do a lot with seat-of-the-pants prompting too. But the top-notch prompt engineers know that seriously learning a wide array of researched and proven prompting techniques is the best way to get the most out of generative AI and likely garner you some good bucks for your skilled deeds.
There are numerous classes covering prompt engineering.
Some of the courses are online and involve an automated system that walks you through various steps. Some classes consist of prerecorded videos along with hands-on exercises. Human-led courses are still actively underway, at times done remotely and in other cases occurring in a physical classroom.
One special way of garnering prompt engineering skills consists of participating in a prompt-a-thon.
When I tell people about prompt-a-thons, their first reaction is often one of modest surprise. What is a prompt-a-thon, they ask curiously, and why would someone attend one?
It is very straightforward, as you will soon learn.
Prompt-A-Thon Defined
First, I’m sure you’ve indubitably heard about or maybe attended a hackathon.
Those are events or activities whereby people come together to write programs to do something in particular for the theme or goal of the hacking endeavor. It is quite common nowadays. Attending can be fun since you get to interact in a place and time to accomplish a usually reasonably sensible mini-project. Lots of satisfaction and personal networking arise. Fond memories are collected.
A prompt-a-thon is essentially the same notion, though instead of writing programs you make use of prompts while using generative AI apps. No programming is usually required. It is all about how to best compose prompts. There are good prompts and lousy prompts. There are prompting techniques that research has shown will tip the AI in a direction that can get you better answers.
The usual prompt-a-thon does several things all at once.
It provides a forum for people to learn about prompt engineering and actively do so in a hands-on mode. A key takeaway is that you end up being able to write more effective and efficient prompts. A secondary or maybe equal footing goal is that you typically focus on some specific realm to hone your proficiency, such as using generative AI for medical uses, financial uses, legal uses, and so on. A third goal is the camaraderie of those you meet and potentially establish collegial friendships that will last throughout your career.
Based On A Chosen Domain
I would estimate that most of the prompt-a-thons that I know of or have been involved in are usually focused on a chosen domain.
Here’s how that works.
Suppose a prompt-a-thon is put together that entails the healthcare domain. The activity will be a day or two days in length and focus on how to best compose and use prompts when it comes to the healthcare realm. Attendees are going to be professionals in healthcare. They know that AI is coming to their desks soon and want to be ahead of the game. Or they are already experiencing generative AI in the workplace and want to do a better job at leveraging AI.
Sometimes a hack-a-thon is undertaken by a company and solely attended by its employees. No outsiders allowed. In other cases, a prompt-a-thon is put on widely and attracts people from all walks of life who are interested in the specific domain.
If you use a publicly offered prompt-a-thon, make sure to look beforehand at how they have structured the event. How hands-on will it truly be (sometimes it is all talk, little or no walk)? Who else will likely be attending? Will the instructor be flexible and accommodating or rigid and act as though a schedule must be strictly kept? Are there any hidden fees besides a base registration fee? And so on.
Do your due diligence.
Recommendation On Structure
Prompt-a-thons vary in how they approach the prompt engineering activities that will take place.
My preference is that the course be explicitly planned and contain at least these five elements:
- (1) Overarching Prompting Goals: A list of the specific goals that the prompt-a-thon is designed to meet — laid out so it is readily apparent where the session will be heading throughout the endeavor.
- (2) Use Of Mini-Project Cards: Various prepared mini-projects that can be used by the attendees should be made available, though not as a lock-in necessarily. In other words, the mini-projects suggest what kinds of focus attendees ought to have but they aren’t limited to just the ones provided. If participants don’t have something else in mind, they are welcome to use the provided depictions.
- (3) Curated Datasets Available: The organizer should provide ready-to-go datasets that can be used during the prompt-a-thon. This will be wise since otherwise much of the time could be inadvertently consumed by finding and wrangling with data. The exception would be if the course is about prompts that entail finding and refining data, in which case the curated datasets might not be as crucial.
- (4) Team And Individual Efforts: I prefer that, if possible, there be team exercises. I’ve found that when attendees interact with each other, the experience gets magnified. I’m not saying there shouldn’t also be individual-only activities, nor am I saying that if team setup is limited you shouldn’t attend. All I’m suggesting is that teamwork and prompting are synergistic.
- (5) Plenty Of Materials And Mentoring: There should be workbooks and guides provided to the attendees. I’ve attended a prompt-a-thon where the instructor did everything verbally and on the fly. I don’t recommend that approach and urge that in-hand materials be provided. Also, some courses have a facilitator who is not an expert in prompting and just tells the attendees to do this or that as based on the guidebook. Sad face. I believe there ought to be an instructor who serves in a mentoring capacity.
Not all prompt-a-thons will have those five elements. Don’t assume that they do. There are ad hoc prompt-a-thons, which, if you know that’s what you are getting into, well, fine, but otherwise my suggestion is to ensure that the activity is planned and hopefully will get carried out according to the plan.
Example In Healthcare
To illustrate how a prompt-a-thon often proceeds, assume that there is a healthcare-oriented prompt-a-thon coming up that catches your eye (this is going to be roughly the same as if it was a finance one, legal one, etc.).
Let’s start with the stated goal of the prompt-a-thon:
- “This prompt-a-thon is for healthcare practitioners who are using or going to use generative AI while on the job in a healthcare practice. You will learn to devise suitable prompts to generate medical summaries and patient notices. Ins and outs will be covered. You will be doing prompting during the prompt-a-thon to make use of the best practices described. Teams of participants will be assembled to undertake a healthcare mini-project. Open time will be available to pursue your specific interests related to the specific needs of the healthcare organization that you are employed by.”
The organizer noted too that data will be available:
- “Curated datasets of fake healthcare data will be provided for use during the prompt-a-thon. The data will be highly realistic and similar to the nature of data you would find in actual practice. The prompting will encompass dealing with data that is at times fragmented, contains errors, and has other real-world issues.”
A sample mini-project in this context might be this:
- Mini-project Card: “Doctor and Patient Communications”. You are to create and then test out prompts that will use the curated dataset to generate summaries of doctor-patient transcripts. The aim is to devise proficient prompts that produce patient summaries of a kind that reduce medical jargon and provide clarity for the patient. An additional aim is to provide summaries for the medical doctor that contain needed medical jargon and retain the essence of the interactions but that provides a condensed version to minimize the time required by the doctor to sufficiently grasp what the medical session entailed.”
As an aside, for my coverage on how to use generative AI to generate summaries, including a medical healthcare example, see the link here.
Prompt-A-Thon As Startup Instigator
A few final comments and then I’ll wrap up.
Think further for a moment about traditional hackathons.
There is a general expectation that if you attend a hackathon, it is usually due to a keen interest in potentially doing a startup business. You go there with a kernel of an idea and hope to find programmers who can turn this into a quick prototype. Or you are a programmer who doesn’t have a clear idea of what to develop but wants to find others who have business ideas, and you can apply your software development acumen to building.
That isn’t quite the same for prompt-a-thons.
In my experience, most prompt-a-thons involve people who want to know more about prompting and see what else they can get generative AI to accomplish. This might lead to some startup ideas, possibly. This is rarer, but still possible.
Remember too that another growing use of generative AI is to have the AI generate code. In that sense, if you were to attend a hackathon and from scratch develop code for a startup prototype, you can potentially do the same via the use of prompting. Make sure to check beforehand if the prompt-a-thon is supportive of that angle. If they aren’t, it would be an uphill battle to try and turn the endeavor in that direction.
Let’s close things with a mindful quote. Acclaimed American author, Vernon Howard, famously said this: “Always walk through life as if you have something new to learn, and you will.”
Go into a prompt-a-thon with your eyes wide open and your mind primed to learn. I can almost guarantee you’ll get something useful out of the experience and pat yourself on the back for having successfully undertaken a prompt-a-thon.