My 2024 Musings On Learning #1 — Generative AI

In the spirit of lifelong learning, I had two major New Year resolutions this year (and one minor one). They pertain to self-development and growth.

First, I wanted daily to be a better version of myself than I was the previous day. Ambitious for sure, but gotta reach for the stars so I can at least land on the moon yeah?

Second, I wanted to be reading, studying, learning, writing, and (hopefully) understanding more about things I’m curious and/or passionate about. Then, to write deeply about them here and elsewhere. (A distant third is to pick up a hobby involving my hands — like pencil sketching. But that’s a story for another day)

This week, I began working on the second resolution when I attended a three-day AI (Artificial Intelligence) workshop online.

Specifically, I signed up for an entry-level introductory learning course on Generative AI (GAI) and Prompt Engineering. (For the uninitiated, traditional AI only performs fixed tasks; GAI however can learn and develop new content and outputs. I’ll say more about Prompt Engineering later in this post)

Ever since the public launch of Open AI’s ChatGPT in November 2022, the world has woken up to the reality that AI isn’t just science fiction.

But what the world perhaps doesn’t realize is that AI has been around for a long time. Long before ChatGPT exploded onto the scene.

The Imitation Game — Learning about Alan Turing

Movie buffs would remember British actor Benedict Cumberbatch’s turn as the renowned mathematician and computer scientist Alan Turing in the film The Imitation Game that hit the screens ten years ago. That was likely the first time Turing’s name was introduced to a whole new generation that’s been benefitting from his important works without knowing or learning about their benefactor.

Turing first came into prominence when he decoded German intelligence for the British government during WWII.

But even more importantly, according to Britannica, Turing was responsible for the earliest substantial work in the field of AI. In 1935, he described an abstract computing machine consisting of a limitless memory and a scanner that moved back and forth through the memory, symbol by symbol. In so doing, it read what it found and then wrote further symbols.

The actions of the scanner were dictated by a program of instructions that stored these symbols. This is Turing’s stored-program concept, and implicit in it is the possibility of the machine operating on, and so modifying or improving, its program. In short, it can self-learn, a distinct characteristic of sentient beings (or so we once thought). In essence, all modern computers are universal Turing machines.

The Thrills and Threats of a Learning AI

Thrills

As an introductory workshop on what Generative AI (GAI) is all about and what it can do, one would expect lots of demos and run-throughs on how users can harness the power of the likes of ChatGPT or Google’s Bard to provide answers to just about anything. And this three-day workshop didn’t disappoint.

From how to create designs of the wildest imagination to explaining complicated concepts simply, there’s certainly no lack of ways that GAI can wow even the most cynical of us. And there are so many platforms one can test-drive GAI. These include, but are by no means limited to: Bard / Bing Image Creator / ChatGPT / Dall-E 2 / Magenta / Midjourney / Perplexity / Stable Diffusion. (The best way to learn what each does is to simply DIY)

It’s quite thrilling to watch as the GAI platforms are unleashed before the eyes of workshop participants. I was especially wowed by how GAI can logically tabulate, in Excel, raw text data we provide in any word processing software. And to do it in mere seconds! Think of the time saved for the busy data entry clerk or intern?! (Do we even need them now??!!)

And Zoom, where the workshop took place, now has a built-in AI Companion that summarises meetings instantly. Meaning, no humans need to take meeting minutes anymore!

Threats

But precisely also because of these wonders, we must now reckon with the loss of many jobs humans used to labor over.

For example, the days of recruiting hundreds of volunteers for congressional campaigns in the USA, look almost prehistoric today. All you need is Ashley, a GAI created to answer questions and handle reciprocal conversations with thousands of voters that sound human-like in every call, everywhere, all at once.

And let’s not ignore the many accounts of professions like lawyers, accountants, customer service staff, traders, etc being replaced by AI in recent times.

To say there is an existential threat from GAI to jobs in every industry is no longer an overstatement.

An educator’s take on AI’s impact on learning

Of course, the threat of copyright infringements and plagiarism, especially in academic institutes of higher learning, is very worrying for someone like me who’s still teaching in these places today.

Will I be able to detect if a student’s work is his/her own? Can I develop and deliver more nuanced assignments to students that ensure they can’t cheat even if they have the latest GAI at their fingertips? Will my students even bother to learn if all they need is a friend called Bard to answer all their queries?

I don’t have the answers.

However, one answer I do have.

After attending this workshop, I was convinced of one thing, at least in the near term.

For now, GAI is still just a machine. It can only provide meaningful outputs if fed clear and descriptive inputs. In short, if you stuff rubbish in it, don’t expect roses to emerge.

According to the workshop facilitators, the way to ‘prompt’ GAI to give you the ‘right’ answers is to take note of five factors. They are contextual precision, directive clarity, inquiry elevation, example-backed guidance, and rules & constraints.

I’ve no time here to dive into each. But even if I don’t expound on each of these factors, it’s not hard to figure out that context and clarity are critical when issuing requests to GAI platforms. Users shouldn’t stinge on sharing as much data they have as possible with the AI if they want the best possible answers. (Although some argue we’ll be surrendering too much data to feed AI’s bottomless tummy, making ourselves vulnerable to hacking and even identity theft!)

Clear Communications is even more vital now than ever

As a communications educator for over two decades, I am heartened by this fact, for it falls squarely within the realm of clear communication. Something I proselytize to my students year after year.

It’s also no surprise to me many AI platforms today hire those from the communications, linguistics, and humanities disciplines to be “Prompt Engineers” (a job unheard of until recently). To help design unambiguous instructions for the GAI to learn better and produce better answers. So learning and mastering effective communication skills is something that I firmly believe will certainly continue to grow in importance.

A comforting thought, when you consider the short attention spans and even shorter text messages, memes, and emojis we pass off these days as effective communications!

In conclusion, AI use will certainly grow in this new year and beyond. So as I continue to learn more about its wonders (and woes), I’ll share more in future posts.

So stay tuned ai?!