Let’s get something clear. This opinion piece is not the work of ChatGPT. Come to that, no op-ed with a human byline should be written by any generative AI tool.
People want to know what the human (or humans, in the case of this joint article) think about the topic they are writing about. And they always will.
With that disclaimer out of the way, let’s address the DALL.E-designed elephant in the media room.
Generative pre-trained transformer (GPT) models will do much of the work that we, in communications and media, are doing today within the next six to 12 months. If it isn’t already.
And, let’s be honest, people are concerned.
Many see it as a threat to job security. Some see it as a bigger existential threat. Are we irreversibly headed towards societies run by self-conscious, self-preserving bots as often played out in sci-fi novels and movies?
Read: Here’s how BigTech is responding to ChatGPT’s popularity
But the English textile workers of the 19th century – remembered by history as ‘Luddites’ – felt the same way when their industry was presented with sewing machines and the like. Despite alarmist warnings that the end of human labor was at hand, it wasn’t. The labor market evolved, human skillsets advanced, and the quality and efficiency of textile products have only improved since.
So, should we take arms against this sea of machine learning-induced troubles and destroy ChatGPT & Co. where they code? (Would ChatGPT know how to weave in a reference to Hamlet’s famous soliloquy here, in a way that makes sense to a modern reader?)
In a word, no. Rather, this is a golden opportunity to propel our productivity, complement our creativity, and enhance the efficiency of our working hours. But we will achieve this if we work with these new technologies. We won’t beat the GPT bots, so we must join them. And this requires a clear and practical plan of how best to optimize them, and how to make them work for us.
This plan needs three things. The first is an understanding of the landscape. Second, is optimizing it. And third is addressing the challenges it brings.
It’s already a crowded house
Though ChatGPT has grabbed most of the recent AI-apocalypse headlines since its launch in November 2022, it is just one of a plethora of tools disrupting communications.
In the language-generation domain, Google has already launched Bard, with Sparrow soon to follow. Anthropic’s rival to ChatGPT Claude will soon be open and available. And the Chinese multinational tech company Baidu will soon unveil Ernie. And then you’ve got the supplementary bots like QuillBot and Word Spinner, which take your AI-generated text and rephrase it so it’s almost non-detectable as AI-generated text – handy for essay scribbling students. Equally handy for speechwriters and editors working on version 71 of a keynote due to be delivered in a matter of hours.
In the audio space, there’s Voice.ai, which benefits from the backing of multi-Big Tech CEO Elon Musk, and which will mimic any voice in the world. Handy for advertisers who want a low-budget Morgan Freeman to do their voiceovers. And Valle, another Microsoft creation, will arrive soon with the ability to mimic any voice in the world (including yours) after listening to it for only three seconds.
And in the creative space, we’ve all seen the mesmerizing outcomes from the likes of Midjourney, Stablediffusion, and DALL.E. Perhaps less known is Dreamix – a video editing tool that can make video edits quicker than you can brief a designer. Good news for the last-minute changes to on-screen text, background music choices, and cropping out the back of people’s heads.
A return on your prompts
So, how do we harness this game-changing tech?
Firstly, you develop your own – like we’re doing at APCO Worldwide with our AI Comms Lab. It’s been rumored that some developers will consider offering their GPT models to companies to customize and optimize. When it comes to building capabilities, upskilling talent, and focusing your efforts on strategic outcomes, this is nothing but a good thing.
One key thing we will need to focus on is change management. There has to be a cultural acceptance that this technology will shift our working dynamics radically. But that’s no reason to dismiss it.
Interestingly, one of the most useful prompts when using ChatGPT is to start by telling it: “Believe you are an expert in (insert field here).” The text that is subsequently generated sounds much more like the human profile you are seeking to mimic, than it would if you start your prompt with: “Draft me…”
We must follow the same prompt. We must believe that this tech is here to help us, rather than replace us. When we accept this, our approach will change and improve.
Staggering as these tools are, there are multiple challenges and limitations.
Though GPT models have some contextual awareness, they still lack a comprehensive understanding of the world around us. For instance, ChatGPT generates responses based on the input it receives but does not have access to all the relevant information and context that you, as a well-read, expert in your field, do.
Further, because ChatGPT is trained on vast data sets to recognize patterns and relationships between words and sentences, its human-sounding output becomes somewhat predictable and monotonous.
This human ‘sounding’ text is also, and importantly, statistics-based. And “statistically-likely sentences are not the same as verifiable truths,” as the World Economic Forum has been quick to point out. We cannot assume that the ‘facts’ ChaptGPT offers have undergone the same rigorous research or journalistic processes that our professions hold as gold standards.
In response to this point, ChatGPT said: “You can trust that I will do my best to provide accurate and relevant information based on the data I was trained on. However, it’s important to remember that I am an artificial intelligence and may make mistakes.”
Fact-checking, and injecting the right tone, style, and strategic thinking, make the human intervention of an industry professional essential.
But this throws out some ethical concerns – who is training the bots, and who gets to decide what is fair and ethical practice?
Biases are easily transferred from the data or algorithms these models are trained on. Privacy is a big issue. Where does the data you are inputting to ChatGPT (an open-source AI tool) go when you’ve entered it into the system?
Remember, when you don’t pay for a product, you are the product. The last thing communications firms want to see is sensitive information become a free-to-access product for anyone with internet access. And then, there’s the hotbed of transparency and accountability. If the AI makes a mistake, who is responsible? You are.
If you are using a GPT model, you are ultimately responsible for the final product that you send out into the world. You should be ready to defend what you have created. This is why your final review and edit are indispensable.
Media and communications are not alone in facing this great wave of disruption. It is one of many industries. But it is perhaps the best placed to harness the change. And so, in the words of the human writer Alan Watts, “the only way to make sense out of change is to plunge into it, move with it and join the dance.”
For more on technology, click here.
