If you’re like me, you’re probably fatigued by the ubiquity of ChatGPT coverage. You can’t escape it. It’s splashed across our broadsheets, you’re confronted with it in the trade press, and you can barely swing a cat by its tail without hitting a LinkedIn boast about some content experiment that’s being conducted. It’s exhausting—and yet here I am, adding to the clutter.
Using a simple query interface in the form of a free text field, ChatGPT extracts meaning from the petabytes of data across the internet, returning outputs based on natural language processing capabilities built into its learning model. For example, a basic query that asks “what’s the secret to making good sourdough bread?”, will return concise and perfectly logical reasoning for how to best achieve the task at hand. The model is intelligent enough to deconstruct complex query strings, without losing syntax or grammar.
Learning models aren’t anything new. They’ve been in existence for decades, ingesting data inputs in gargantuan volumes, and burning computational resources to train the models into a semblance of logic we can extract value from. Every bot, every virtual assistant (hey Alexa!) and most recently, every AI-powered conversational interface has been the precursor to this moment in time we’re seeing unfold upon us.
So, what’s changed?
Anyone who works in tech is familiar with the maturation lifecycle of technology development. As new technologies go from concept to nascent existence and into the mainstream, users tend to follow along with a similar adoption curve pioneered by early movers, before the greater public at large catches on. The point of intersection typically occurs when one or all of the following conditions are met; price elasticity drives accessibility, product readiness reaches the point of mass availability, and/or the tools hit an ‘ease of use’ threshold, where anyone can have a go thanks to limited barriers to entry.
Furthermore, there needs to be a perceived value of what’s on offer. OpenAI—the artificial intelligence startup behind the technology, has done an admirable job of capturing the lay man’s interest in deep learning. They did this by first launching DALL.E to the masses in 2021, where with just a few prompts, you could ask DALL.E to produce a complex visual image. At the time, the novelty value of being able to dictate a machine-produced image was extensively captured in the media, with amusing anecdotes written about art experts who were pitted against each other, challenged to decipher ‘real’ versus AI-based imagery. When ChatGPT came along in November 2022, it hit that intersection point I mentioned earlier, providing a quantum leap forward in our understanding of how AI can bring relevance and add value to our lives.
It’s been widely reported that Microsoft has ambitions to embed version GPT-4 into its Bing search tool interface. Imagine a reality where you no longer are expected to trawl through half a page of sponsored ads listings before taking a punt on any of the links displayed, hoping you’ll be sent straight to the information needed. Convenience aside, enhancing our relationship with how we use the internet has broad use cases we should be embracing, especially when it opens the door to technology inclusion. Using ChatGPT to communicate, access and share information among those with physical impairments or neuro-divergent traits, has the potential to improve the quality of life for parts of society currently being left behind.
Amidst the troves of commentary, there’s the usual camps of naysayers and supporters, going at each other with their own interpretation of where this will lead humanity. My brain isn’t big enough to process the full depth of these arguments, so I remain cautiously excited about the utility aspect of these tools. They play nicely into what is inherent to the job of any modern Marketer: an automated toolkit to facilitate test and learn for customer experience optimization.
So much of what we do as Marketers—from running A/B tests for email marketing subject lines, adding dynamic content to version communications, swapping imagery to test different audience reactions, writing and publishing helpdesk content to name just a few, is predicated on our ability to provide alternative outputs quickly and easily. This is the part I find fascinating. And while I appreciate it’s not a silver bullet, ChatGPT has enormous potential to short-cut some of the legwork to help get us from point A to B. I’m thinking customer service query handling, interactive product learning content, AI-powered personalized advertising, survey response sentiment analysis, all of this is within the realm of what’s just around the corner.
That is, if we can ever get back into the interface.
