When Will AI Replace Me? Not For A While.

Image by Steve Johnson on Unsplash

I don’t know what it says about my personality that I’m always polite to a fault when I ask ChatGPT to do something for me. Maybe it’s because I’ve read too much science fiction and listened to too many apocalyptic podcasts about the looming robot takeover of humanity that I want to remain on the good side of my future overlords. Or it could be that I’m just painfully Canadian.

I use Chat quite a bit. I use it when I’m procrastinating on tasks I really should be doing. I enjoy throwing tough assignments at it, like this morning when I asked it to write me an essay in Japanese on the impact of social media on people’s mental health. It delivered – very competently I might add. I then asked it to translate the essay into the Yukaghir language. While it not surprisingly told me it wasn’t equipped to deal with this obscure, endangered language, it at least recognized that Yukaghir is a language indigenous to the Russian Far East and gave me what I assume is a competent Russian translation. (I don’t speak or read Russian so I’m in no position to comment.)

I also use it for actual work. I’ve found it particularly useful as an aid in producing long, tedious-to-write content of the sort that goes into project proposals. Ask ChatGPT to write a response to an RFP for a website migration and revamp for a government agency and what you’ll get is a massive amount of content that covers all the important bases of a website upgrade project proposal, complete with its best guess at a timeline and budget. The names, numbers, and other details are not going to be correct, but for writing like this, which is by its nature formulaic, Chat is a brilliant time-saving device. It takes a lot less time to edit an existing document, fixing the actual information as you go along, than to write one from scratch.

I also use Chat to help me write emails, another chore I find tiresome. At the end of the work day I’m officially out of brilliant things to say to people in an email and I lean on the AI for ideas.

Like many people in my profession, I initially felt threatened by the advent of ChatGPT. That is until I actually started using it and quickly found its limitations. For example, I once asked Chat to write me a haiku on the subject of web accessibility. This is what it came up with:

Web's open gateway,

Inclusion guides every click,

Accessibility.

Let’s talk about this for a moment. Firstly, this is objectively bad poetry. It manages to be both highly literal and ambiguous at the same time, while lacking any sort of rhythm or word combinations that sound good. Secondly – and this is what really surprised me – it’s not even a proper haiku structurally, unless there’s some way of saying the word “accessibility” in five syllables. I don’t expect poetic brilliance from AI (at least not yet), but I at least expect it to be structurally correct. This one’s an abject fail.

I know what you’re probably asking: could you do any better? Hold my beer:

Keyboard enabled,

Alt text beckons, colours guide,

Internet for all.

Not Shakespeare, perhaps, but at least it’s correct and moderately poetic. And unlike Chat, I managed to talk about accessibility without having to wedge the word “accessibility” into the text. And I managed to do it in under ten minutes, which, while an eternity in AI terms, isn’t bad for a human writer. And all the speed in the world is useless when it doesn’t get you to the right destination.

This may seem like a frivolous exercise, but Chat’s response to this simple task told me everything I needed to know about AI and its potential to threaten my livelihood. ChatGPT can do a lot, and indeed I’ve come to rely on it for some regular heavy lifting, but from the standpoint of crafting nuanced, emotionally intelligent writing, our current AI has a long way to go. Chat even said so itself when I asked it if it was a threat to human writers. In its own words, “Creativity, intuition, and empathy—hallmarks of human cognition—are difficult for machines to replicate fully. Therefore, rather than viewing AI as a threat, writers can leverage these technologies as valuable tools to augment their creative process.”

While I can’t help but worry that Chat is simply telling me what it knows I want to hear, it echoes what I’ve seen of AI-generated writing thus far. It’s always factually accurate and word-perfect, and for some forms of writing this is more than adequate. But when it comes to essays, blog posts, or any other form of writing meant to be truly absorbed, it’s generally obvious when something has been written by a human being versus a machine. For one thing, human-generated writing is generally replete with personal anecdotes or observations, much like this blog post is. For an AI to write what I’m writing here, it would take a quantum leap in artificial sentience, and I have no reason to think this is immediately forthcoming.

Of course, I could be completely wrong and we’re in fact only a few years or months away from AI that can write better poetry than Leonard Cohen and more gripping page-turners than Stephen King. Even if this happens, though, I still have to believe that human beings will continue to value communication with other human being, unmediated by machines. While I’m confident that my cognitive abilities surpass those of my dog, I know that when she gets around other canines at the dog park or the Humane Society where we take her for agility training, she interacts with them in a way she can’t with me. I’m simply not a dog and will never quite be able to see the world through one’s eyes – or nose, as is more pertinent.

In sum, no, I’m not particularly worried about AI putting me out of work. Moreover, if and when AI becomes sentient and essentially omnipotent, I figure I’ll have more to worry about than my capacity to make a living producing written content. And I’ll be far from alone; we’ll all be out of work and at the mercy of our robot overlords. Which, again, is why I make a point of being polite to Chat. I even refrained from giving it grief over its terrible poem. We’ve all written stinkers in our time. It’s how we learn.