As ChatGPT blows out its first birthday candle, it’s time to get real about the good, the bad and the ugly of Generative AI tools.
When my friend Julie* started a new job in a very different industry this year, she quickly found an unlikely ally in Max. If she was stumped by an insider-jargon term in a meeting, Max would quietly translate for her via text. Max could also simplify the complex stuff she needed to pick up quickly, or give her insights into issues on the fly.
Max also helped her come up with new cocktail recipes and dance playlists on a girls’ weekend – with absolutely no judgment.
But Max isn’t a new colleague, or an old friend with industry experience, or even a person.
Max is what Julie’s ChatGPT called itself when she asked it for its name.
She says she chats to Max almost daily, and only wishes he could do PowerPoint too.
On November 30, 2022, tech start-up OpenAI released ChatGPT to the world. It quickly saw the fastest tech adoption rate known to mankind, with 100 million weekly users by February 2023.
ChatGPT (and other generative AI assistants including its Anthropic rival Claude) are the first technologies to seriously disrupt knowledge work.
As my friend Julie quickly realised, ChatGPT is more than a very fast text generator. It can simplify really hard concepts – “explain it to me like I’m five years old” – or summarise meeting notes and create action items.
It might be the best intern you ever hired. And it doesn’t make typos or roll its eyes when you ask it to try again.
In January 2022, I pondered whether AI tools would become a valuable sidekick, or a bland-copy-churning supervillain. One of the things I worried most about was what might happen to the quality of my writing if I started outsourcing my thinking.
So here’s my take on a year of GenAI – as well as some thoughts from those putting AI to the test in different ways.
The good: a writing buddy when the words won’t come
I don’t have time for writers’ block: deadlines press down on me every single day. Now, instead of hovering above the Google search bar with the urge to type ‘help’, I have ChatGPT, Claude, or TextFX at my beck and call.
Need one more headline? Ask GPT-4. Honestly, its ideas will all be decidedly average – but one word could trigger a whole new line of thinking. Looking for better workshop discovery questions or activities? Claude, GPT’s slightly more helpful and honest cousin, has given me some useful thought-starters. Can’t think of an alliterative word for a headline or a simile for a visual metaphor? TextFX is weirdly random.
The truth is, most people find writing hard and time-consuming. And I suspect that is what is really driving the rapid experimentation with AI assistants.
There’s the small business owner, who tells me he’s now saving $2,000 a month on blog writing. He says it’s been massive – he can now publish two blogs a day, and it’s transformed his search results. And he’s learned the better the brief he gives it, the better the results.
AI has also taken hold in the very human world of HR, where recruiters use it to write job descriptions, ads and generic replies. And it’s a gift for speakers. After delivering a compelling keynote this year (on AI), the presenter admitted to me GPT-4 had come up with her entire structure and some highly original analogies – turning what would have been a two-week job into a few hours.
Creative agencies are using it too. One of my favourite design partners says she used it weekly to brainstorm ideas, ask it to explain complex client strategies or products and find visual metaphors.
“I just wish it could do humour or wit better,” she tells me. “Even when you ask it to use a more approachable tone it feels a bit cringe!”
The bad: everything is starting to have a robotic same-ness
Ah yes, the cringe. It can be incredibly frustrating (and time-wasting) to keep pushing prompt after prompt and getting even more generic or nauseatingly clichéd responses.
There’s also not much of an advantage when everyone else has access to the same tech. At a recent HR tech conference, one recruiter shared that so many candidates are using AI tools as live-coach during interviews, all their answers sound exactly the same.
She wants to hire someone for who they really are, not who GPT makes them out to be in their keyword-stuffed application and pitch perfect pre-interview.
I suspect most clients feel the same about businesses they choose to work with.
And while the tech has certainly improved over the last year, its responses have a Truman Show-esque fake sheen. GPT-4 still makes stuff up. It has implicit biases. And it can’t share the vulnerable mishaps and universal truths that make a great, relatable story.
Not yet, anyway.
The ugly: it’s still not quite right
A client recently shared a blog draft with me for my opinion. Her usual freelancer hadn’t quite hit the mark with the copy but she couldn’t work out how to give that feedback. She asked me if I could re-work it, and what I would suggest.
My instincts told me that writer might have leaned a little too heavily on GPT. The article lacked human empathy, and it felt basic and unoriginal.
Writing is not just a case of typing one word after another. It’s about getting those words into the minds of other people, so your message can have impact or create change. And that’s the risk for freelancers who believe ChatGPT will be the magical answer to the eternal question: how can I earn more by doing less?
I polled my LinkedIn community to ask how they use generative AI, and was a little surprised that 23% are using it to write entire blog posts or emails. That’s a lot of bot-generated content out there already. While I have no concerns with using ChatGPT as a thought-starter, I can guarantee to all our clients that their words will be written by a human.
There has also been much hype about the new jobs that will be created in the wake of AI – like prompt engineers.
But I believe there are two skills everyone will need in the very near future:
- the ability to fix AI-generated content so it feels genuine, useful and original
- the ability to write better than a bot.
If you need help with either, we’re here for you.
* Not her real name
Unit 2543-53 Bridge Road
Stanmore, NSW 2048
Subscribe to the W. Insider
No junk. No horn-tooting spam. Just some useful tips and relevant updates. Every once in a while.