How soon will this AI impact you and your career, and to what extent?
Have you and your colleagues tried ChatGPT yet? As mentioned in my mid-December and year-end newsletters, the future is not upon us; it’s here.
Within five days of its release last month, more than a million users had signed up to take advantage of this artificial intelligence (AI). I was among the early adopters. Speaking with someone who teaches at a major university, it’s clear students are also jumping on the ChatGPT wagon. This will create another level of challenge in screening for plagiarism.
Scroll down to see how ChatGPT responded when I posed a question about its potential impacts on the assistant career
While it’s still early days, with room for increased sophistication, ChatGPT is already being touted as a valuable tool for speech and content writing. Think of its potential for drafting emails, reports, social media and marketing materials, travel planning, translations, research and more. Let’s also consider how this may impact your role.
Open source? What’s in a name? While ChatGPT is owned by a business named OpenAI, that does not mean it’s open source, which was my initial impression.
To learn how to use it, begin with play
Learning and gaining knowledge have serious, positive implications, yet we can have fun in the process. That’s how I’ve approached ChatGPT, beginning with a bit of play before leading up to more substantial questions.
You begin by creating an account. Then, tell ChatGPT what you’re after. Just don’t expect your results to be entirely accurate!
I directed ChatGPT to write a haiku about the work I do. Here’s what it delivered.
Training for success,
Minutes, resilience, change,
Tools for a brighter future.
Now, that may well reflect some of what I do, and the result did provide three lines that, in accordance with haiku writing, do not rhyme. So far, so good. However, the results didn’t adhere to haiku principles of having five syllables in the first and third lines, and seven syllables in the second line. As such, I popped in a reminder about how many syllables each line should contain and requested a redo. In ChatGP parlance, that means clicking a “Regenerate response” icon. What came up on the next attempt?
Assistants learn fast,
Resilience in change they grasp,
Minutes well managed.
With names like mine, you may be unsurprised to learn I then moved on to limericks. I requested one outlining some of my upcoming travels. I provided my name and some of this year’s upcoming destinations. ChatGPT came up with the following. The first iteration, which I’ve not shown here, did not adhere to the conventions of limerick writing. After the second attempt (“With sights to be seen galore …”), I gave further direction, resulting in the version describing me as a “traveler”.
Why did I want to edit the first and second versions? To begin with, the first limerick consisted of four lines rather than five. When writing a limerick, the first, second and fifth lines should rhyme, as should the third and fourth lines. I did not advise ChatGPT that, traditionally, the first, second and fifth lines of a limerick will have seven to nine (or, arguably, 10) syllables, and the third and fourth lines would have fewer … and you can see the system veered away from such norms. Conventions aside, one line in the first limerick read that my travels “were a joy to gawk”! When I responded, I instructed the AI to replace the word “gawk”.
Details matter, and this is where good human assistants can have positive impacts. Each iteration contained errors in tense, as I’d asked for a limerick about upcoming travels. That’s a detail, yet imagine using ChatGPT to write a report, only to have it speak about upcoming expenditures, plans or events in the past tense. This goes back to the importance of the human assistant proofreading all those deliverables.
Country- or culture-specific spelling: Depending on your country, ChatGPT’s use of a single “l” in “traveled” and “traveler” may or may not be deemed correct. So, if you and your colleagues rely on this AI, you may want to see how the system respond to instructions reflecting your language choices.
Translation services, anyone?
I often turn to DeepL for translations, and find it helpful. Here’s a look at ChatGPT’s translations of a simple statement. As you may guess, I’m excited to make this first trip to Cape Town, South Africa this March, and to seeing many of you when I present on Day Two of this year’s Executive Support LIVE Cape Town conference! This conference, hosted by Lucy Brazier and her team, will run March 24th and 25th.
If conducting research or relying on it for an article or speech, be aware that 2021 marks ChatGPT’s knowledge “cutoff”
I decided to try ChatGPT for information on this year’s inflation and recession trends. Below, you’ll see ChatGPT’s response included a statement that 2021 marked its knowledge cutoff. In this instance, that was a non-issue. If you were to use it to research medical (“What vaccinations should I secure before travelling to X?”), geopolitical (“Tell me about Russia’s attack on Ukraine.”) or other matters, you could potentially be relying on dated information. That said, I’ve seen ChatGPT make repeat references to its knowledge cutoff date, so our own common sense and independent thinking should also come into play.
If you direct ChatGPT to draft an article or report, as I’ve done, remember that, as of the date I’m publishing this article, its knowledge base –that cutoff – does not extend beyond 2021.
Does that mean we shouldn’t use it? Not necessarily. I’d be mindful, though, of current limitations.
You don’t like the first answer? Try again. As you’ll see at the base of the screenshot above, ChatGPT has a “Regenerate response” option. When you click on it, you’ll usually see a differently formatted response. This was the case even with math questions. When I asked the square root of 144, the initial response came in the form of a sentence. When I directed the system to regenerate its response, it came back with a simple “12”.
Knowing it’s imperfect, would you trust ChatGPT for travel planning?
Heck, we’re also imperfect. I gave this a whirl, and asked ChatGPT to provide a three-day itinerary for a visitor to my city, Vancouver. It came up with a good list, although the sequence of spots to hit on each of the three days appeared in an order that could sometimes see a visitor zig-zagging across the city. You could take that itinerary, though, and map out your routes on another platform.
If you’re looking for restaurants for hosting business meetings or special events, you’ll want to keep ChatGPT’s 2021 knowledge cutoff in mind. I posed questions about top restaurants in New York City, and about top pizza spots in Milan, knowing any restaurants that may have opened or closed since 2021 would not feature in the response … and that some restaurants that may have declined or risen in popularity since that cutoff date may be either inappropriately identified among leading restaurants or missing from such subjective lists.
As someone who’s designed, planted and sometimes (I kid!) weeded a couple of large gardens, I love exploring gardens when travelling. I asked ChatGPT, “What’s the best place to see beautiful flowers in England?” The resulting list of 10 gardens was impressive, yet I can think of at least two I would have included on such a list. Again, such matters are subjective.
What if your principal is heading to New York City for business, or you’re going there on holiday and you want to sort out transportation options? I gave that a try and found the answers helpful.
As with we humans, things sometimes don’t work. There also appear to be times of particularly high demand or other issues, as in the example below, that mean you won’t get your answer right away.
What if you or your principal had an appointment at 10 Downing Street in London? If you were taking the Tube/London Underground System rather than a car service to visit the Prime Minister or some of his staff, how would you reach it from a nearby Tube station?
Do you remember AskJeeves? Playing with ChatGPT had me thinking back to the early days of internet searches and how the emergence of Google and other search engines impacted earlier providers such as AskJeeves/Ask.com. To what extent might ChatGPT displace search engines upon which we currently rely? I decided to ask the system for a map to accompany those instructions and, as you’ll see, Google Maps are safe for now.
Hmm. So is this AI likely to – or designed to – supplant popular search engines?
What might this AI advancement mean for your career?
You knew I’d ask
I did ask, and here’s how ChatGPT responded. I anticipated, as should we all, the acknowledgement of potential for job displacement – another way of saying job losses – within the career. However, another part of the response had me debating whether to laugh or roll my eyes at the lack of awareness. I decided to challenge that response.
Chat GPT got the “repetitive” part right. Clearly, though, the response reflected a lack of understanding of just how complex scheduling can be.
I challenged the response … and received an apology, which we know not all humans readily offer!
Responses may vary
… even when you’re posing similar questions. As seen a couple of screenshots earlier, I asked, “What impacts is ChatGPT likely to have on the assistant career?” Have a look at the responses, below, when I subsequently tweaked the question to ask, “How might ChatGPT impact executive assistants and their careers?”
Do you subscribe to the response, “Increased focus on technical skills and less emphasis on interpersonal skills”? I do believe tech skills will continue to be highly relevant, yet I would be surprised if the emphasis on interpersonal skills declined. I do concur with the need for assistants to be adaptable, with the potential for job losses (“Decreased demand for human executive assistants …”), and with potential for AI systems to augment the work of humans.
What does ChatGPT know about yours truly?
I began by asking about this website. I figured ChatGPT may have some info, since I’ve been publishing Exceptional EA since 2013, long before ChatGPT’s knowledge cutoff. Have a look at the response.
What, then, did ChatGPT know about me? Read on.
Again, we can expect variations in responses to similar questions. I posed a similar question two or three days later.
Could ChatGPT be a ready source of contact info, or information on professional associations?
I began with info I could readily verify. ChatGPT missed some options, including my Eventbrite page, Instagram feed and more. The advisory at the base of the response was entirely reasonable.
I decided to also ask ChatGPT about some clients for whom I provide training.
I’m an IMA Training Partner, and decided to see what ChatGPT could tell me about this professional association, founded by Sonia Vanular in 1974, in Europe. ChatGPT offers options for readers to give a thumbs up or thumbs down response, so you can provide feedback on its responses.
Here’s what ChatGPT had to say about another couple of clients, Canada’s Association of Administrative Professionals (AAP), and England’s Strategic PA Network.
“I am just a machine”
Do you find yourself attributing human qualities to AI, or saying please and thanks when or if talking to AI devices? When I found myself beginning to thank ChatGPT for one of its responses, that led me to ask , tongue in cheek …
You’ll want to know …
You’ll want to know …
As mentioned, ChatGPT has limitations. It identifies them front and centre when you access your account. They are, and I quote, “May occasionally generate incorrect information”, “May occasionally produce harmful instructions or biased content” and one I’ve already mentioned, “Limited knowledge of world and events after 2021”.
Watch for my Weekend Poll on you and ChatGPT
Now that we’ve walked through my early experiences over the course of a few weeks on ChatGPT, please watch for my next Weekend Poll. I’ll pose questions here and on LinkedIn, so we can gauge whether ChatGPT has already begun impacting how you carry out your career.