ChatGPT and the end of white-collar work as we know it — or maybe not quite

Article content

Brendan Caldwell was sitting at his desk in his downtown Toronto office on Feb. 6 to craft his annual Super Bowl party invitation. It is this piece of writing, more than anything else he bangs out over the course of a year, that the chief executive of Caldwell Investment Management Ltd., a financial services firm with a broad portfolio of clients, pours his heart and soul into.

Article content

The party is a tradition, and part of the fun for the past 16 years has involved the host setting down his thoughts and predictions for the big game in essay-like form, filled with punchlines, arcane historical references, arguments for and against the respective foes, and some insights and witticisms, for 200 potential guests to read.

Article content

This year, in a bid to stay “au courant,” Caldwell decided to do something a little different. Having read about the wonders of ChatGPT — an artificial intelligence chatbot developed by OpenAI LLC that launched in November and gripped the imagination of the planet by pumping out answers to user questions of all kinds — he ceded part of the invite’s authorship to the bot.

ChatGPT muddled some football facts, lacked in writerly flair and simply wasn’t as good as the genuine, exclusively Caldwell-written invite typically is. But judged purely as a piece of writing, its efforts were good enough (full disclosure: I am on the invite list), and it was produced at a rapid-fire pace that has some believing the bot is an existential threat to white-collar work as we know it and not just another gee-whiz phenomenon.

Article content

If an AI gizmo can write something decent and reasonably intelligent in the blink of an eye, it very well might replace human writers, marketers, lawyers, university professors, accountants, bankers and many other white-collar professionals who deal with the written word as part of their job. Or so the thinking goes.

It’s a big change from those ground underfoot by technological advances in the past. They were often the Joe and Jane Does of the world slugging it out on the assembly line, or driving the horse-and-buggy equivalent of their age into history’s dustbin.

But artificial intelligence is driving change at all levels of the economy, and the horse and buggy of today might just be you, the desk jockey with a university degree or two, and the unwavering certainty that you are not replaceable. Unless, perhaps, it turns out you are.

Article content

Noah Waisberg knows a thing or two about desks and the people who occupy them. He was once an associate at Weil, Gotshal & Manges LLP, a high-end, international law firm specializing in mergers and acquisitions and headquartered in New York.

Due diligence is a major part of M&A work, and that work traditionally involves junior lawyers reading through reams of contracts. It is mind-numbing, time-consuming, painfull
y repetitive stuff, but it represents billable hours, and due diligence accounts for a good chunk of a client’s fees — fees they aren’t always particularly happy with. Waisberg imagined there must be a better way.

“The system didn’t seem sustainable to me,” he said.

He was no techie, but the Toronto native met one in Alexander Hudek at a coffee shop on Queen Street West. The pair co-founded Kira Systems in 2011 and worked to harness artificial intelligence’s power to develop contract analysis software for law firms.

Article content

The resulting new technology didn’t herald widespread layoffs in the legal profession, but instead enabled lawyers to review a huge mass of contracts in a fraction of the time it previously took, and to provide clients with a broader, much more comprehensive legalistic bang for their buck.

“It allowed the lawyers to do the work that was actually worth $500 an hour,” Waisberg said.

Kira Systems counted 18 of the 25 largest M&A firms worldwide as customers when Waisberg and Hudek sold the company to North Carolina-based legal software maker Litera Corp. in 2021. The deal allowed them to keep a copy of the underlying technology, which Waisberg is now applying to contract reviews in the corporate space with his latest AI venture, Zuva Inc.

Article content

Waisberg and co. created a happy marriage between lawyers and artificial intelligence, and some believe a similar bliss can be attained with ChatGPT as it blazes a trail from an object of curiosity to an application companies will presumably be scrambling to employ for who knows what end.

But just nine per cent of Americans believe artificial intelligence will do more good than harm, according to a recent poll from Monmouth University in New Jersey, and 65 per cent believe university students are going to use ChatGPT to cheat.

One of the things about AI, and technology in general, is it tends to get better over time. Put another way, imagine how good something is today, and think about it being twice as good two years from now, and twice as good again two years after that, and so on. The ChatGPT of 2023 isn’t going to be the ChatGPT of the future.

Article content

“I think there is a point where you would be crazy not to be scared of that,” Waisberg said.

But the one-time lawyer sketches a scenario where AI is able to mine every single word a writer has ever written and then convert all that verbiage into the best of its best, giving the writer a starting point for the next story’s first draft. AI in this hypothetical scene could empower the writer to think more deeply about the article, and how a story should be constructed, and the essential points it should make.

In other words, fear not unnecessarily, he said.

Of course, the list of jobs that have slammed headlong into innovation in the past is lengthy. Anybody remember the milkman? Equally long, however, is the history of perceiving computers, robots and, more lately, artificial intelligence as harbingers of widespread employment doom destined to eradicate jobs on a massive, permanent scale.

Article content

This human capacity for panicking about the great technological unknown was in evidence at the highest levels of the United States government at least as far back as the 1950s, according to a U.S. Bureau of Labor Statistics report. The computers were going to take over, the narrative went, kicking workers to the curb and plunging America into a jobs’ crisis. The gadgets, of course, did take over — to a point.

Consider that a record 19.6 million Americans were employed in the manufacturing sector in 1979, according to the labour bureau. Those numbers have been in decline ever since, but the overall number of jobs has actually increased, marked by gains in the leisure and hospitality industry, education, health care, and the professional and business services sectors.

Article content

Losing your job stinks, and losing it to some newfangled gizmo must be doubly rough, but there seems to always be other jobs that need doing.

Humans also have something going for them, a built-in advantage artificial intelligence can’t match: o
ur humanity.

Artificial intelligence has billions and billions of data points to mine, and ChatGPT can spit them out in tidy, paragraph form, albeit with factual errors. But our brains allow us to problem solve, and think critically and analytically — thereby catching and correcting those mistakes — while our essence, as flesh-and-blood beings, lets us form bonds with other people.

We foster trust and camaraderie, inspire one another to fight the good fight, and figure out the hard stuff that needs getting done or altogether new stuff that nobody has done before.

Article content

“A lot of the time, in real-life situations, situations are complex — it is not a simple layup — and it takes human intervention to understand a problem,” Hubert Pun, a professor of management science at Western University’s Ivey Business School in London, Ont., said.

The calculator is a good example, he said. In the old days, addition, subtraction, multiplication, long division, rounding to the nearest decimal point and crunching numbers of all sorts was a labour-intensive human task essential to business.

Then along came the calculator, which wasn’t a bad thing since accounting types could suddenly do math more quickly, freeing them to use their analytical skills to find creative new ways for their employers to earn a few more nickels — and maybe hide a few from the taxman — and build better accounting mousetraps.

Article content

His point? Embrace change.

“It is up to you — the user — to use the tool optimally,” Pun said. “If you can optimize the tool, you’ll get ahead, but if you cower in fear in the face of it, and run away, then you’re going to become irrelevant.”

If you cower in fear in the face of it, and run away, then you’re going to become irrelevant

Hubert Pun, professor of management science, Ivey Business School

But grabbing hold of artificial intelligence isn’t just about saving oneself, according to Ed Clark, the former chief executive of Toronto-Dominion Bank and now chair of the Vector Institute, a not-for-profit dedicated to advancing AI research, and a partner at Radical Ventures Investments Inc., the largest AI-targeted venture fund outside of China.

Clark was a young civil servant economist, with progressive leanings, when he wrote a research paper for the Department of Finance in 1974 that said baby boomers would cause per-capita income to grow quite rapidly, since Canada had twice as many people of working age than retirees. The flip side of that, he cautioned, was the situation would be reversed in 2020. The boomers would be retiring, meaning a much smaller, working-age population would need to step up its productivity for income to keep growing.

Article content

Alas, Canada hasn’t stepped up. The country ranks 20th among developed nations in gross domestic product per capita and productivity, according to the Organization for Economic Co-operation and Development data, lagging behind a who’s who of Europe, the U.S. and the United Kingdom.

Clark recently reflected upon his government research paper from his vacation home in Florida, and not from a viewpoint of nostalgia, but from the position of a 70-year-old-plus “geezer” who is big into artificial intelligence. He sees AI as the “electricity” of our age and the answer to Canada’s productivity woes.

“We have this double pressure, where we are not doing as well in terms of productivity per worker, and we have the same problem all G7 nations have as an aging population, and so productivity really matters,” he said. “And along comes this new technology that has the opportunity to change productivity.”

Article content

Should the price of change be certain jobs, Clark believes employers have a “moral obligation” to train workers for other jobs.

“AI combined with humans makes the human a way more productive person, and, therefore, they don’t get laid off, they just become more valuable to you,” he said.

It is an upbeat line of reasoning, but we all know things don’t necessarily work out so well for the rank and file when moral obligations tangle with the almighty profit motive.

Back in Toronto, the investment manager who dabbled with ChatGPT for a Super Bowl invite was riffing on artificial intelligence and its potential impact upon his livelihood. Caldwell wouldn’t discount being made irrelevant by AI at some future date, but what he holds to be true in the present is that people would rather talk about “sex” than money.

Article content

Personal finances are among the great taboo topics. Discussing money, along with our hopes, fears and goals for our financial future, is ideally done with a human being one can trust. That person isn’t always going to beat the market, but they will at least pick up the phone when you call.

“We are talking about AI like it is magic, and it isn’t,” Caldwell said. “It is a tool, like the cotton gin or the steam engine.”

As for his ChatGPT-assisted Super Bowl invite, he figured it was worth a “Hail Mary,” but he isn’t ready to convert.

“What ChatGPT’s AI produced was like the NFL official who effectively decided the outcome of the Super Bowl itself,” he said. “However well intentioned, it did not ultimately produce a satisfactory result.”

• Email: joconnor@nationalpost.com | Twitter: oconnorwrites