The AI Future, Judging by the Past

One question that even ChatGPT can’t answer is about the future of artificial intelligence. Technological futurism is littered with widely off-the-mark predictions. In 1977, the CEO of DEC, then a leading computer company, opined that “nobody would ever want a computer in their home.” Other calls were much too optimistic, or still unfulfilled; where’s my flying car, our fusion power? Still, while AI forecasts are similarly challenging, if history does repeat itself, it offers some lessons for both investors and economists.

First, in most cases, there have been long lags for the full economic consequences of a new technology to show up. The Nobel economist Robert Solow famously noted in 1987 that “you can see the computer age everywhere except in the productivity statistics.” Often, the first innovation — the computer — requires subsequent investment spending and further innovations before it truly flourishes. When Solow spoke, firms were just buying their first desktop computers, software was unfriendly to the non-specialist, and business computers weren’t hooked up to the internet until the middle of the next decade. There was another long wait between the adoption of the internet and today’s widespread use of virtual meetings.

That lag has us skeptical about some of the widely circulated forecasts for the GDP lift from AI. A PWC study, for example, cited a 15% boost to North American real GDP by 2030. That would entail a roughly 2% annualized acceleration, essentially a doubling in trend growth, that would require a much quicker full flowering of the technology than we saw in the dawn of computers or the internet.

Second, we’d caution equity investors that first-mover advantages aren’t what they’re often cracked up to be. That’s been true in spreadsheets (anyone remember Visicalc, Lotus 1-2-3?), search engines (Lycos, Altavista), browsers (Netscape) desktop computers (Xerox, Altair) and cell phones (Blackberry, Nokia). Suarez and Lanzolla (Harvard Bus Rev, 2005) found that first movers typically fail to gain a lasting edge when the market and/or the technology develop rapidly. That might well characterize the path for AI-related goods and services, so choose your high-flying AI stocks with due care.

Part of that risk is tied to the chance that something even better comes along that supplants the types of AI applications being considered these days. In the tech world, there’s a graveyard of once-dominant products, and related companies, that fell victim to a better mousetrap. We still have cell phones, browsers and desktop computers, but there’s not much of a market for fax machines, word processors, or CD players these days.

Other AI forecasts drill into the implications for workers and careers. An IMF study estimated that 60% of advanced economy jobs will be impacted by AI, equally divided between those that AI will help by improving worker productivity, and those where workers will fully be replaced by AI tools.

But history shows that when it comes to technology, it’s always been thus, and we shouldn’t worry about ending up with a 30% jobless rate. Technology change shifted massive numbers of workers off of farms and into industry, and then in turn shrank the labour needs of the manufacturing sector. But these changes freed up the labour needed to expand the service sector into the major employer it is today. Banks have far fewer tellers, but more complex products that couldn’t exist without IT investments have meant offsetting job creation in higher value-added and better-paid positions.

In the 1970s, the typical office building was filled with secretarial typing pools and filing clerks whose jobs vanished as people used desktop computers to type their own reports, and filing cabinets were supplanted by electronic documents. But consider all the white collar jobs that were created in the internet era: social media managers, web designers, TikTok influencers, data analysts and so on. Canada’s unemployment rate was at multi-decade lows in early 2023, even without all the jobs that various technologies replaced over time.

And we’ll still need economists, even if an AI system could summarize today’s payroll report. This commentary wasn’t written by ChatGPT; some of us still like to think for ourselves.