How artificial is that intelligence? (And four other myths you need to bust when it comes to AI)

I asked my favourite AI app recently to map out a baseball road trip for an upcoming summer weekend. It came back with a great itinerary — except the recommended Pittsburgh game was in 2023.

The same evening, I tried to get a ride-share in downtown Toronto, only to discover the AI-powered app hadn’t factored in a series of conferences and concerts — and a streetcar malfunction — that sent wait times (and surge pricing) through the roof.

Small frustrations, for sure, but an indication of how imperfect artificial intelligence is right now, and how patient we’ll need to be on our meandering journey through a brave new AI world.?

In the original Brave New World (1931), Aldous Huxley wrote about the dangers of too much technology and not enough humanity. Nearly a century later, we’re still struggling to find the right balance between human judgment and technological efficiency. Add beauty and purpose to the human side of the ledger, and the need for balance is even greater.?

I’ve been thinking a lot more about this while engaging in a flurry of late-spring tech conferences (cue the traffic) exploring the bold projections about AI, and particularly Generative AI.?

RBC Economics and Thought Leadership is trying to add to the conversation with a new research paper on organizational readiness, examining how far behind Canada sits on the adoption curve. It’s a matter of national concern, given the need for more advanced technologies to improve our productivity. For organizations, in both the private and public sectors, there’s also a clear need to invest more in AI, to stay competitive.??

But equally, there’s a risk that many enterprises will be disillusioned with the results. Or worse, they may develop dependencies on AI systems — to run a call centre, for instance — that won’t be so easy to unwind.?

Our research found:

— only one quarter of Canadian businesses are even considering Gen AI, with fewer than 10% actually working with it;

— Canada’s historically slow rate of tech adoption is holding the country back from rapid gains that others, especially the United States, are starting to see;?

— the prevalence of small businesses in the Canadian economy is another barrier to adopting and scaling Gen AI;

— we need to play to our strengths in professional and financial services, as well as health care and education, as they have large data sets and clear use cases;

— manufacturing may be the greatest opportunity, with Gen AI systems beginning to transform the supply chains that Canada has long been a part of.

To help Canada strengthen our competitive position in the application of AI, we’ve teamed up with the Creative Destruction Lab , and over the next year will jointly engage with a cross-section of sectors and organizations to shine more light on some of the opportunities, and address some of the doubts and myths.

CDL devoted a recent day-long AI Supersession to help with that, including an excellent myth-buster session with its founder Ajay Agrawal and 美国麻省理工学院 professor Sendhil Mullainathan , one of the leading AI authorities in the U.S.?

They laid out four myths that every business needs to come to grips with:

1. AI is accurate. It’s not, and that’s ok, as it still has a better batting average than most of us. “We are looking for immaculate perfection,” they said. “We hold machines to a higher standard, and dislike machine harm a lot more than human harm.”?
2. AI is good at menial tasks. Not always, even when it comes to quick, simple, repetitive tasks, like telling us what to wear. ChatGPT, and other large language models, not only are at risk of hallucination — that is, drawing conclusions without good information. They’ve led us to hallucinate that they can take over the mundane.
3. AI needs to learn more from us. In fact, we may be the ones who need to do the most learning. We’ve gotten used to apps and body sensors telling us how we’re doing physically. Imagine getting real-time feedback on pretty much anything we do electronically, from calendar management to email efficiency.
4. AI amplifies human bias. It’s a fashionable argument, with little evidence — not least because we rarely understand and define our biases. AI manages to step right over them anyway, to get to clearer decisions.

That word “decision” may be the most important part of the AI journey. Our economy is nothing more than the sum of decisions — by consumers, workers, investors, capital allocators and regulators — and AI is increasingly informing, guiding, even making those decisions.?

Think how the Netflix or Spotify algorithm have impacted your entertainment decisions, for better and worse. Think again how that might scale with grocery choices based on your physical and financial health — and how that may, in turn, impact purchase decisions from food producers and farmers.?

It is indeed a brave, and concerning , new world.?

In that world, can we ensure our integrated decisions with AI are both productivity-enhancing and prosperity-sharing??

We will need to ensure we’re feeding the right amounts and quality of data into Gen AI systems, and fairly allocating computing power (and the associated electricity demands) to the most pressing needs. Most importantly, we will need to continue to elevate human ingenuity and our ability to judge, assess and critique, so we can hold AI, and each other, to account.