It may seem crazy, but the world’s biggest retailer is working on a scheme to use artificial intelligence to decide what you will want to buy before you’ve decided to buy it and send it to your home.
AI, after years lurking in the background, is about to get significantly more intrusive.
And according to the experts, despite any protests to the contrary, we’re going to get used to it.
Sometimes called “deep learning” or “machine learning” to distinguish it from the science fiction monsters that wrench power from human hands, the science of algorithmic prediction has suddenly become more powerful and pervasive than most of us realize.
But as Canadian artificial intelligence whiz Ajay Agrawal, co-founder of the startup Kindred AI, says in his new book, Prediction Machines, the power of machine learning is only half the reason it’s worming its way into so many parts of your life.
Powerful and cheap
As an economist, Agrawal insists the second, and possibly more important reason, is that the cost of AI is plunging. And as he says, “Cheap changes everything.”
Agrawal, who among many other things is also a professor at the University of Toronto’s Rotman School of Management, says if it isn’t already, AI is about to get in your face. AI will be telling you what to do and it will be telling you in its own words.
For search engines, 80 per cent accuracy is fine, but the artificial intelligence that runs self-driving cars must be nearly perfect.(Paul Chiasson/Canadian Press)
“I think what will be, over the next 24 months, the most noticeable AIs that the average person will deal with will probably be in those two formats, through recommendation and through chat,” he says.
The process has already started as shopping sites suggest what you should buy next and Siri responds to your questions, but expect it to be more obvious and more widespread.
Most people who don’t think they use artificial intelligence are merely those who don’t realize they are using it.
Until now, AI has flown under the radar in most of our lives, doing things such as preventing credit card fraud, filtering spam and smoothing out the action in video games, to mention just a few examples.
‘AI first’ strategy
No longer merely an interesting research project with an exciting future, companies large and small are adopting a strategy of “AI first.” And that will change everything.
“If you want to understand the priorities of a technology company, first look at the seating chart,” the New York Times reported earlier this year.
The boss of Google, Sundar Pichai, had reorganized the corporate seating plan so that he was surrounded by the company’s artificial intelligence team, demonstrating the No. 1 priority in the organization. Other technology companies have done the same thing after deciding AI is the most important key to their success.
“What doesn’t generally get as much press coverage is that when something becomes first, something else becomes second,” says Agrawal.
“Effectively, what it means to be AI first, it means that a company is prioritizing prediction accuracy above everything else, so above short-term revenues, above user experience,” he says. “Potentially, above privacy.”
A conversation among Google employees on the Silicon Valley discussion site Quora reveals the reason why. Whereas an 80 per cent chance of getting the result you’re after is good enough for Google searches, it simply doesn’t make the grade when AI is booking your flights or operating your self-driving car.
We are talking about a goal of close to 100 per cent predictive accuracy, which is astounding.
That accuracy is also why online retail giant Amazon is working on a plan to send you the products you were just about to order. It has already registered a patent for “anticipatory shipping.”
Amazon is working on a scheme to send you tomatoes and toilet paper before you know you want them — and experts predict you’ll end up liking it.(Ben Nelms/Reuters)
Based on reams of data from you and billions of dollars of other purchases, Amazon expects to be able to so accurately predict what you want to buy that it will simply ship the goods to you and arrange pickup for anything you decide not to take.
That is the extreme form of the AI-driven process referred to as recommendation, where a predictive algorithm will know before you do when you are going to run out of tomatoes or toilet paper and what movie you want to watch on your home entertainment system.
Better than monkey brains
“It might actually bring more variety into your movie consumption than you might have if you just leave it to your own monkey brain,” says Agrawal.
The other place where AI has begun to compete with the mere monkey brains that invented it is conversation. Now, that person you are having a conversation with on the phone may actually be a machine.
In one example, the Canadian startup Ada Support was able to handle 94 per cent of customer questions with AI, referring only six per cent to human operators.
Google’s Duplex, an AI-based phone assistant, is so good at idiomatic English, as portrayed in this example, that there have been demands that AI technology be required make it clear in any conversation that you’re not speaking with a human.
But will you accept being told what to do by a computer? At first, no doubt there will be a lot of resistance. Current AI strategies often include something called “human in the loop,” where the technology offers a series of suggestions and the human makes the final choice.
But Agrawal says once people realize they are getting good advice, they get sick of saying yes, yes, yes, yes.
“I think as the machines make better and better recommendations, we’ll be more and more willing to just give up a lot of these decisions to the machine.”
Follow Don on Twitter @don_pittis