Machines Think. Humans Experiment.

If you've been in tech for a while, it can feel like Groundhog Day. Come in on Monday morning, and a CEO of some start-up is breathlessly announcing that they have reinvented air. Again.

That's why it was so refreshing to hear Katy Yam, director of marketing for the red hot Element AI, recommend that companies take a very pragmatic approach to artificial intelligence ("AI"), one of the hottest technologies of the 21st century (which has been around since the 1950s). 

Start small and try with AI. If anyone’s trying to sell you this huge AI package, be super skeptical.
— Katy Yam, Director of Marketing, Element AI
 
 

Yay, Katy. We're fans of starting small, and starting in general. You'll learn more from your mistakes than anyone who's still studying the textbook or moving Post-Its around endlessly on butcher paper. Whether it's chatbots or AI or voice-enabled assistants or augmented reality, just start. Identify a good use case or two, and run a pilot. Why is this so important? Because innovations build upon prior innovations. And if you don't muck around with the base innovation, you may never catch-up to the future applications.

For example, if you're a retailer who's not experimenting with augmented reality ("AR"), you may already be behind. That was the message from Roots, Snapchat, and Amazon, at DX3Canada. Amazon, hyper-focused on price, selection, and convenience, is using augmented reality to increase customer convenience - and shorten the sales cycle. Sure, IKEA did it earlier, but...this is...Amazon (most common question we get from retailers: "how do I avoid getting Amazon'd?"). Amazon Canada's President, Tamir Bar-Haim, explains from the stage at DX3Canada:

Amazon's been using one of the base technologies in AR, image recognition, since at least 2011, when it allowed ordering of products based on a scan of a bar code. That's since morphed into Dash, a wand that let's you scan a barcode or simply say the name of a product, then dumps it into your cart, ready to be delivered by...Amazon Fresh. Natch. This is not the newest, most breathtaking application of image recognition, but it's so damn...useful:

 
 

Other more recent, very practical applications of image/barcode recognition include OLG's May 2017 lottery ticket scanning app, which lets you scan your ticket to see if you've won, from your couch. Handy for lottery players, and handy for those of us in the line-up at Shoppers Drug Mart on Fridays. If you understand the "base" tech, bar code scanners and image recognition, you understand how an Easter bunny can pop out of the floor at a grocery store, courtesy of Zappar and UK grocer Asda. 

 
 

Why on Earth would a retailer want to interrupt someone's shopping experience by having an Easter bunny pop out of the floor? James Connell, VP Marketing at Roots Canada, explained in under 7 seconds:

It has to be more entertaining to come into a store than it would be to sit around at home in your underwear and buy a Roots sweatshirt.
— James Connell, VP Marketing, Roots

Yep, today's retailers are also in the entertainment business, and AR helps them do it. Now let's switch gears from AR to AI. AI's actually been around since the 1950s, when John McCarthy created programming language LISP and Alan Turing asked an important question and proposed the Turing Test.

I propose to consider the question, ‘Can machines think?
A computer would deserve to be called intelligent if it could deceive a human into believing that it was human.
— Alan Turing, 1950

And let's not forget Norbert Wiener, the MIT professor who in 1948 invented the field of cybernetics (computer technology as a means to extend human capabilities) and observed that feedback is a key feature in almost all learning (a foundational principle of machine learning). 

AI's been around since before most of us were born. It's just that now we have enough data, and access to data (the cloud), along with compute power (machine learning software) to accelerate it.

AI runs the gamut from regression analysis (which becomes more accurate with the more data you feed it), to machine learning (use of pattern recognition and algorithms by computers to learn new things) to neural networks or deep learning, and finally to artificial general intelligence (what Turing envisioned - and we are not there yet). For all the breathless fascination with AI, the current applications are oh-so-practical. Which is a good way to start. 

 
einstein vision.png
 

For example, Salesforce's Social Studio (formerly Radian6) now offers Einstein Vision, which detects images and the context they are used in. The application? Figure out where your logo is being used. Are influencers touting your product? Is it showing up in conjunction with certain events or places? Einstein can tell you. 

Other useful applications of AI are in hospitals and cities. Brian Donlan of Cisco explains how Oakville Trafalgar Memorial Hospital, working with Cisco and construction company Ellis Don, uses Internet of Things ("IoT") sensors and intelligent algorithms to monitor vulnerable patients.

Chatbots, such as 1-800 Flowers' Gwyn, are a very practical application of AI. Gwyn is powered by Watson, IBM's natural language-based AI platform. Gwyn helps users choose a gift based on occasion and the sentiment they are trying to convey, and she gets smarter with every interaction. Other clear applications of chatbots are frequently asked questions and customer care, particularly for repetitive questions not requiring a tremendous amount of empathy (ie: "what movie should I see, TIFFbot?").

Useful applications of AI abound. The key is to find those applications in your business, and get started.

Flashback to 1997. I am General Manager of Dell Canada's Home and Small Business Group, responsible for one of the largest advertising budgets in Canada. We've just launched one of Canada's first e-commerce sites and built a complex revenue model that factors in call center calls, abandon rates, close rates, average revenue per unit...and I am exhausted. Bill Sharpe, head of Dell's agency Sharpe Blackmore, runs in breathlessly. "Heather, we think if we build a regression analysis with all this sales call data that's coming in, we can probably figure out how to accurately predict the number of calls we'll get, based on seasonality, the offer, region, what media type we're using, and that sort of thing."  Me: "Bill, you're an ad guy. What are you doing proposing a data analytics project? Besides, there are too many variables - like colour vs. black and white, day of the week, size of ad. It's just too much data." #TrueStory. Within a year, Bill's crazy model was predicting calls to within +/- 3% accuracy.

We are all standing on the shoulders of giants, and innovations build on prior innovations. Best to start experimenting today. 

 
groundhog day.jpg