• by Mike Azzara , Op-Ed Contributor, June 1, 2017

Being a fast-follower may not work when it comes to artificial intelligence. That’s a potentially huge problem for the vast majority of companies — the mainstream of American and global businesses that have watched and waited as wave after wave of new technologies emerged over the course of the last 20 years.

From the PC to “local” networks to the Internet to cloud, mobile and social. For each of those tech waves you could turn out fine if you waited, learned from the pioneers in your industry, and then followed quickly with the benefit of their lessons. Common principles emerged from early adopters of cloud computing, the mobilization of work, and the socialization of business that enabled fast-followers to start out, on day one, at a much higher level of implementation sophistication than their predecessors.

But with learning technologies like AI, you’re not the only one who needs to learn lessons — the technologies need time to learn, too. And every company’s situation is different, especially when it comes to marketing.

While there will certainly be lessons learned that “raise the floor” in terms of foundational AI/machine learning technology infrastructure, it simply does not stand to reason that application-level progress of company A (or companies A through Y) will be of any benefit to company B (or Z).

A very closely related challenge that many companies are running into as they consider AI initiatives is the “toddler problem,” according to Mike Nicholas, a founder of Born, the new AI-focused agency subsidiary of MDC Media Partners launched in November 2016.

The toddler problem is this: AI programs start out as dumb models that must learn on the job in order to achieve greatness. The machine learning algorithms that today can determine, in fractions of a second, the best route for you to take across town, or across the continent, began their artificial lives as empty-headed toddlers. They tried and failed, tried and failed, etc., and each time learned a bit and became a little bit better. The more times they tried, the more they learned. After billions or even trillions of tries — interactions — they’re faster and smarter than any program anyone could have written.

Remember how “challenged” newborn Siri was to understand what we were saying? Well now there’s a $249 pair of headphones that will simultaneously translate real-time spoken language among English, French, Spanish, and Italian.*

The toddler problem is a big concern for brands because it leads to a chicken-and-egg problem (sorry to mix metaphors). AI requires a lot of data — i.e., interactions. You need thousands or millions of interactions before your AI agent (or chatbot) gets super-smart. The only way to get all those interactions is by exposing your toddler-level AI to customers and letting them hack away at it.

That, Nicholas says, stops most brands dead in their tracks. No way will I let a 3-year-old represent my brand! But then it won’t get smart.

Marketers who are used to fine-tuning their work to perfection before letting it into the light just can’t wrap their heads around this problem. Fans of human brain physiology and evolution will note that this is the same phenomenon which, in humans, requires our species to have a far, far longer childhood than any other species on earth.

One approach some businesses are trying is to stand up an internal AI to support customer service reps. That way, the rep becomes a human filter between the AI’s recommendation and the customer, ostensibly preventing the stupidest mistakes from damaging the brand. The AI learns through interactions with the customer-facing humans.

The challenge here, though, is getting enough interactions, quickly. The brands with the guts to release toddler AIs to the world will get a first-mover advantage that probably overcomes all but the worst reputation-damaging screw-ups from the toddler.

As my AI Insider column partner, Sarah Fay, pointed out in her keynote talk at MediaPost’s Marketing AI conference, when you think about it — and when you do the math — first-mover advantage in the AI era will be virtually impossible to overcome. Which brings me back to my opening sentence: Being a fast-follower may not work when it comes to artificial intelligence

*Sci-fi subplot note: I try hard to convince people to read science fiction to learn about new technologies, since so many of the geeks inventing the stuff grew up reading it and are now trying to build what they read. With regard to the “universal translator,” “Star Trek" got there first (though by calling for the invention in 2151, it missed the mark by more than 125 years).

But “The Hitchhiker’s Guide to the Galaxy” got there funniest: "The Babel fish is small, yellow, leech-like, and probably the oddest thing in the universe. It feeds on brain wave energy, absorbing all unconscious frequencies and then excreting telepathically a matrix formed from the conscious frequencies and nerve signals picked up from the speech centres of the brain, the practical upshot of which is that if you stick one in your ear, you can instantly understand anything said to you in any form of language: the speech you hear decodes the brain wave matrix."