Two weeks ago, Google bought Clever Sense. Some are billing this as a response to Siri. Some see it as Google going after Yelp or Wolfram Alpha. And some have a hunch it’s about Hunch, which I once called the iPhone of Search.
I think the answer is none of the above -- but, in a way, all of the above.
With Clever Sense, Google is trying to embody the original Ask Jeeves -- a butler-like digital assistant that knows who you are, what you want, when you want it, where you want it, and how you want it. In my mind, that would be the perfect search engine.
Alfred at Your Service
Meet Alfred. Alfred is “your personal robot who recommends places based on your favorites.”
Available as an app for Android and iOS, Alfred “provides curated recommendations using Clever Sense's Serendipity Engine.” Essentially, he’s an app-ssistant.
But, wait, there’s more: “In tune with the location, time, and user's intent, his suggestions are personalized for each situation.”
So how does the sausage get made?
“The Extraction Engine built into the Clever Sense Platform curates large amounts of unstructured crawled data by leveraging natural language processing, statistical machine learning, and data mining algorithms.”
And how does Alfred know how you take your sausage?
“Start with teaching. Tell me about few of your favorite places to go out, and I'll have some new ideas for you.”
After downloading the Alfred app, I answered a series of questions about the kinds of places I like to go for various occasions from a quick lunch (Jimmy Johns, of course) to a Saturday night out with friends (wherever Billy Dec is tweeting).
Alfred then takes these “teachings,” combines them with the lessons he’s already been taught by people with similar preferences, does his fancy statistical machine learning, and then serves up recommendations.
When the Student is Ready, the Master Appears
Eric Schmidt has often said that the perfect search engine would return only one result for each query -- the one you want at that moment. His remarks at this year’s All Things Digital Conference shed some light, and quite a bit of foreshadowing, around the value of Clever Sense and its trusty app-ssistant, Alfred.
“We’re trying to move from answers that are link-based to answers that are algorithmically based, where we can actually compute the right answer. And we now have enough artificial intelligence technology and enough scale and so forth that we can, for example, give you -- literally compute the right answer.”
But Clever Sense goes well beyond one-box results. Alfred doesn’t even wait to be asked questions.
Clever Sense promotes the idea of “Zero-Query Search.” After all, Alfred’s got “personalized, relevant recommendations for you from the start. Not sure whether you want ramen or ravioli? I can show you a stream of recommendations right when you open up the app, so you can consider all your options.”
Be Careful What You Wish For
There’s no shortage of subplots with Google’s acquisition of Clever Sense and its continued pursuit of search perfection. Here are three storylines I’ll be watching.
1. Monetization: Is Zero-Query Search really in Google’s best interests, given that the company makes almost all its money from ads alongside search queries? Maybe, if the price marketers are willing to pay for Alfred’s recommendation is high enough. From the T&Cs of the Alfred Android app: “Correspondence from Clever Sense may contain commercial messaging, such as advertisements and offers from Clever Sense and/or select partners.”
3. Scale: Alfred works pretty well with restaurants, but how will the Clever Sense Serendipity Engine do for other verticals like retail or travel? And what about noncommercial search activity where intent is a bit less clear -- or, as George Michie puts it, “utterly ambiguous?”
At the recent Search Insider Summit, I led a panel of expert chefs in an experiment to cook up the perfect search engine (video and slides here). The recipe called for just the right blend of user experience and relevancy algorithms. We debated all the various ingredients, from query format (voice, text, audio, or images) to SERP integration (products, news, maps, video, images, websites) to SERP interaction (transactions, recommendations, one-box answers, refinement tools or links).
At the end of the day, the only thing we could agree on was that perfection is in the eye of the beholder. In that sense, one man’s clever is another man’s cleaver. Indeed, when it comes to cooking up the perfect search engine in Mountain View, let’s just hope Alfred doesn’t go all bats*it on us. And, for all you Michael Gough fans out there, that’s an “h” not a “u.”
The approach of Google seems to rely too heavily on trying to understand as much about us as possible: our preferences, our historical patterns, etc. But humans are an unpredictable bunch. We are often subject to spontaneous trains of thought, usually sparked by content we are engaged with. We hadn't planned on taking a certain path ahead of time, nor had we taken this unpredictable path of discovery in the past. A traditional search engine, no matter how sophisticated, wont address this need. I agree "Zero-query" search is where search is headed. But the big question is, what data is driving zero-query? Our platform is betting that content, context and the "impulse-buy" drives zero-query, as opposed to the Big Brother, in-your-shorts personal data mining path that Google is spending billions on. Stay tuned.