2016: The Year of AI Made Real

In this series, professionals predict the ideas and trends that will shape 2016. Read the posts here, then write your own (use #BigIdeas2016 in your piece).

Last month, I found myself in the same hotel bar as IBM Chief Scientist Grady Booch and our conversation eventually turned to a hot topic: the near future of AI.  We weren't talking strong AI like “open the pod bay doors Hal,” but the much more common (i.e. real) intelligence called “applied AI” that’s quietly sneaking into our homes and businesses.  I’ve written quite a lot about both kinds of AI lately, because barring a major civil unrest or natural disaster it’s likely to change the way we live more than any other type of innovation.  I just don't think it will happen the way most people think.

IBM, of course, is well known for their Watson brain, which has long moved on from crushing Jeopardy opponents to new careers helping oncologists make more informed treatment decisions, and mashing up old recipes with new food science to create entirely innovative dishes. Parts of Watson (like speech synthesis, question parsing and its core algorithms) could one day be the foundation of a broadly conscious system that instantly changes everything about our lives and even how we think about life itself. Today, however, Watson’s intelligence is applied to very specific problems in search of very specific solutions. These soft variants of AI have been slowly and, for the most part, quietly changing how we live.  In 2016 the change will be hard to miss.

Applied Artificial Intelligence is the Only Game in Town

We can all agree that this year hasn’t given us any reason to believe that the T-1000’s are any closer to our doorsteps, but 2015 has given us some serious clues that the “applied” variety of AI has reached an inflection point that will move it from niche to mainstream. These task-specific AI have all been in the works across many industries for quite a few years.  But the journey from R&D labs into people’s homes and businesses seemed painfully slow up until now.  I believe 2016 will be the year that AI becomes real for a greater majority. What was bleeding edge this year — from home automation to cars — will quickly become commonplace. 

Much More than the Google Fleet

In 2016, the most obvious instances of applied AI will be in our cars. At CES earlier this year, Mercedes showed off their F015, a seemingly windowless, auto-piloted car that looked like it would be at home in a sci-fi film. If you asked me then, I would have said that we’re still many years off from autonomous driving outside of Google’s private fleet. Only 10 months later, Tesla vehicles are not only self-parking, they are fully auto-piloting at freeway speeds — albeit only on well-marked roads. Tesla’s AI can’t yet read traffic lights or stop signs, but it can read speed signs, avoid hazards and inform you with some insistence that it’s gotten out of its depth and request human assistance.  What’s more, it’s unlikely that Tesla owners will have to wait years for AI enhancements.  New “awarenesses” like responding to traffic lights can be delivered as painless, over-the-air software updates.

AI in cars won’t be limited to the open roads for long. Next year will also see autonomous driving in dense urban traffic become commonplace. Both BMW’s “Active Assist” and Volvo’s “Pilot Assist,” with their partially automated steering, breaking and acceleration for heavy traffic are slated for 2017 models that will show up next fall. By the holidays next year, you will either be shuttled in an auto piloted car or you’ll be sharing the road with one. Either way, very few drivers in the Western world will be able to change lanes in 2016 without considering that the agent operating the vehicle next to them may not be human. The experience might be subtle, but the reality is enough to give us all pause at the shift in our technological environment.

A Handheld Assistant that Actually Assists

The second area where AI will become real for many people will be with assistants and automation. In recent years mobile assistant technologies like Siri, though eminently available, have seen sluggish adoption as a direct result of limits to their practical intelligence.  Apple’s acquisition of the personal assistant app Cue in 2013 has unquestionably made Siri smarter (try “Hey Siri, what planes are overhead”) but interactions with the mobile assistant still feels more like a Web search than a conversation with an intelligent helper. 

The challenge of having a conversation with an artificial agent, versus simply querying it, comes down to an agent’s ability to parse natural language.  Question parsing, though not traditionally thought of as intelligence, is at the heart of why we still prefer to “ask” humans questions and still just “query” our mobile devices.  For example, I can ask Siri “how long will it take me to get to work,” and using my phone’s GPS, my work details in my contact card and traffic information from Apple Maps’ data service TomTom, Siri will give me a time estimate based on the quickest route it finds. This is a cool query, but it’s just that.  If I were to ask Siri, “How long will it take me to get to work if I take the 101?” it instantly falls down. Getting a valuable answer to a question like “Hey Siri, if I was in Hong Kong right now and at my favorite hotel, how long would it take me to get to the airport?” would be simply unimaginable. Outside of AIs like IBM’s Watson, the technology still isn’t available. That will change soon.   

This summer, SoundHound (a Bay Area company that is working to turn speech into understanding and meaning) began releasing unpolished demos of a mobile assistant that could parse complex questions and even multiple questions at once. It easily handled questions like “what is the population of the capital of the country where the Space Needle is located?” or “when is the sun going to rise two days before Christmas in 2012 in Tokyo?” Being able to “talk to” the Hound app versus rigidly speaking a series of keywords is a game changer that I expect will show up big next year.    

X.ai’s “Amy” is another highly focused assistant AI that’s poised for the mainstream in 2016.  Amy describes “herself” as “your AI powered personal assistant for scheduling meetings. You interact with me as you would to any other person — and I’ll do all the tedious email ping pong that comes along with scheduling a meeting.”  This is no exaggeration.  If fact, my first experience with Amy was scheduling a meeting over a long weekend and it was not immediately obvious to me that Amy was AI.  The interaction, via email, was done completely in natural language and if there wasn’t a small footer in the mail that read “x.ai – artificial intelligence that schedules meetings,” I wouldn’t have known better. 

I’m not suggesting that Amy could pass a Turing test, but the point of applied AI is that it doesn’t have to be a fully realized intelligence — it just needs to do a specific task well.  If you haven’t engaged with Amy yet, expect that in 2016 you will.  I’m seeing “her” CCed on more and more conversations. You should also expect the trend of highly specialized assistants to grow next year.  They will be deceptively “human” in how they engage with you — that’s the point — so you may have to watch for them closely.

The Great Machine Intelligence Divide

AI is not just for consumers on one hand and for large enterprises on the other. In my work at GoDaddy, 2016 will be a big year for applied AI as well.  For the past 24 months we’ve been building the world’s largest predictive analytics and machine learning system for small business — collecting the anonymized “data exhaust” from the billions of interactions with our business customers (and their customers in turn) and returning that data to our customers in the form of insights.  Like auto-pilot in cars or smarter assistants in mobile phones, the result of our work will be an applied intelligence, purpose built to do one thing extremely well.  In our case, it will guide the small, independent entrepreneurs that make up the bulk of our customer base toward better business decisions.

Though the strong AI of sci-fi fame looms out there — either as a promise or threat or both, applied AI is at our doorstep now and poised to change the way we live and work now.  There’s a lot to look forward in 2016 — let’s see how far we can move the needle with AI in a single year.

Thoughts or Questions? 

I’ll be available in the evenings (West Coast) this week if you want to talk more about applied AI—just leave a comment.