AI: Another Banner Year
We are set to conclude another banner year for applied machine learning and artificial intelligence, in what has now proven to be a long, successive line of banner years. As we swap out the calendar, it seems an opportune time to review what was forecasted last year (and how accurate it was) and look ahead to what I think is coming next.
What Was Correct
At the end of 2016, I penned my expectations for AI in 2017. These predictions included: continued advancement in chatbot technology, an uptick in the battle over the home voice assistant, the launch of the first real, road-ready fully autonomous car, a significant expansion in MLaaS offerings, an invigorated battle in ML-oriented hardware, and a major breakthrough in quantum computing. Today, I feel confident in saying that I was a solid 5-for-6.
Chatbots have continued to expand in use and talent, especially across the marketing world. The ecosystem of chatbot start-ups has continued to flourish, especially as tooling and services improve, and estimates now expect the chatbot market to reach $1.23 billion by 2025.
The battle for the home voice assistant has become dramatic and expensive, as expected, with both Amazon and Google expanding their offerings (and mixing in some feuding, as when Amazon recently refused to carry YouTube on it’s Fire platform). Apple is expected to launch a home voice assistant to expand Siri, and eyes are on Microsoft to see what they will do with Cortana. According to estimates, Amazon sold “millions” of Alexa-enabled devices during the 2017 holiday season, and has likely sold over 20 million devices since launch.
nuTonomy and Lyft launched a real-world autonomous ridesharing pilot late in 2017, with select customers being given access to fully autonomous rides in the Boston Seaport. Both companies have stressed that the program is not merely a test, but is a real feature of the product, with hundreds of autonomous rides completed already. nuTonomy parent, Aptiv, is said to be planning a major expansion of these efforts, and is also planning a driverless car technology hub in Boston (the current home of nuTonomy).
MLaaS took center-stage at the 2017 AWS reInvent conference in Las Vegas, when Amazon launched a full menu of powerful new offerings, including SageMaker (which hopes to democratize access to even the more advanced ML tools with an end-to-end model building and deployment solution). In direct competition with existing offerings from Google Cloud and Microsoft Azure, the battle for ML services has really heated up, and each is actively seeking dominance in the space by touting top-tier partnerships and customers (and reducing prices).
Specialized AI hardware has started to gain serious momentum, which began with the launch of the TPU by Google in 2016, and has continued with the Nvidia Volta, the Intel Loihi and Nervana, and Neural Engine (amongst several others). While the GPU remains the industry favorite for high-horsepower compute, ASICs are becoming more attractive to those who demand unusually high workloads against very specific computational use cases.
What Was Incorrect
Sadly, we didn’t see any great breakthroughs in quantum computing, as predicted. Research carries on (and expands), and all the major technology players have a hand in the space (including some non-traditional players, such as JP Morgan, who just announced a big partnership with IBM). Smaller-scale advancements were seen, of course, but the big and elusive prize of an affordable, large-scale quantum computing system remains unsolved.
Maybe next year?
What Is Next
I don’t expect to see anything resembling a slowdown in the pace of adoption or expansion of practical application in 2018. If anything, the slope of the curve will likely continue to steepen as existing but young markets start to mature, and new verticals start to emerge. Here are a few areas where I expect to see notable advancements or announcements in the coming year:
Healthcare. For a few years now, there have been a great many academic publications and early commercial efforts in applying machine learning to healthcare, mostly with a focus on finding ways to leverage the impressive classification powers of these systems. For example, in late 2017 MIT announced a model that proved incredibly accurate at detecting early signs of breast cancer, with an accuracy rates well over 90%. I expect that 2018 will prove a pivotal year in the growth and eventual commercialization of these diagnostic opportunities.
Finance. For a myriad of complicated reasons (some regulatory, some cultural), the finance industry, despite being historically on the cutting edge of most analytical and predictive advancements, has not widely adopted machine learning (with notable exceptions). However, as the fintech industry continues to heat up and attempt to challenge the establishment, I expect this trend to quickly change. More start-ups will rise with a focus on providing advanced analysis and automated trading platforms, and more banks and hedge funds will pour money into developing (or supporting) platforms that will assist (or supplant) their traditional efforts.
Driving. While I may have made mention of autonomous driving last year, it would be disingenuous to not mention it again, with a slight twist: I expect 2018 to be the year that autonomous offerings really go mainstream. Lyft, Uber, and others will make a serious and sustained push toward genuine adoption, city-by-city, and people will start to become more comfortable with the technology. I dare say, if the stars align in just the right way, I expect the first consumer sale of a fully autonomous-capable car to happen toward the end of 2018.
Security. The international threat of cyber warfare (brought to the foreground of the public consciousness following the 2016 Presidential election), combined with the ever present threat of an increasingly-connected world, will likely result in an explosion of interest in AI-assisted security. Money has been pouring into this industry for a few years now, and I expect broad adoption to really start ticking up in 2018 (along with more than a few additional public and damaging hacks, which will only work to further spur demand).
Talent. With a worldwide focus on machine learning, which only continues to broaden and intensify, the number of new practitioners will steadily yet notably rise (alongside worldwide demand). The growth in MLaaS offerings and ongoing improvements in common frameworks certainly promise to widen access to at least the basic tools and workflows of machine learning systems. Combined with an ever-expanding array of learning opportunities, such as offerings from Coursera and edX, and new degree programs from popular universities, I think the title of data scientist will become quite common. Which is very good news, as there is a dramatic shortage in data scientists; IBM expects demand for data scientists to rise 28% by 2020, and Tencent has argued that we are short by millions of qualified AI engineers.
What About Travel?
It should be clear by now that machine learning and AI are set to transform nearly every industry and market; travel is no exception. So what do I expect for this space in 2018?
While I am certainly biased, I firmly believe that the biggest opportunity is in deep and focused personalization. Several of the major players in travel booking, recommendations, and reviews have made tepid forays into personalization, but no one has yet won the space for a simple reason: everyone is much too focused on revenue maximization at the expense of the customer.
People crave products that understand who they are and will help them get things done, quickly and easily. The first company to truly serve those desires, delighting their customers with a unique, highly-personalized experience that helps them get what they want with minimum effort, will secure a beloved position in the category (and will fundamentally change the way people search for and make decisions about booking travel). That is what Lola is trying to do, and I have no doubt that the rest of the industry will follow suit.