In machine learning, it’s not wise to make predictions too far out. Just a few years ago, it was common knowledge that computers might be great at finding patterns in numbers and data but they couldn’t recognize images as well as the eye of an insect. Last year, Google researchers proved that observation to be obsolete.

The Significance of Pattern Recognition

The Google FaceNet algorithm achieved 86 percent accuracy in identifying faces through a range of lighting conditions and at a variety of angles. Facebook had already demonstrated 97.5 percent accuracy at matching faces, but FaceNet’s achievement went further. The machine was able to describe which faces looked the most similar and which were the most distinct.

Better pattern recognition will be critical for machine learning in 2020 because patterns define the human experience. People know each other by the way they look, not by passwords or fingerprint scans. As people go to work, they judge the weather by the look of the sky and the commute by the movement of cars on the road ahead.

The real challenge for the next generation of machine learning tools will be not in learning how to think but in learning how to see. Algorithms will need to learn how to make sense of what their cameras capture and report it back to humans in a natural human language.

Applications of Machine Language in 2020

Managing transportation in the city of the future is a top priority for smart cities and it will take next gen machine learning tools. In a study of smart city initiatives by Juniper, researcher Steffen Sorrell explained, “Facilitating the movement of citizens within urban agglomerations via transport networks is fundamental to a city’s economic growth. Congestion reduces businesses’ competitiveness, and contributes to so-called brain-drain.” Sorrell estimates that traffic management and smart parking initiatives based on machine learning will save 4.2 billion man-hours per year by 2020.

A good indication of where machine learning is headed comes from an app created by Cubic Transportation Systems (CTS). Their vision for an app called NextCity would act as a platform for all modes of transportation within any given city, from public buses and rental bikes to parking garages and ride shares. The app would instantly be updated by citywide sensors about traffic delays, lanes closed, service changes and anything else that might affect travel times.

The latest advances will soon be tested out in Miami, where CTS and Passport have won a $33 million contract to modernize the city’s online transport hub. Passengers will be able to pay for rides with mobile wallets like ApplePay or wearables like the AppleWatch and the Samsung Gear.

$3 Million for Inspiration

That may not be feasible for every city but it outlines what will soon be achievable in the transportation sector due to advances in machine learning. To speed up that process, IBM and the XPrize Foundation are offering a $3-million incentive for teams to solve the world’s biggest problems using AI and machine learning. Issues around parking, traffic flows, smarter energy use and rapid transit in the cities of the future will be at the top of the list.

Another promising direction involves apps like the Wit.ai natural language processor. This helps drivers in connected cars replace their keyboards or touch screens with voice commands. The machine learning part of this API improves its ability to extract meaning from commands based on context and examples.

It may be hard to believe, but 2020 is just over three years away. That’s a lifetime given the speed of development in machine language right now. This is definitely one area where the current limits are not on the technology itself, but on our own creativity in telling the machines what we want them to learn.