A quick Web search seems to paint a bleak picture of algorithms. Pair the word with “interact,” and the first few hits include “When algorithms control the world,” “Can you trust your algorithms?” and “Why we should expect algorithms to be biased.”
They’re definitely disrupting the job market. Research cited by The Conversation estimates that nearly 47 percent of U.S. jobs could be lost within the next two decades because of algorithmic business. (This extends to members of my own profession: “By 2018, 20 percent of all business content will be authorized by machines, which means a hiring freeze on copywriters in favor of robowriting algorithms,” Datafloq reports.)
Robot fears aside, a definition is in order. An algorithm is essentially a recipe for doing a set of tasks. In the age of big data, this means everything from Amazon suggesting books you might like to Uber’s dominance in sending a car right away.
To thrive, businesses must learn to harness the growing power of algorithms. Here are a few points to keep in mind:
The age of apps is declining. In the 2020s, IT research company Gartner says, cloud algorithms in the form of digital assistants will supplant apps to guide us through daily tasks. “People will trust software that thinks and acts for them,” Senior Vice President Peter Sondergaard says.
Algorithms will in time learn from experience to become smarter and generate unexpected results. And then they’ll spawn further algorithms, producing agents from agents or robots from robots.
Algorithms fuel the Internet of Things. Devices are reaping gigantic amounts of data that algorithms can analyze and then automatically act upon at the right moment. Sondergaard says that “by 2020, smart agents will facilitate 40 percent of interactions.” Algorithms will also account for context—crucial to truly understanding what’s happening within a process, business or device. They’ll be able to find the right business moments, make important connections and forecast all manner of behavior.
Don’t trust too easily. How can you be sure algorithms aren’t misleading you? Algorithmic differentiation helps determine their accuracy based on a certain data set, and whether they’re subject to data volatility, according to Datanami. An AD test, for instance, can show whether real-life data inputs match a data model—thus helping you decide whether to discard the model or rely on less costly data collection to get the job done.
Not understanding the mathematic fundamentals behind algorithms can lead to blindly trusting the answers they generate, AD expert Uwe Naumann says. No business wants to invest millions in data mining only to discover that it’s randomly generating numbers.
To help further round out an algorithm strategy, check out these tips, courtesy of The Conversation:
- Figure out which processes can be automated, and think boldly when you do.
- You’ll need to gather and store lots of data so that your algorithms can make validated decisions. Find relevant data sources and apply the Internet of Things to tap into new data sources.
- Collect high-quality data. Feeding your algorithms anything less will generate poor results.
- Test, iterate, train, validate and improve your algorithms in a repeating cycle to develop better ones that add value.