Artificial intelligence (AI) is permeating our lives — just not in the ways we might have expected from reading sci-fi novels or watching robot apocalypse-themed movies.
Instead of having live robots walking around, doing our dishes and engaging us in conversation, AI exists primarily in web and mobile apps designed to help us with small intellectual chores, like finding out when the Civil War began or where the nearest taco restaurant is located.
Until recently, most of these AI developments have been designed to make consumer processes easier; for example, digital assistants take on the role of an intermediary search engine to process a vocal request and fetch appropriate results. Now, the trend is starting to shift toward app development — at least in an early stage.
Rather than introducing a layer of AI to help users make use of a given app, these AI programs are operating in the background, making the apps better. They’re less glamorous than the human-like interactions programs like Siri that are capable of mimicking human speech patterns, but they’re even more useful — even if you never see them.
Google’s RankBrain update
Google is known in the search engine optimization community because of its frequent, manual algorithm updates that shake up rankings and frustrate webmasters. Now, its algorithm is starting to update itself.
Google RankBrain is technically an addition to Hummingbird, an algorithm update that centered on identifying semantic patterns in human speech and delivering results that match a user’s intention (rather than individual keyword phrases). RankBrain works by clarifying complex, ambiguous or hard-to-understand queries so the search engine can fetch better results.
As it learns new correlations between semantic phrases and successful results, it will update itself to serve queries better in the future. If applied to more areas of Google’s search algorithm, eventually the software could “learn” to update itself with little to no human interference, resulting in a self-modifying app on a constant cycle of improvement.
Wikipedia’s new AI assistant
Wikipedia is also developing an AI assistance algorithm to improve its massive store of information. Previously relying solely on human editors, Wikipedia’s new Objective Revision Evaluation Service (ORES) automatically identifies problematic or inaccurate edits to articles and assigns them quality scores, helping editors catch them quickly and easily. As ORES spends more time identifying damaging edits, it will learn faster and more efficient ways to catch similar edits in the future.
The astounding pace of modern AI
As reported by Bloomberg Business, 2015 was a breakthrough year in the world of AI. It’s not the type of new developments that are coming out, but rather the pace that these developments are being produced that’s astounding. The rate at which new learning algorithms are developed is faster than ever, and new AI programs are rolling out almost constantly to address new problems.
AI is more affordable, more practical and more quickly evolving than ever.
The developments at Google and Wikipedia are just two examples of this. As quantifiable examples, image recognition error rates have fallen from 42 percent to just over 5 percent between 2011 and 2014, and the number of AI software projects at Google has climbed from less than 100 in 2012 to more than 2,700 this year.
There are several reasons why there’s such a renewed burst in AI development. First, cloud computing is more available and more cost-efficient than ever, giving more people and companies more power with which to innovate. Researchers also have access to more information with more plentifully available data, meaning more people can learn new things in the AI field more quickly and efficiently.
What it means for future apps
AI is more affordable, more practical and more quickly evolving than ever. Major tech players like Google, Facebook, Apple and Amazon are already finding new ways to incorporate the technology into their consumer-facing products. But more importantly, they’re starting to use it as a substitute for the demand for human innovation.
Rather than relying on humans to find and apply improvements, apps will theoretically be able to find and apply updates to themselves, freeing up the tech workforce to fuel even further AI developments. The step beyond that involves AI algorithms developing even better AI algorithms. At that point, we’ll be well into the technological singularity — but that’s a topic for another day.