Printer Friendly

Why Is There Only One "Machine Learning First" Company?

Have you noticed how much more accurate Google search results have become in the last year? ... Machine learning.

Have you seen the improvements when you search for pictures and videos online? ... Machine learning.

Have you been surprised by the predictive word and phrase suggestions in Gmail? ... Machine learning.

Have you experienced improved language translation from Google Translate? ... Machine learning.

Google and a few others are creating and playing an entirely different game than everyone else.

All these improvements were the result of years of effort and experimentation in developing and perfecting machine learning models and software. Hundreds of computer programmers have been shifted from writing new search and translation algorithms, to writing machine learning models capable of performing search and translation better than humans can. The improvement in performance is significant, as is the reduction in the amount of code and effort required to customize the models. Jeff Dean of Google has reported that 500 lines of machine learning code in TensorFlow has replaced 500,000 lines of traditional code in Google Translate. With the right machine learning model, the system performs better, requires less code, and is significantly easier for human staff to learn, master, and maintain.

At Google, machine learning is gradually changing how all the company's products and services work. Josh Cogan, another senior Google engineer, has stated that the company is currently using hundreds of different machine learning models across all its products, services, and internal operations. The company's leaders have declared Google a "machine learning first company." This statement means that when creating or improving a software product, Google looks first for a way to do it with machine learning as opposed to hand coding it the traditional way. And Google is not the only company making big investments in machine learning. Amazon, Microsoft, IBM, and Netflix are all making similar moves to evolve their products and services and build an advantage over competitors who have not yet grasped the power of machine learning.

What makes machine learning so powerful is its ability to examine millions of data points, find patterns, and make decisions that adapt its performance in real time. Human business analysts can't work as broadly or as quickly as machine learning algorithms can. The company that masters these tools essentially doubles, triples, or even quadruples its workforce at a fraction of the cost of adding human employees.

This trend in cognitive labor mirrors the transformation of physical labor during the industrial revolution, when machines replaced large numbers of manual laborers, reducing the time to create products while also improving their quality. Industrialization eliminated some kinds of jobs, but it also created new ones; someone had to manufacture, monitor, and maintain the machines. The role of the human shifted from making products to making (and fixing) the machines that make products. This shift is occurring now for workers who make software. They will become coders, monitors, and maintainers of the machine learning models that will provide the software's capability. Millions of coders who know how to program in the traditional manner will have to learn to program or monitor machine learning models; many other workers who now spend their days analyzing data will need to learn to analyze and measure the performance of machine learning models, instead of working directly with data.

That change is coming--it's even underway--but it's not here yet, at least not on a wide scale. Business media would have us believe everyone is already turning machine learning into improved business performance. But actually, only a few are. It is still very early days; the power of machine learning has been applied only to a few, quite similar classes of problems. Only the rare one-tenth of one percent of companies are on the frontiers, unraveling their businesses, experimenting with machine learning, and finding ways to remake their businesses with these tools.

Discovering how to apply machine learning more broadly remains a challenge. It cannot be applied to all business IT problems just yet. The databases and transaction processes associated with sales and order placement, for instance, have yet to be subjected to machine learning. Machine learning algorithms can examine images, audio, video, streams of transactions, medical images, radio signals, and other similar forms to identify patterns that are not explicitly obvious by other methods; those patterns can then be mapped to actions to be taken--that's how the software replaces humans. But it does not work when the problem is unique or too unusual, when there are too few previous instances to generate a pattern.

In a 1993 NPR interview, science fiction writer William Gibson said, "The future is already here; it's just not very evenly distributed." Google and its peers are on the front edge of the future with respect to machine learning. It might be time for the rest of us to begin learning how to play this new machine learning game. The machine learning revolution will change the way we write software, process data, respond to customers, and design products. As a result, it will increase the velocity of business and the velocity of change within a business, or a market, or an industry. This future has already happened at Google. When will it happen for your company?

Roger Smith is the Chief Technology Officer for AdventHealth Nicholson Center. He has also served as the CTO for US Army Simulation and for Titan Corp. and as a vice president of technology for BTG Inc. A member of RTM's Board of Editors, Smith has led technology innovation for medical, defense, software, and computer systems. He holds a PhD in computer science and a Doctorate in business administration.

Orcid: [iD]

DOI: 10.1080/08956308.2019.1661081

Published by Taylor & Francis. All rights reserved.
COPYRIGHT 2019 Industrial Research Institute Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Smith, Roger
Publication:Research-Technology Management
Date:Nov 1, 2019
Previous Article:The Technology Fallacy: People Are the Real Key to Digital Transformation.
Next Article:Defining the Tipping Point.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters