Ed Fernandez Blog 2000x1000px

The Race Towards Enterprise-Level Machine Learning Applications

Founded 12 years ago, the Spotify music streaming service is today as ubiquitous as radio. Where it differs from radio, however, is that whenever any one of its millions of listeners tunes in, they find a music playlist customized to their individual taste. This magical experience is made possible through Machine Learning (ML) technology, a branch of Artificial Intelligence (AI), which “learns” listening preferences to create customized playlists that are potentially in the billions of song combinations. This massive scale of personalization is only made possible by the advent of ML technologies. While this is illustration applies to the music industry, ML applications cut across multiple industries, making it necessary for corporations to explore ways to use the technology to fend off competitors.

“Software is eating the world, but AI will eat up software.”

Jensen Huang – NVIDIA CEO

This illustration is perhaps the perfect stage-setter for the next iteration of advancements in enterprise digital technology applications. Rapid advances in ML are now seen as having the potential to supplant traditional software programming within the enterprise context. To further understand how ML applies to the enterprise, let us first look at another simple illustration.

Machine Learning vs. Software Programming

Imagine for a moment that you have gone to the local farmer’s market to buy apples. When you arrive, you find you must select between good apples and bad apples. If we were to build a software program to make this choice, we would have to input rules to tell it how to do so. So, for instance, we would say select for size, color, and origin. If you want the program to use more parameters, you must input these as well. The result is a program that does one thing repeatedly in what is known as traditional rules-based software programming. This type of software is what currently powers most enterprises today. Next, let us look at how a machine learning algorithm would complete the same task.

Unlike traditional rule-based programming, ML utilizes data-driven rule discovery. In this case, instead of putting in a set of commands, we would feed the algorithm a data set that includes the various sizes of apples, various colors, various origins and so on. The algorithm would then combine all these pieces of data in various ways to generate an outcome. Assuming we only accept large red apples from France, the algorithm would arrive at this same conclusion through reductive computation. That is, it will eliminate all the combinations that result in a rejected outcome. In this way, the algorithm can self-train to look for greater nuances, in the same way that a human would when determining a set of choices.

Ai Business Case

The idea that it is possible to train algorithms to make choices has tremendous applications in the enterprise. Consider the company Arterys. Offering medical imaging cloud AI, the company uses machine learning to process radiology scans to identify anomalies. By using each subsequent scan as a basis to improve future results, the AI can spot tumors faster and more effectively than a human radiologist would. However, it is not enough to look at such awe-inspiring examples to know that ML is poised to accelerate in the enterprise setting. One need only look at the amount of money going into ML to see a rapidly accelerating trend.

Venture Capital Investment Growth in ML

According to CB Insights, in Q1 of 2012, there was only one publicly disclosed merger and acquisition or M&A deal in the ML space. By Q1 of 2017, that figure had soared to 34  publicly disclosed deals. While tech giants like Google and Amazon are leading this wave of acquisitions, the same report shows that other legacy businesses like IBM, Nokia and GE are also getting in on the action. This rapid acceleration in the space demonstrates an increasing urgency to acquire the necessary technology to apply ML in more mainstream ways. What is shaping up is the greatest enterprise platform revolution since desktop computing.

Enterprise Platform Revolution

As with all technological revolutions, adoption always follows a bell curve of what is known as the hype cycle. Referencing the Gartner hype cycle research methodology, we find ML just beginning to come off the peak of inflated expectations. From the chart, Gartner predicts that ML is two to five years away from the plateau of productivity, a point that represents a mainstream platform revolution. For enterprises looking at ML, now is the right time to begin experimenting with the technology as it provides first-mover advantage before laggards move to adopt the technology.

The real opportunity ML represents, however, is its industry agnostic nature. Companies across industry verticals can find useful and productive applications to boost their competitive advantages. ML-as-a-Service infrastructure investments from tech companies like Google, Amazon, IBM, and others provide a ready opportunity for forward-thinking firms to start experimenting with ML without having to make massive investments.

Enterprise Machine Learning Adoption Drivers

Firms that are still unsure about investing in ML must know platform revolutions take the form of massively disruptive self-perpetuating cycles that leverage emergent technologies to accelerate. In the case of ML, there are five key drivers of adoption:

  1. Data
  2. Hardware
  3. Algorithms
  4. Tools
  5. Expertise

Data

Data is the foundation of ML. Today, petabytes of data are available for ML purposes. Intel CEO Brian Krzanich calls data the new oil. In the same way oil fueled an entire industrial revolution, he sees data as the new oil fueling the ongoing digital transformation revolution.

Hardware

To process all this data, AI-focused chip development like NVIDIA’s Tesla GPU as well as chips from other companies like Intel, AMD, and Qualcomm, is on the rise.

Algorithms

Deep Learning, a type of complex ML, is at the core of some of the most advanced AI applications such as IBM’s Watson and Google’s Deep Mind. Such algorithms are creating new possibilities like natural language processing, autonomous cars, image recognition, among others.

Tools

Currently, available ML tools represent capabilities that allow firms to utilize these ML resources in practical and scalable ways. Such tools include ML-as-a-Service, ML APIs, and open source ML technologies like TensorFlow, Keras, Microsoft Cognitive Toolkit, among others.

Expertise

Human expertise brings ML to life by modeling potential use cases. As perhaps the most crucial component of all five, this represents an opportunity for corporate leaders to be at the forefront of discovering ML applications that can help vault their firms to the next level of growth.

Real-world Machine Learning Examples

Mount Sinai Hospital Deep Patient – Medical Diagnosis

Incorporating hundreds of thousands of anonymized patient records, Mount Sinai Hospital’s Deep Patient can diagnose hard-to-catch ailments by processing patient data and cross-referencing with machine-learned data.

Waymo – Autonomous Cars

By subjecting autonomous vehicle (AV) ML algorithms to thousands of miles of real-world driving, Waymo is training its autonomous cars to one day drive safely with no human intervention.

Google Duplex – Autonomous Personal Assistant

Google Duplex, an advanced personal assistant AI, can call a business and, using natural-sounding speech, book an appointment. This application is but the tip of the iceberg of what is possible.

Google Maps – Transit Optimization

Crunching billions of data points sent in from Android devices running Google Maps, Google Maps not only correctly estimates transit ETA’s, but also provides alternative routes.

Strategic ML Application

Companies like Google and Baidu typically have upwards of a thousand engineers focused on ML applications. While this may not be viable for most companies, it is important to have an ML strategy in place. Such a strategy could mean either internal provisioning of resources or partnering with startups in the ML space. In either case, not having an ML strategy in place means exposing the firm to disruptive threats from other forward-thinking players currently experimenting with ML in internal innovation labs.

WEBINAR: Towards Zero Clicks: Machine Learning and AI for Every Business

Machine Learning is the next frontier in enterprise software applications. Where desktop computing gave rise to the current information age, machine learning is poised to create enterprise computing capabilities we are only beginning to comprehend. This article is an excerpt of a one-hour deep-dive webinar into enterprise machine learning by Ed Fernandez, co-founder of innovation advisory firm NAISS, early-stage, and startup VC and mentor at Singularity University. To watch a recording of the webinar, click the button below.

WATCH NOW
info@svicenter.com 1-650-274-0214