Advertisement

Machine Learning of the Next Decade: The Promises and the Pitfalls

By on

WANT TO STAY IN THE KNOW?

Get our weekly newsletter in your inbox with the latest Data Management articles, webinars, events, online courses, and more.

pg_mlnxtdcd_072616Machine Learning (ML) technology has gained rapid strides in the current enterprise software ecosystem, as it perfectly balances the ease of adoption, deployment, and market maturity with strong business benefits. On the flip side, Machine Learning technology also threatens some core elements of the human experience such as natural processes, human labor, or traditional workflows. This article explores both the positive and negative impacts of such an all-pervading movement that is rapidly engulfing the Digital Age.

Although ML is helping to tackle some of the most difficult Data Management issues like explosive data volumes, the Cloud, and Big Data, it cannot and will not be able to stand on its own without suitable intervention of human decisions makers. Many experimental ML projects such as Google Car have proven such an assertion beyond any doubt. It is widely acknowledged that Machine Learning of the next decade will be one of the biggest technological contributions to human society, but at what cost?

Machine Learning: A Transformational Force of Future Computer Technology

Machine Learning possesses all the characteristics of a transformational enabler of next-gen,  software technology.   Currently, we are witnessing the impact of Machine Learning in Artificial Intelligence (AI) and Robotics in its infancy; however, as the next decade unfolds, we will begin to see traditional enterprise software development getting replaced by ML techniques. There are some benefits and some difficult challenges ahead for the global software development community ahead. What follows is a quick review of some of those challenges and benefits of Machine Learning in enterprise software development.

The Promises

Machine Learning Will Provision Technology Beyond the Data Center

A recent article Google Wants to Bring Machine Learning to Everyday Devices states that Google is actively working with Movidius to deliver Machine Learning locally to mobile devices without the presence of Cloud platforms. As typical mobile interactions create more unstructured data, ML will allow these devices to interpret images and audio in a perfectly natural way. Currently, mobile ML software applications are limited by the hardware’s power and processing capabilities, which forces mobile users to depend on Cloud platforms. Google is planning to use the Movidius- MA2450 chip, which enables super-charged computation without the need for a Data Center. Movidius reiterates that future mobile devices would have the power to decipher images and audio with more speed and accuracy—thus leading to a highly personalized experience for the user.

Enterprise Software Development Will See Major Transformations

The LinkedIn Pulse article titled Machine Learning Will Power the Next Generation of Enterprise Software presents an interesting picture of the future generation of software applications aided by Machine Learning:

  • ML Stacks, which help accelerate Data Science innovation, are finally being embraced and used by developers and third-party vendors.
  • Machine Learning techniques become more effective and efficient when applied to large data sets. Newer technologies like Hadoop that enable the use of massive data have made Machine Learning techniques fully achievable now that were once hitherto unrealized.
  • The sudden rise of mobile and social data has necessitated ML to work with large external data bases and data sets.
  • Gradually, advanced Machine Learning techniques like clustering or regression are gaining pivotal importance in transforming information into tactical intelligence for businesses.
  • Now, Software-as-a-Service (SaaS) systems can be reinforced with Machine Learning models to deliver better intelligence and forecasts.
  • The future ML models will act as an enabler for more domain-specific business solutions. For example, advanced ML techniques like sentiment analysis, churn prediction, or forecast analysis are rapidly disrupting the traditional sales processes.

A recent article on how the Chinese government is employing Machine Learning in criminal identification provides an informative use case of its current advancement into many different areas of society.

IBM Braintrust Remains Hopeful for a Better Future

In the article, Through Machine Learning, IBM Braintrust Sees Better Days Ahead, Fortune magazine reveals the outcome of their meetings or interviews with the top brass of IBM. The Fortune magazine reports that IBM’s CEO, Ginni Rometty, envisions many business opportunities through advancements in Machine Learning of the next decade. 

IBM executives feel that Cognitive Computing, or Machine Learning, as represented by IBM’s Watson, is “a technology that learns as it goes by ingesting data.”  Computer systems like Watson have the power to learn on their own, without any human intervention. IBM’s Ginni Rometty sees $2 trillion worth of business opportunities in the ML market of the future.

The Pitfalls

The Biggest Obstacle in the Future of ML: Humans vs Software-driven Robots

It has been suggested that Machine Learning projects are much tougher to harder to manage and execute than other technological projects. Typically, what appears to be 80% accurate on the first week of development may very well slide down to 20% accuracy the next week, or may never even succeed. Further, the threatening prospect of complete automation in human tasks or robots taking over as software developers is the next big challenge for ML projects. To remain realistic, technical leaders or projects heads should think of making machines and humans work together.

A very good example of this human-machine conflict is a self-driving car that could potentially replace all human drivers if it can ever run on auto-pilot with 100% accuracy. But what about a self-driving car that can drive with only 99% accuracy? That means the car still has a 1% chance of killing passengers, pedestrians on the road, and damaging properties. Google’s recent diagnostic tests on self-driving cars revealed that approximately 15% of the disengagements on the order of 30,000 miles could have resulted in accidents if driving control was not handed over to a human driver.

Will Next-Gen Machine Learning Affect the Lives of Future Citizens?

In How Will Machine Learning Impact Your Life, the author predicts some likely trends that may transform the future lives of citizens in ways far beyond the ordinary human comprehension. The predicted technological trends include:

  • The growth of the millennial workforce to about 50% of the total workforce by 2020, and to 75% by 2025.
  • Worldwide proliferation of smart devices with rise of automated, ML-aided, connected mobile networks.
  • The erosion of traditional human labor or jobs.

Currently, we will probably see a wave of smart algorithms predicting everything from diseases to weather, but soon, there will be self-driving cars, automated financial advisors, or robotic healthcare professionals. Machine Learning, in conjunction with robotics, will not only transform human-machine interactions of the future, but may even potentially threaten thousands of jobs conducted by humans. Another popular thought trend indicates that while newer technologies will push human labor into oblivion, they will also provide fresh job opportunities for the entrepreneurial mindset. The millennials are advised to study the current skills shortage patterns, and start developing skills relevant for the future Machine Learning Era.

Machine Learning Week 2016 provides a quick review of the latest and smartest developments in the field of Machine Learning; they also have a free Machine Learning e-book.

In conclusion, one can only say that while Machine Learning of the next decade promises improved medical diagnostic systems, better fraud-detection models, or automated decision-support systems, enterprises will no longer require a gigantic R&D budget to adopt Machine Learning techniques. The current challenge though is how to make imperfect or nearly-perfect Machine Learning models to work in the existing workflow.

Leave a Reply

We use technologies such as cookies to understand how you use our site and to provide a better user experience. This includes personalizing content, using analytics and improving site operations. We may share your information about your use of our site with third parties in accordance with our Privacy Policy. You can change your cookie settings as described here at any time, but parts of our site may not function correctly without them. By continuing to use our site, you agree that we can save cookies on your device, unless you have disabled cookies.
I Accept