The Cloud Can Prevent Machine-Learning Failures

Andrew Froehlich
8/8/2018
50%
50%

For years, machine learning (ML) has been touted as the next great evolution in enterprise IT. Yet in many cases it's failed to deliver. There are three key reasons for this.

First, the AI aspect of machine learning requires big data that's properly curated. It's no secret that big data has been a challenge for businesses on its own. Thus, the requirement that for ML to work, it requires big data to work -- it simply compounds the issue.

The second hurdle when working on machine-learning projects is the fact that one must understand the current limits of the machine-learning platform and be able to ask the right questions that the ML and AI will be able to accurately answer. The learning curve to gain these skills is steep and there is a global lack of qualified data scientists to fill open roles.


Now entering its fifth year, the 2020 Vision Executive Summit is an exclusive meeting of global CSP executives focused on navigating the disruptive forces at work in telecom today. Join us in Lisbon on December 4-6 to meet with fellow experts as we define the future of next-gen communications and how to make it profitable.


Lastly, ML platforms available today provide the necessary tools, but not much more. Thus, to build a model specific to your business, you largely must develop it from scratch. This is time consuming and requires the right skills to do properly. For the most part, there aren't many open source or commercial ML platforms on the market that are specific to solving problems for various business verticals.

It's safe to say that even in 2018, machine learning is a high-risk, high-reward proposition if you go it alone. The bulk of the risk lies in capital expenditures for designing, configuring and maintaining the machine-learning platform. This is where machine-learning-as-a-service (MLaaS) comes into play. Leveraging the cloud allows businesses to eliminate much of the upfront capex risk as well as initial design, rollout and ongoing maintenance of the ML platform. The business can instead focus on proper data curation and API development to create the right analytics. It also significantly increases a business's ability to at least begin the process of experimenting with machine learning without a massive investment. This fact alone should bring in many new businesses to the machine-learning market that previously shied away due to huge upfront costs they couldn't justify.

MLaaS comes in two distinct platform flavors. First are the "Swiss Army Knife" style platforms that can do most everything you throw at them. The caveat, of course, is that these platforms require a great deal of effort to develop and fine tune to get the desired results. The second MLaaS platform type that's gaining popularity is the one that fits a specific niche. This MLaaS provider niche may be a unique business vertical -- or it could cater to a specific type of machine learning you want to leverage. Either way, these platforms get you much closer to what you want from a business sense compared to a generic platform that you must develop from scratch.

Related posts:

— Andrew Froehlich is the President and Lead Network Architect of West Gate Networks. Follow him on Twitter @afroehlich.

(0)  | 
Comment  | 
Print  | 
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
More Blogs from Column
Like Lucy always does to Charlie Brown, 5G is likely another new technology that teased optical networking vendors with massive, sustained spending, then quickly yanked back expectations.
More acquisition activity in the optical space is inevitable, but it might not happen for a while until some newly coupled companies get to know each other just a little bit more.
Cablecos must go beyond platform upgrades like DOCSIS 3.1 to stay truly competitive with the growing number of FTTH offerings from rival providers.
And how these changes offer fresh opportunities for video providers to engage sports fans in meaningful new ways.
ADVA's Prayson Pate looks at the key considerations when selecting the compute foundations for universal CPE.
Featured Video
Flash Poll
Upcoming Live Events
April 8, 2019, Las Vegas, Nevada
May 6, 2019, Denver, Colorado
May 6-8, 2019, Denver, Colorado
September 17-19, 2019, Dallas, Texas
October 1, 2019, New Orleans, Louisiana
October 2-22, 2019, Los Angeles, CA
October 10, 2019, New York, New York
November 5, 2019, London, England
November 7, 2019, London, UK
December 3-5, 2019, Vienna, Austria
December 3, 2019, New York, New York
All Upcoming Live Events
Partner Perspectives - content from our sponsors
Huawei Shows 5G in Action at MWC
By Ken Wieland, for Huawei
Huawei Heats Up Microwave for 5G Backhaul
By Ken Wieland, for Huawei
Huawei Services Bring the Best 5G Into Reality
By Steven Wu, President of Consulting & Service Solution Sales Dept., Carrier BG, Huawei
All Partner Perspectives