Machine learning’s potential is undeniable; Will it be the most disruptive class of technologies over the next few years? Author Girish Raichur - Associate Manager
www.blueoceanmi.com |
[email protected]
What is machine learning? Computing systems that become smarter with experience | past data + human input = “experience”
Background In October 2015, Elon Musk1 called artificial intelligence “our greatest existential threat,” and equated making machines that could think, with “summoning the demon.” In December 2015, Stephen Hawking2 said “full artificial intelligence could spell the end of the human race,” while Bill Gates3 said he was “concerned about super intelligence” only a few decades away. However, if the human race is at peril from killer robots, the problem is probably not artificial intelligence. It is more likely to be artificial stupidity. The difference between those two ideas says a great deal about how we think about computers. In the early 1990s, executives and managers welcomed information technology — databases, PC workstations and automated systems - into their offices. They saw the potential for significant business gains. Computers didn’t just speed up processes or automate certain tasks, they could upset nearly all business processes and allow executives to rethink operations from the ground up. Thus reengineering movement was born. Now it’s happening again! Machine learning, already one of the most versatile technologies of the past decade, will gain even more traction in a digital environment. Machine learning executes AI (artificial intelligence). Algorithms enable computers or machines to detect patterns, predict future outcomes and train themselves on how to best respond in certain situations. The idea has been around for 50 years, but big data and the need to make sense of it, is driving its recent adoption, along with great improvements in many AI disciplines.
Use cases and benefits Machine learning already has a variety of applications ranging from marketing field to behavioral pattern assessment, and healthcare, with accurate and early detection of complex diseases, to infrastructure with smarter urban planning. It is employed in a range of computing tasks such as spam filtering, optical character recognition (OCR), search engines and computer vision. Netflix's movie recommendations, Amazon's product recommendations, Facebook's ability to spot your friends faces, dating websites that match you with potential dates — these are all early examples of machine learning. Today, machine learning algorithms enable computers to communicate with humans, autonomously drive cars, write and publish athletic reports and even help locate terrorist suspects. There is also excitement around the prospect of applying machine learning to fraud management and speech recognition. As the field grows, many more industries are expected to be affected by automation. Some of the impacted jobs that ZDNet cites as topping the list are field technician, insurance underwriter, tax preparer, translator and fast food employee. Machine learning’s potential is undeniable, and it is viewed as a “decade-long opportunity.” Gartner has opined that smart machines will be the most disruptive class of technologies over the next 10 years, while a July 2015 TechCrunch article states that enterprise software is about to undergo a radical transformation powered by machine learning. There’s an abundance of examples across organizations that demonstrate what can be done with machine learning. Here are just a few4: Sales and marketing – Machine learning models used for product recommendations can be built to predict which product a customer is most likely to buy. A customer’s profile is taken as the input (customer activities, recent purchases and personal information) and mapped to the predicted likelihood of the customer responding to a given offering. Risk and fraud management – Machine learning in fraud detection is typically used to map descriptions of transactions to their likelihood, indicating whether an ongoing transaction has a high likelihood of being fraudulent. In the case of credit risks, it can map the loan applicant’s details (demographics and credit/payment history) to the likelihood of them defaulting on a given loan. Smart transportation – Traffic optimization can be achieved through an understanding of traffic patterns using sensor data, accidents and roadworks. A machine learning model can predict delays or road obstructions and recommend a faster route for public buses, as well as for consumer and commercial vehicles. Sources:
1. 2. 3. 4.
http://observer.com/2015/08/stephen-hawking-elon-musk-and-bill-gates-warn-about-artificial-intelligence/ http://time.com/3973500/elon-musk-stephen-hawking-ai-weapons/ http://time.com/3973500/elon-musk-stephen-hawking-ai-weapons/ http://www.gartner.com/smarterwithgartner/what-we-can-do-with-machine-learning/
copyright 2017 Blueocean Market Intelligence
1
Supply chain processes – Machine learning models in asset performance management can take the operating conditions of assets, such as wind turbines, solar panels and nuclear reactors as input, and predict when failures can occur. The objective is to decrease maintenance costs and minimize downtime. Healthcare – Machine learning models in early-warning systems for employees can analyze sensor data in hazardous environments — such as measurement of air quality, equipment performance or employee productivity, or even atypical behavior — to predict the likelihood of accidents. This application has been widely adopted to alert truck drivers of potential accidents.
Current scenario and players Machine learning is considered to be a key catalyst in redefining and expanding what is possible in computer science. The interest in this approach is at an all-time high and causing a flurry of activity among competitors, venture capitalists and emerging companies, both product and research-related. Major vendors like Amazon, IBM (Watson), Google (an algorithm created by Google-mastered Go, the ancient Chinese game, in March 2016), and Microsoft, are acquiring and investing in R&D and new products, while companies such as Uber and GE are staking much of their future on machine learning. There is a pattern-finding race among competitors and firms that are maneuvering for position, poaching researchers, setting up laboratories, and buying startups, with Google being the number one acquirer so far. Forrester estimates that the number of machine intelligence startups in 2014was more than 2,300. Machine learning had a major public breakthrough of sorts in March 2016, when an algorithm created by Google mastered Go, the ancient Chinese game with more possible board configurations than there are atoms in the universe . Google's self-driving car has also become a recent machine learning case study. Additionally, several companies, including Facebook, Google, IBM, and Microsoft, have donated machine learning development projects to open source. The next wave of software isn’t expected to emerge from traditional software leaders. Rather, it is being pioneered by consumer Internet players like Facebook, Google, and Twitter, who have been using machine learning techniques for years to analyze and act on large volumes of data, and are seen as having a disproportionate advantage.
“
Risks
Currently, machine intelligence has been more disruptive in an industrial sense than on a personal level. However, this is expected to change. Companies are now starting to compete in a new battlefield - the move to mobile.
”
While machine learning’s role as a crucial addendum to big data platforms and services is a top benefit and despite significant interest in its potential, machine learning is not without risks. Data quality is a primary limitation, along with human bias. Other hurdles include user attitudes and lack of skill sets, enterprise organization and cultural norms, and high costs. Results are only as good as the data that is provided, making data quality one of the top limitations mentioned due to lack of available, accurate data; sampling error or bias; model complexity; data leakage from the future; and more. Human bias is a closely related risk. According to TechCrunch, in order to use machine learning and public data responsibly, we need to have an uncomfortable discussion about what we teach machines and how we use the output. Hurdles also remain around user attitudes and skill sets, and enterprise organization and cultural norms. In short, as the technologies are developed further, social, business and personal attitudes about smart machines and the nature of work must also adapt and evolve. Gartner says that successful adoption of machine learning is predicated on finding talented data scientists who can execute the technology while understanding its pitfalls and limitations.
copyright 2017 Blueocean Market Intelligence
2
Conclusion If machine learning is going to fulfill its potential and enhance the quality of our lives, it will require a change in how we view our relationship with machines. It requires moving beyond the “on-demand” approach we typically take towards technology and instead allow it to be an interruption. This next generation of data products is going to be more intrusive and we have to come to terms with that before we reap the benefits.
Sources: 1. http://www.recode.net/2016/6/1/11833340/bill-gates-ai-artificial-intelligence
2. http://observer.com/2015/08/stephen-hawking-elon-musk-and-bill-gates-warn-about-artificial-intelligence/ 3. http://time.com/3973500/elon-musk-stephen-hawking-ai-weapons/ 4. http://www.zdnet.com/article/hpe-touts-biz-analytics-resurgence-in-apac-with-machine-learning/ 5. http://www.forbes.com/sites/louiscolumbus/2016/06/04/machine-learning-is-redefining-the-enterprise-in-2016/#5b03806a5fc0 6. https://hbr.org/2016/02/companies-are-reimagining-business-processes-with-algorithms 7. http://whatis.techtarget.com/definition/machine-learning 8. http://www.popularmechanics.com/technology/a16630/get-much-smarter-about-machine-learning-in-2-minutes/ 9. http://bits.blogs.nytimes.com/2015/07/11/the-more-real-threat-posed-by-powerful-computers/?_r=1 10. http://www.kdnuggets.com/2015/08/gartner-2015-hype-cycle-big-data-is-out-machine-learning-is-in.html 11. http://techcrunch.com/2015/08/02/machine-learning-and-human-bias-an-uneasy-pair/ 12. http://searchitoperations.techtarget.com/feature/IT-pros-get-a-handle-on-machine-learning-and-big-data 13. http://www.forbes.com/sites/bernardmarr/2016/02/19/a-short-history-of-machine-learning-every-manager-should-read/#39424ef0323f 14. http://www.techinsider.io/machine-learning-as-important-as-the-internet-2016-3
copyright 2017 Blueocean Market Intelligence
3
About Blueocean Market Intelligence Blueocean Market Intelligence is a global analytics and insights provider that helps corporations realize a 360-degree view of their customers through data integration and a multi-disciplinary approach that enables sound, data-driven business decisions. Since we live in a highly dynamic and multi-dimensional world, we believe the most effective business decisions come from a synthesis of data streams and not from one-dimensional sources. Using our 360 Discovery approach, we ensure the comprehensive use of all available structured and unstructured data sources, enabling us to bring the best to bear against each engagement. Strong decision support is enabled through a combination of analytics, domain expertise, engineering and visualization skills brought together in harmony. Leading companies have benefited from our partnership with financial growth, 360 views of their markets and competition, and improved customer acquisition, satisfaction and retention. For more information or to request a consultation, please email
[email protected] or visit us online at www.blueoceanmi.com. United States | United Kingdom United Arab Emirates | India
