Machine learning is moving into the mainstream thanks to advances in processing power, deep learning, and Big Data.
Computers generally don’t take sick days or pass on germs, most of the time they don’t argue or answer back and they never have hangovers or wake up in a bad mood. In theory that could make them the perfect digital co-worker.
But there is a problem. Unlike the human brain, computers have historically suffered from a key limitation – they must follow the rules laid down by programmers. If it is not in the code, it does not compute.
Now however, advances in processing power, Big Data, and a technique called Machine Learning, is changing all that. Machine Learning was first defined in 1959 by Arthur Samuel as a, “Field of study that gives computers the ability to learn without being explicitly programmed.”
The basic idea behind Machine Learning is to teach a computer to spot patterns and make connections by showing it a large volume of data. In one form of Machine Learning called supervised learning, the computer is fed example inputs and their desired outputs, with the goal of generating a general rule that connects inputs to outputs.
Examples of the use of this form of Machine Learning in today’s consumer-centric world include the recommendation engines built into Amazon and Netflix services, and the face recognition capabilities of Facebook.
Meanwhile cloud service providers including Microsoft, IBM, Google and Amazon’s AWS unit have all launched Machine Learning service that let developers build smart, data-driven applications that can learn based on past data.
The modern theory of Machine Learning grew out of the study of pattern recognition and computational learning theory in Artificial Intelligence in the 1980s. However, it was the recognition in the past 10 years of the parallel processing capabilities of graphic processing units (GPUs) (developed initially for the video gaming industry) that has turned Machine Learning into a practical and cost-effective reality.
This increased computing power, coupled with other improvements including better algorithms and deep neural networks for image processing and ultra-fast in-memory databases like SAP HANA are the reason why Machine Learning is one of the hottest areas of development in enterprise software today.
Another key factor is and the growing availability of structured and unstructured Big Data from a rapidly expanding number of sources including IoT sensors, digitized documents and images. This data can then be used to ‘train’ machines and enable them to make accurate predictions and recommendations.
Business users and other enterprises that are able to leverage their data in this way should be able to increase their revenue, improve customer satisfaction and reduce cost. In a very real sense, Machine Learning transforms enterprise data into business value.
As a result, enterprise software vendors have been investing heavily in Machine Learning – either organically or by acquisition – and have made it a strategic priority. SAP, for example, has committed to eventually make all enterprise applications intelligent and widely available by building Machine Learning on SAP HANA Cloud Platform
Unlike some of its competitor, SAP’s strategy is customer and ‘apps led’ – starting with business problems, then identifying the appropriate business Machine Learning services and technology to address the issues.
At SAP, the push to make all apps intelligent is being led by a team of software engineers, researchers, designers and business developers in the SAP Innovation Centers in Singapore, Potsdam, Walldorf, Ra’anana (Israel) and Palo Alto. The team is led by Markus Noga and is part of the SAP Innovation Center Network headed by Jürgen Müller.
As part of the effort, the SAP team is piloting five applications:
- Automated SalesForecast: Only 44 percent of executives feel that their company is managing sales effectively and only 22 percent of companies are able to make accurate sales predictions. SAP is using Machine Learning to predict which opportunities will close and recommend sales representatives the best possible actions to move the deal forward based on data in SAP Hybris Cloud for Customers and unstructured text from emails and the Web.
- CV Matching: Recruiters spend substantial amount of time going through CVs trying to find the best match for open positions. With the use of Machine Learning, SAP has been able to significantly reduce the time to shortlist the best candidates for a particular job – or the best job for a particular candidate.
- Invoice Matching: Manually matching payments to invoices is one of the most labor-intensive processes in accounting, and is often handled by shared services centers. Machine Learning can significantly increase automatic matching rates providing a real world example of an intelligent digital co-worker.
- Social Media Customer Service: Social media community managers and support agents can be overwhelmed by posting volumes, i from Twitter and Facebook for example. This app automatically tags and clusters in-bound messages and suggests appropriate responses.
- yaaS Recommender: Most e-commerce shops suffer from low conversion rates as they have no, or only, basic recommenders. yaaS recomender analyzes full customer clickstreams to provide tailored, contextual recommendations in real time.
These examples demonstrate how Machine Learning can bring intelligence and additional value to the digital enterprise. In the process, they could also change the relationship between man and machine further freeing people from mundane and repetitive tasks in order to focus on what humans do best; Think and create.
As Bill McDermott said during his SAPPHIRE NOW keynote: “I think very strongly that intelligent applications will fundamentally change the way you do work in the enterprise and the way you collaborate with your trading partners outside of the enterprise.”