Proliferation of Data Driving Machine Learning
Machine learning is a hot topic, but what applications is the technology driving and what drives machine learning? The short answer, concludes a new research report on machine learning, is data—lots of data.
“Data is available in volumes never dreamed of, which means the algorithms have more to get their teeth into,” concluded a survey by market watcher 451 Research, noting that “machine learning lives or dies depending on the quantity and quality of the data available.”
Moreover, the rise of application containers in software development “means small [machine language] applications can be placed into software without breaking the rest of the application,” the report’s authors continue. With enterprises large and small going digital and computing resources ubiquitous, there’s plenty of data available with which to train rather than simply program machines.
Despite great strides by hyper-scale infrastructure providers such as Google (NASDAQ: GOOGL) in areas like image and speech recognition and language translation applications, the analysts are reserving judgment on whether broader application drivers will emerge anytime soon.
Machine language “is trending now because it is seen as a new competitive weapon—to make more money or save money—as well as being a new butterfly to chase,” the market watcher asserts in a report released earlier this month.
Along with a “huge influx” of machine learning-based security applications for detecting ransomware and other cyber attacks, the report authors note that “security in big data—in the sense of Hadoop-scale analysis—is exciting because it sees correlations across data silos over long periods of time.”
Along with data, the driving force behind machine learning is the ability to train rather than program machines as a way to automate repetitive tasks. In IT operations, therefore, machine learning is catching on as a way to automate routine processes so developers can focus more on, for example, getting distributed applications out the door. Hence, machine learning is gaining traction in the datacenter for automating IT operations such as incident response using step-by-step training methods. That approach is gaining “traction with telcos and service providers as they seek to become more efficient in selling and solving customer problems,” the researchers said.
Telecom providers, as an example, operate networks using machine-learning techniques that can respond to shifts in traffic patterns while introducing predictive analytics as workloads shift across those networks. “As companies automate and mobilize mission-critical processes, low latency and error rates will need to be addressed,” the report notes. Response rates must be significantly improved, the authors added, and network operators “will need to reduce latency to the single digits.”
Among the other key challenges to wider adoption of machine learning is the availability of training capabilities. While machine-learning technology is ready for primetime, “we believe market education will play a major role for companies to start integrating smart bots into their customer service organizations,” the 451 reports notes. “The main barrier across all use cases is the availability of training sets.”