Cloud Machine Learning - The Rise of Machine-Learning-as-a-Service
With more and more technology being floated into "the Cloud" I wonder what wonderful, life-enriching nutrients are going to rain down next. In one of my all-time favorite movies, super-nerd Flint Lockwood sends a machine into the clouds and causes it to rain hotdogs, pizza, and ice-cream. Now that would be pretty awesome.
Big Data Analytics Made Easy?
Last week, HP Haven OnDemand opened for business with its Machine-Learning-as-a-Service (MLaaS). Although it has been around as a freemium service for a few years now, the recent announcement was more about pricing that extends with usage and is backed by enterprise Service Level Agreements (SLAs). That is probably quite significant coming from the $50 billion behemoth that is HPE. So are we witnessing yet another cloud-based revolution and what can we expect to rain down over the next few years?
Haven OnDemand's HPE IDOL search engine (the big data and analytics driven search engine that’s the successor of HP Autonomy) is the pattern-matching brain in the machine that is being used by developers to create some pretty novel applications.
For example, Blink is a mobile phone speed dating app that monitors user’s faces to make sure that the correct body parts are in conversation. RingDNA, an Inside Sales Software company, has started using the API to perform live conversation analytics to understand sentiment and predict outcomes of sales interactions. Another startup, Social Capital, is using the service to support HR departments in profiling job applicants based on their timelines in social media; tweets, photos, videos, warts and all. Yes folks, that moment we all dread is becoming a reality: our social media is going to catch up with us! But I suppose it’s better to be judged by a machine than an HR director.
Do you remember Watson, the IBM super computer that actually won at Jeopardy live on TV? Well Watson didn’t take his winnings and retire. Guess what, Watson too is in the cloud and it is also learning. BBC's “Tomorrow’s Food" program even featured recipes developed by Chef Watson that analyzed millions of recipes and by identifying previously unimagined patterns actually developed new food combinations that people like.
IBM’s vision for Watson extends far beyond Chef Watson’s crazy recipe ideas. Watson also has in-cloud cognitive APIs for language, vision, speech, and data. The APIs have names like “Tone” and “Emotion Analysis” and are able to detect anger, disgust, sadness, fear, and joy. With this kind of capability some might start to wonder whether Orwell's Big Brother was indeed called Watson; but I think we should be fair; all technology has always had the propensity to be used for good and evil.
Machine Learning Use Cases
Just consider the potential for cloud machine learning in various use cases. For example, utilizing machine learning in e-commerce to predict shoppers’ preferences through log data and then show them personalized products that they are most likely to buy. Or in the recruiting industry, machine learning can “learn” feedback from recruiters and key influencers in order to systematically improve the effectiveness of future candidate search and match. I am going risk using a major cliché, but it seems the possibilities are endless.
These examples do suggest that the humble search engine is on the verge of revolutionizing society, yet again, when combined with big data, predictive analytics, and machine learning (as Hadoop creator Doug Cutting said here). For search technology enthusiasts like myself, despite the lack of hotdogs and ice cream falling from the cloud, it's still pretty awesome to know that the machine is actually learning by pattern matching.
It turns out that search as a pattern-matching technology can be leveraged to find patterns not just in structured data like words, but also in unstructured data like words in the context of real life things like expression in faces, intonation in conversations, and even flavors in recipes.
If machine learning is Arthur Samuel's 1959 definition of "a field of study that gives computers the ability to learn without being explicitly programmed" then indeed we might be witnessing a new era where cloud machine learning makes its insights available for application developers to build the next-generation of life-enriching big data applications powered by search. Still thinking about how machine learning fits your business use cases? Start a big data assessment to find out.