We are witnessing the rise of algorithm economy to create, share, and remix algorithmic intelligence at scale to enhance Machine Learning (ML).
ML has moved beyond the ability to just apply the best algorithms on a given data set to learning predictive models and applying them to new data and integrating the applications that we use to make those applications “intelligent”.
Creating Application Programming Interfaces (APIs) are part of the solution to make it workable with a range of softwares and hardwares, but there’s more.
The combination of ML and Internet of Things (IoT) is a source of innovation for predictive applications that run on a variety of devices. Nowadays, most organizations have initiated IoT on underlying sensors, devices, and “smart” things and develop human-impact IoT use cases for boosting food production, cutting carbon emissions, transforming health services and making consumer products for everyday use.
Amazon echo, for instance, functions as an all-in-one problem solver. Alexa, the device’s vocal persona, answers questions, controls smart devices, acts as a speaker, and even shops—all processed by voice commands.
Moov is a next-generation fitness wearable that aims to improve the user’s exercise goals by acting as a coach. Through a trio of motion sensors, Moov analyzes the wearer’s movements during activities, such as running or cycling, and transmits both audio and visual feedback in real time.
The sleek Ninja Sphere is an all-in-one platform that connects an array of smart devices. Using simple hand gestures, users can do things like call an Uber, shut off the lights, or manage home climate control. The Sphere also uses location, environment, and device data to make informed recommendations to users.
Magic Leap is a wearable device that projects 3-D images directly onto the wearer’s eye to create uninhibited, 360-degree augmented reality. While the company has been silent about a release date, Magic Leap promises to put a new spin on the way we interact with our physical environment.
Some companies are managing data, leveraging IoT infrastructure, and developing new business models from them , since IoT typically generates volumes of critical data, which poses a scaling issue. We start creating a different model for each user of our intelligent things and end up having more models than there are data scientists in the world. For that, Salesforce uses “AutoML” strategies, to automate ML and 3rd-party ML-as-a-Service platforms.
MLaaS platforms (internal or 3rd-party) is a viable solution that removes infrastructure concerns and helps in ML integration. MLaaS eases the deployment of predictive models, as well as the experimentation with various techniques and datasets (till we choose the right model to deploy).
3rd-party MLaaS platforms like DataRobot, Dataiku , BigML or MLDB.ai provides an open source solution to build our own platform. For Deep Learning, platforms like Nervana (now part of Intel) have implemented speed-ups all the way from high-level library choices down to leveraging custom silicon.
Startups such as Datatrics are using these platforms to build innovative products like predictive marketing solution called “Next Best Actions” for SMEs that give them actionable insights.
While ML and IoT together collect data of increasing variety, to predict more and more aspects of people’s behavior and take action and impact this behavior, we should simultaneously consider how this data must not play with sensitive information about individuals, and impact on society. Long-term and wide adoption of machine intelligence and intelligent applications requires solutions to learn from data in a privacy-conscious way.