“Easy” probably isn’t the first word that comes to mind when you think of machine learning (ML) and AI. But that might not be the case for long.
While the complexity and capability of ML continue to grow – e.g. in the last year alone, deep learning has significantly advanced NLP, GANs and recommender systems– the ML software development process is getting simpler.
Use ML Without Reinventing the Wheel
That’s because, in the race to determine who will become the industry standard AI platform, some of the industry’s major players are opening up their AI capabilities to third-party developers. For example:
Google’s AutoML allows developers to create ML software through a drag-and-drop process.
AWS’ APIs and Deep Learning AMIs help developers add intelligence to their products in a plug-and-play manner, and provide the infrastructure and tools to accelerate deep learning in the cloud.
Microsoft’s Computational Network Toolkit helps developers create deep learning models for things like speech and image recognition.
Apple’s Core ML allows developers to easily integrate ML models into apps using just a few lines of code.
With AI capabilities at their fingertips, developers can, for the first time, incorporate ML into custom applications in a way that’s not exorbitantly expensive or time-consuming.
Finding the Right Models
As platforms open up their ML engines, and more ML models are freely available, developers will need to focus on finding the right models for their needs. For example, if a developer is building an app to help health providers better manage EHRs via voice recognition; they might want to consider a deep learning model like Deep Boltzmann Machine (DBM). Similarly, if a developer is creating a retail app to recommend or promote purchase pairings based on previous user data (e.g. if someone buys chips, they’re likely to buy beer), they might want to incorporate association rules through something like the Apriori algorithm.
But success doesn’t just depend on choosing the right model. There is a whole host of other things that have to happen to make an intelligent app work – things like training the model with the right training set and testing it to evaluate the performance of the algorithm on a particular function/problem. But the hardest work (actually creating the ML engine itself) is already done. Developers with ML integration expertise (like DVmobile) can help organizations work through the rest of the checklist fairly easily. And as a result, we’re beginning to see the first big wave of third-party intelligence-integrated apps hit the market.
If your organization is ready to ride this wave and experience the power that ML can bring to applications, operations, and customer engagement, let’s connect.
Guest blogger for DVmobile, busy mama, & strategy wiz.