Machine learning, the field of artificial intelligence based on self-improving algorithms, is not overly new to the business world. But with techniques such as deep learning, an approach informed by the brain, increasingly coming to the fore, businesses are identifying new and potentially transformative uses for the technology.
A significant leap forward for deep learning came in 2012, leading it to become the undisputed king of machine learning approaches. These included neural networks being shown to outperform other models in images and speech, while facial recognition took a step forward as Google trained its model to recognise that ubiquitous internet genre, cat videos. Following on from this, a new generation of important machine learning tools and software arrived, including TensorFlow in 2015, while hardware costs decreased as the highly parallel nature of graphics processing units (GPUs) began to be taken advantage of..
According to Algorithmia’s 2020 report, the main use cases for machine learning currently being explored by businesses are in customer service (i.e. chatbots) and internal cost reduction. But there are a vast array of potential applications. Dynamic pricing, for instance, where a system learns about factors like customer interest, demand and history to adjust prices and entice purchases. Or the automation of tedious manual processes such as classifying and labeling data sets - images, for instance. Churn modelling is another highly prized area of exploration, affording companies the ability to predict which customers are likely to be lost and allowing corrective measures to be undertaken.
It’s clear that machine learning has a huge amount to offer enterprise - but it’s tricky to know where to get started. AI for AI’s sake is a real and ever present danger, with the proliferation of AI solutions meaning some are inevitably akin to snake oil. Even when objectives are clear, there can be a long and winding road to deployment thanks to factors such as difficulty scaling, machine learning technology evolving while a project is underway and difficulties assigning correct budgets thanks to different levels of understanding across company hierarchies.
While open-source machine-learning systems are available, they require both significant hardware and the right kind of data science expertise (made harder by the fact the world is experiencing a deficit in people with such skills).
Enter the cloud, one possible solution to this quandary, allowing businesses to leverage the necessary hardware without the capital expenditure and affording them stable platforms with which to work. Companies can now look forward to the benefits from machine learning without having to themselves becomes technology specialists.
Google is perhaps the company most associated with machine learning, thanks to its development of the open-source TensorFlow platform, as well as its association with one of the most advanced machine learning companies - DeepMind and its programs such as AlphaGo.
Intended for enterprise use, Google Cloud’s AI Platform combines and integrates different aspects of the machine learning pipeline, from data storage and labeling, to training to deployment.
Lucas Ngoo, co-founder and CTO of ecommerce marketplace Carousell said: “In retail, it’s important to provide customers with easy access to alternative products or recommended add-ons. We train our own machine learning models with TensorFlow on AI Platform, and we automate the periodic retraining of these models with Kubeflow Pipelines. Together with AI Hub, useful for sharing models between data scientists, we can now iterate faster on our models, and automatically deploy them to staging and production.”
Amazon’s cloud service, AWS offers a wide range of machine learning solutions on the cloud, with Amazon claiming that more machine learning happens on its platform than anywhere else. Of particular note is Amazon SageMaker, which is focused on simplifying the process of building, training and deploying machine learning models. It does this in part through a web-based visual interface allowing for the uploading of data, the tuning of models and comparisons of performance.
AWS has also developed specific hardware for machine learning, with an inference chip known as Inferentia, which is intended for sophisticated applications such as search recommendations, dynamic pricing and automated customer support, and is accessible through the cloud.
One prominent customer of AWS’s approach to machine learning is Disney, which is using machine learning to tag and categorise its old content with metadata.
Microsoft’s Azure cloud platform has built-in machine learning services for enterprises looking to bring machine learning models to bear. With a stated focus on MLOps, the subset of DevOps dealing with correct machine learning development practices, it includes both code-based and drag-and-drop environments in order to accommodate users of all skill levels.
Azure also has a focus on the potential perils of machine learning, building in so-called ‘responsible machine learning’ solutions to mitigate bias in models.
Matthieu Boujonnier, Analytics Application Architect and Data Scientist, Schneider Electric, said: "With Azure Machine Learning, we can focus our testing on the most accurate models and avoid testing a large range of less valuable models. That saves months of time."
With the proliferation of machine learning services on the cloud driving down costs and opening up possibilities, expect companies of all shapes and sizes to leverage the technology going forwards, opening up new methods of customer interaction, as chatbots are proving, and highlighting areas in need of efficiency.