Artificial Intelligence (AI) – it’s the topic that’s setting the business world alight – self-teaching computer programs with the ability to make predictions, and decisions, more accurately than humans.
Sounds great, but don’t you need a brain the size of a planet, not to mention millions of dollars worth of the latest computer hardware and software in order to take advantage of it?
In other words, isn’t it something only available to multinational corporations or governments?
Well, no! In fact, organizations of all sizes are already taking advantage of the technology that is fundamental to the “fourth industrial revolution” – thinking machines – and millions of us are already using it in our everyday lives, often without even knowing about it.
When you search for information on Google, shop on Amazon or browse for entertainment on Netflix or Spotify, you’re putting AI algorithms to work. And just as cloud service businesses have spent recent years deploying these solutions to assist their customers, an increasing number of tools are becoming available which businesses can plug into, to offer their own intelligent, automated services to customers.
In this article, I aim to introduce some of the most popular ones available today. Of course, they all vary somewhat in cost and complexity, but they all have the same aim – to enable organizations to harness the incredible power of AI (specifically machine learning and deep learning) through the cloud.
Google’s Cloud AI service is designed to let businesses start building intelligent analytics and predictive algorithms on top of their data, even if they have limited experience and previous exposure to working with AI.
If you don’t know anything at all about AI or machine learning, being Google, it also offers consumer access to many of its features without any upfront costs – perfect if you want to familiarize yourself with the basics before making a case for trialling the platform within an organization.
Projects are set up through a simple console interface, and the whole service takes a modular approach, allowing Google’s different intelligent tools and services such as AutoML, Tensorflow (Google’s open source machine learning library) and Cloud Natural Language to be integrated when they are needed.
All in all Google's suite of cloud-based AI solutions has proven itself to be an excellent choice for individuals and organizations wanting to get started on learning the basics of machine learning and AI and provides plenty of scope for moving into more sophisticated and complex deployments as your confidence and understanding grows.
Paperspace may be the least recognizable name of the providers and platforms covered here if you're new to AI, but among the machine learning community, its platform is highly regarded as a powerful and versatile solution.
Like Google, Paperspace offers cloud-based access to powerful GPU hardware as well as a suite of libraries and algorithms to run on them – after a bit of reading up on the basics, all you need to do is plug in your data sources and start training.
Its machine learning/ deep learning interface is known as the Paperspace Gradient, which enables you to get started training neural networks with a simple click-based interface, or run Python code for more complex operations. Free accounts are available, and GPUs (including some of the most powerful processing units commercially available today) can be hired by the hour when you’re ready to start training on large datasets.
Those who have tried out a number of these cloud systems often comment that Amazon’s service is less straightforward to get up and running than some of its competitors. This is particularly true if you have no prior experience with setting up cloud services.
However, it makes up for this by being compatible with a huge range of different data standards, frameworks, and platforms. These include both open source solutions and those provided by other proprietary vendors.
Amazon provides two machine learning frameworks – Amazon ML and Sagemaker, which are suited to tasks of different levels of complexity. A drawback of Amazon ML is that it currently doesn't support unsupervised learning methods, meaning it requires labelled data in order to train algorithms. It does auto-select the most suitable machine learning method, based on the data provided, meaning that users don't need to understand different classification and regression methods in order to get started.
Sagemaker requires a bit more of an in-depth understanding of data science techniques and fundamentals but does allow the use of unsupervised learning methods, through clustering as well as its Neural Topic Model method, that classifies data into "topics" based on their frequency and context within a dataset.
Overall, Amazon’s AWS machine learning platform is one of the most well-established and widely-used ML-as-a-service offerings in the marketplace, and while it may not be the simplest to set up, is well supported with a great set of documentation online and a mature user base.
IBM’s cloud AI service is based around it’s “cognitive computing” engine, Watson. Watson made news headlines (and generated a lot of publicity for AI in general) when it beat a human champion at the question-based TV quiz show Jeopardy in 2011.
This perfectly demonstrated a key feature of Watson, which bases its business intelligence capabilities around a question-and-answer format, built on natural language recognition.
Since then, IBM has launched Watson as a dedicated, cloud-based AI business solution to compete with offerings from the likes of Google and Microsoft. In particular, Watson has proven its worth in the healthcare field where it is used to read and understand patient data, make diagnoses and suggest effective treatments.
Plug-in services and solutions include Watson Studio, which is used for training machine learning models and data preparation, and Watson Assistant which makes it simple to build and deploy chatbots and virtual assistants using your organization’s data. An example application here could be a customer service bot, which uses product and sales information, as well as customer feedback, to suggest relevant products and services.
Microsoft Azure AIAzure is Microsoft’s suite of cloud-based services, which includes a fully developed set of AI applications designed to be installed across an organization with a minimum of necessary hassle and knowledge.
As with others mentioned here, Microsoft includes pre-built, customizable applications for deploying machine learning and deep learning such as its Cognitive Services and Azure Machine Learning. However, it distinguishes itself with a focus on "edge" computing applications. Edge is a newer buzz-word in computing which refers to analytics and decision-making which is carried out as close as possible to the point at which the data is captured. A great example is an "intelligent" security camera system, which, instead of sending all the visual information it captures back to the cloud for analysis, is able to recognize and detect anomalies itself, minimizing the need for sending data back and forth, increasing efficiency and saving bandwidth.
So, as you can see, businesses are currently somewhat spoiled for choice when it comes to ready-made, easily-configurable cloud platforms from which they can deploy AI and advanced, cognitive analytics for prediction and decision making. Most of the services mentioned here come with a free trial, or free basic user access, so it isn’t hard to dive in and start to investigate what you can do with the data you have available. While they differ somewhat in terms of USP and “extra” features, as well as the level of experience required, they all offer the core services you will need to start preparing, training and deploying AI in your business.
It's undoubtedly true that AI is no longer a tool that can only be used by large, well-resourced businesses. In fact, as adoption increases, it's likely that those organizations that choose to ignore its potential could find themselves being left behind and struggling to compete.