Dec 03 2018
If there are two major buzzwords in tech right now, they are “Cloud Services” and “Artificial Intelligence,” so it’s inevitable that they’d be combined. AI and the cloud are complex though, so here’s our primer to understand what AI cloud services do, and whether they could matter for your business.
Cloud providers are offering AI as a service as part of a “service stack” — a combination of features, functions, and resources that allows businesses to run simple and complex AI and machine learning algorithms on data and apps that run in the cloud. From cognitive computing to conversational AI and data lakes to batch processing, the range and depth of AI services does vary between providers.
There are two main drivers behind AI as a service — “Compute” and “Data.”
Compute refers to both the range of services you can run against data, and the power (speed, efficiency, capacity) of those services. Typically, AI compute services consume a large amount of “cycles” from cloud infrastructure like graphic processing units (GPUs) and central processing units (CPUs).
As you introduce more factors into AI processing like increased data points, different simulations, larger data sets, etc, the amount of GPU and CPU cycles increases exponentially. Because cloud services are typically charged “on demand,” each GPU and CPU cycle will have a cost associated with it. That’s important when it comes to managing your budget for AI processing.
Compute services also expand beyond simply running AI algorithms — they can power big data frameworks and languages like Hadoop or Spark. They can allow processing across containers and virtual machines. They can power batch AI services and work across both traditional and serverless environments.
Whereas compute provides the tools to get AI insights, data provides the raw information that the tools extract those insights from. Data is the lifeblood of AI — it is only through processing large datasets and creating simulations and models that machine learning really delivers benefits for your business. Machine learning models run complex AI algorithms against your data to teach themselves and provide you with information so you can tweak parameters and processing until you get a high-quality output.
In many cases, how the data is stored won’t be an issue. Most AI as a service works across relational databases, serverless environments, data lakes, and other use cases equally well, although functionality does differ between providers.
Of course, another vital factor of AI computing is accessing and integrating the data through APIs, and allowing for easy input, output, and processing via AI tools. Cloud providers are building a range of AI tools and APIs to make it easier to build bespoke services on top of the AI as a service platform.
If you’re thinking of moving to AI as a service, talk to us first. We can help to understand what your business needs are, and match them up with the cloud provider who will give you the most value and insight.