Prediction for AI in Networks for 2018

This article is written by Entuity’s CTO, Lee Walker.

2017 was a busy year for all of us in the IT services industry! The network gained a lot of value, often being cited as the core of business success and the driver of exceptional user experience. The focus has shifted to be on the users, not the technology.

We saw rising sentiment to move from reactive models to proactive, whether that be by exploring the benefits of machine learning, or by refining how data is presented in order to help IT professionals get to what they need, faster.

We saw another year of major disruption by cloud services as more businesses migrate applications and workloads. Vendors have had to stay on their toes, developing multi-cloud strategies in order to stay ahead.

What’s in store for 2018? This is the first in a five-part series exploring my top 5 predictions for the year ahead. First up, Artificial Intelligence!

2017 saw AI’s prominence grow

Google’s artificial intelligence program, DeepMind, made headlines in 2017. First by beating several of the world’s top Go players, an ancient and complex game of strategy and intuition which many believed could never be cracked by a machine. Then, in December, it was announced that DeepMind had beaten the world’s top chess computer program, StockFish 8, playing 100 games and losing none – after learning the rules of chess just 4 hours earlier. Quite a feat! Though it’s worth noting that with both of marvels, DeepMind was given the rulebook for both games and then played itself over and over, learning from its experience, and that’s a key consideration. This is where AI differs from Machine learning – if “AI” is defined as the overall intelligence, “Machine Learning” can be defined as the algorithms that help it learn to be more intelligent.

AI beat the top Go players last year

Defining AI Use Cases

Machine learning in network monitoring is still very much in its infancy because there is no hard and fast rulebook for managing a huge, sprawling, distributed network. Before we apply machine learning, we first have to understand where it can be of most benefit. We’ve seen various methods employed so far including pattern discovery and prediction that learns what is normal on a network (dynamic thresholding) and helps to prevent issues before they impact the business. Attempts have been made to reduce alert fatigue by learning to filter important alerts/events from extraneous ones. And we’ve seen collective intelligence, whereby anonymized customer data is pooled and insights shared for more proactive remediation, for example; a customer reports an issue with a patch, that insight is fed proactively to other customers before they apply the same problematic patch.

In 2018, we can expect to see a clarification of the use cases for AI in networks. What repetitive tasks can AI lend it’s digital hand to? Can AI remediate without human intervention? Machines learn from patterns, but before we can teach machines, we must first understand the human patterns of behaviour. How does a human solve a network issue? How does a human drill down to find root causes? The network is a complex web of interconnectivity. AI will have it’s place in helping us navigate it, but rather than replace humans, AI will augment our abilities to do tasks more efficiently, allowing us to move from a position of reactivity, to focusing on generating greater business value.

AIOps is about bringing together technology and people

AIOps brings technology and people together

We’ll see growth in the field of AIOps (Artificial Intelligence for IT Operations). AIOps uses machine learning to garner continuous insights from big data. IT data is often siloed in different platforms, and AIOps encourages aggregating that data into a single observational platform. It’s then possible to connect the dots, join universes and see trends where only disparity previously existed. But AIOps is about more than just consolidating data and technology, it’s about bringing people together. Removing technical obstacles and simplifying user experience gives people more time to do what’s important – connect, communicate and create.

As AI creeps into network management software, we’re seeing more and more vendors start to re-market themselves as “analytics” platforms. Descriptive analytics (the analytics used to describe things that have happened or are happening now in your network) are being superseded by Predictive analytics (the analytics that predict what might happen in the future). It may be some time before we start seeing the next level of analytics – Prescriptive (the analytics that predicts what will happen, and then prescribes a course of action, monitors and learns from its actions) – because this type of analytics requires quite advanced levels of AI. It’s going to be an exciting year for AI, that’s for sure!

Next week, Lee will talk about his 2nd prediction for 2018 – Cloud.

Share This

Share this post with your friends!