The Artificial Intelligence (AI) vs Machine Learning (ML) Debate
AI has been a forefront topic in our minds. We’ve seen AI emerge as a major theme in movies and TV shows, many of us use it every day when we ask Google and Apple maps to gauge traffic, and our smart assistants (Cortana and Alexa) seem to epitomize the AI experience. But is AI really any different than machine learning in UC? A buzzword isn’t necessarily a negative unless you are trying to be specific and descriptive. So, while machines are automating all kinds of tasks including helping optimize UC functions, using these buzzwords interchangeably is doing everyone a disservice. Despite working together to optimize unified communications, AI and ML are different in a couple important ways.
Let’s break this down and start where it all began. The term “artificial intelligence” was first coined in the 1950s when Claude Shannon introduced a proposal called “Programming a Computer for Playing Chess.” Years later, the idea came to fruition, and 1997 saw one of the most famous instances of machine vs. man, when Deep Blue, an IBM supercomputer, won a chess match over World Chess Champion Garry Kasparov. For the most part, what was considered AI then isn’t considered AI today, and what’s considered AI today might look very different in the future.
One of the biggest misconceptions in the unified communications space is that AI is the brain behind the operations. In reality, it’s the other way around – ML is one way of achieving AI. The statistical techniques to “learn” with data without being specifically programmed (a.k.a. machine learning) allow a computer to perform tasks that normally require human intelligence (i.e. artificial intelligence). If entity A and entity B are both asked a question, and you can’t tell which entity is a machine, then AI has been achieved.
Machine Learning and UC Intertwined
More specifically, ML is a field of computer science that gives computer systems the ability to use algorithms to progressively improve performance on a specific task. Whereas some algorithms always produce the same output, an ML algorithm analyzes data and outputs to improve its program every time. A traditional program spits out data, but ML takes data and incorporates it into the program. In UC, ML algorithms can analyze data from systems and users – signaling events and errors, endpoint characteristics, device characteristics, network characteristics and server inputs – to identify and address issues before they become incidents.
ML also enables targeted user engagement. User data like typical UC usage and quality, endpoint and devices used, common times and locations for communication, frequent contacts and session profiles can help inform the creation of a user persona profile. This enables more automated user engagement and can drive better user behavior to improve overall UC quality. For example, if End User A has experienced five bad calls in a given month, the ML algorithm can deduce the root cause of the issue and propose a solution. AI becomes involved when we take this information and incorporate decision-making to give us output on how to resolve End User A’s call issues. You can see this in action with Unify Square’s patented UC-Core™ technology, which leverages AI technology to sharpen visibility in UC environments and deliver actionable insights.
Why AI Needs Love, Too
If it wasn’t for artificially intelligent machines, machine learning in UC couldn’t exist. For example, people generally wouldn’t perceive a conference call as artificially intelligent. But, when we introduce ML to help optimize the call experience and use proactive engagement to calculate active speaker quality and probability, ML can proactively identify and resolve call issues. If the users have the opportunity to “rate my call” afterwards, an ML algorithm could help deduce what went wrong and prevent it from going wrong again in the future.
With system usability being one of the biggest things holding UC back, an AI-enhanced UC system will allow new features like meeting setup, recordings, and transcriptions to be automated. In fact, a direct correlation between those who adopt AI into their UC systems and UC success has been found in Nemertes’ 2018 Unified Communications and Collaboration report. The ability to improve interaction with data is pushing massive organizations to find efficiency and effectiveness improved with collaboration.
Going Beyond Machine Learning in UC
As hinted at above, achieving AI does not require machine learning in UC. In fact, there are many subfields of AI like neural networks, deep learning, and natural language processing. However the symbiotic relationship between AI and ML lends itself to being confused with a false interchangeability. It’s important to remember both occupy different, albeit complementary, functions. Next time an employee experiences a bad conference call, IT can look to ML to identify the problem and propose a solution, all with the help of an artificially intelligent machine.
For more on AI and machine learning in UC, check out our On Demand webinar covering the growing role of AI in UC from Alan Shen, VP of Consulting, as well as this article in Computer Weekly, coauthored by Shen and two of our closest partners (Arkadin and Jabra).