Rich Tehrani CEO of TMCnet challenged me to write on real business use cases for artificial intelligence (AI). While the term AI has been around since the 1930s led by Alan Turing who created the Turing Test aka Imitation Game considered to the benchmark for AI determining whether something was artificial or human.
Today, everyone uses the term AI in everything from toasters to clothes like it was a panacea or answer to any technological innovation. Personally, I have worked on AI since the mid-1980s and find there is still a lot of hype but not a lot of real results. Yes, there is a lot of research and development with many so-called AI products, but they aren’t really intelligent at all, they are just fast analysis of data.
Much of AI today comes in the form of machine learning where machines gather vast amounts of data using mathematical formulas called algorithms to determine results. This is not AI, just more hype. I prefer the term machine intelligence rather than artificial. In addition, from decades of research and analysis there are, in my humble opinion, three key forms of machine systems that would be considered as critical to any AIQ test. They are vertical, lateral (horizontal) and oblique forms of algorithms that can be applied to software-defined networks (SDN or SD-WAN software-defined wide area network) and other technologies. In that light, here are some of the ways that machine learning can be used for SDN.
First, data gathering is what machines do best. Gather all the “big data” together like would be found in a SDN network, then apply algorithms aka “rules of thumb” or formulas to the data and see what you find. The data from SDN network traffic, user logins, and security attacks would be analyzed by an AI system.
Second, bad algorithms are like bad or old data that generally lead to bad results and bad outcomes. An algorithm is like a cookbook recipe which there are thousands. By building, testing and testing again the algorithm will lead to better results not just bad or “dirty data” results. The key point is a great algorithm applied against bad or biased data will only make the outcome worse.
Third, networks evolve and with new devices, services, applications and new forms of security attacks. This means that like with any new kind of human thinking about finding a machine or truly intelligent solution may bring real insights to solving real-life problems. Then apply algorithm models built with guesses as to their outcome as no algorithm is perfect and then build models that evolve the algorithm to react, predict and fix network issues.
Finally, here are just some of the benefits of machine learning algorithms on Software-Defined Networks (SDN-SD-WAN) for communications technology users and providers:
- Improved QoS – benefit from on-demand and dynamic traffic performance
- Agility – on-demand network shaping, orchestration and provisioning
- Root management – enhanced component diagnostics and troubleshooting
- Disaster Recovery – provide resiliency and auto-reconfiguration for man-made crises and natural disasters
- Security – no absolute security but faster response to and management of security crisis/mitigation
- Technology – new approaches to chatbots, voice digitization, contact center call routing and more
- Device management – migration to server and network virtualization accelerates BYOD provisioning
- Manageable first and then provide scalable growth for customer applications
- Cost (fewer boxes and lower-layer interfaces) with staff consolidation reducing duplications and inefficiencies
- Improved traffic and capacity forecasting and physical plant geo-forecasting
- Provider – new ways to manage platform API integration for carrier/network operator integration
- Strategic – design, test and build new simulations and data visualization systems and change algorithm models for evolving business demands
Summary – While the focus of this article is AI for SDN-SD-WAN communications technology, there are also other significant AI applications for weather, military, education, global issues, industrial, healthcare, finance and others with significant investments in each underway.
By taking a broader view of what AI is more than just a series of algorithms applied to a particular task, new and better outcomes can be realized and greater integration of AI into problem solving. Lastly, if you ask why we need quantum computing the answer is that the larger the dataset, the more complex the algorithm is needed, requiring greater machine processing for testing and re-testing the results. More details on this concept will be presented exclusively in TMCnet Selling UP Market UP Margin Technology to SMB & Enterprise Certification class at ITEXPO / MSP Expo part of the #TECHSUPERSHOW. Register for this pre-conference now.