I have been at Cisco many years and seen a few transformational events affect network engineers. After I first joined Cisco in 1998, IP telephony came on the scene and disrupted not only classical circuit-switched communications but those of us in networking as well.
More recently, automation, programmability, and software grabbed everyone’s attention, promising to upend the way we think about, design, and operate networks.
Today, you can’t go 10 feet (3 meters for those of you outside the US) without someone talking about AI. CES 2024 has just wrapped up, and Artificial Intelligence (AI) was everywhere.
Whereas IP telephony and automation may have been reserved to specific places in IT, people from all walks of life are discussing and using AI. The networking industry specifically is shifting rapidly around technology and talent acquisition as well as product integration and focus. All this change and hype can leave people confused and wondering how they can wrap their head around what AI is and how to best make use of it.
Before going too much further, it’s important to mention that AI is not just one thing. It also isn’t new. AI is a class of machine learning that uses math to analyze large sets of data to make predictions, provide automatic classification, and summarize large data sets.
In fact, a few years ago, when this notion of machine learning (and AI) started to first emerge in networking, I took an online course that focused heavily on the math behind machine learning to do things like cluster various test results to identify anomalies and analyze a large set of number images, providing new images so AI could correctly identify the number it receives. This same kind of predictive and classification AI is already baked into several Cisco products. Have you ever been on a Webex with someone mowing outside, yet no one on the call can hear anything? That adaptive noise canceling is a form of AI.
Capturing all the attention today, however, is generative AI, a new use case that, as its name suggests, is about generating new content by analyzing an array of existing content. This generative AI receives a prompt that specifies what type of content to create and in what fashion, and it uses its underlying model to satisfy the request. Different generative AI tools exist for different media and different use cases. Here are some ways of using it:
- Ask various generative AI systems to write a short story in the style of Ernest Hemingway.
- Create a picture of you shaking Albert Einstein’s hand.
- Convert a video you shot on your phone into a silent movie.
Because of the viability and availability of generative AI, the buzz surrounding it has grown to an all-out symphony. People have been trying out different use cases for it as well as performing prompt engineering to fine-tune the results. As with all powerful tools though, we must take care. ChatGPT, for example, warns you that it can make mistakes.
Network engineering is an interesting use case for generative AI. You can ask a tool, such as ChatGPT, to build a simple OSPF configuration for a Cisco router, and it will not only generate a sample config, but it will also explain part of it. The initial config it generates will have placeholders in it, but you can prompt ChatGPT to then regenerate the config with sample values in the placeholders. This makes for a nice way to bootstrap a network testbed or help crystalize new networking technologies by way of example.
How can you learn more about AI, its use cases, and its caveats? Cisco U. has you covered. Cisco U. All-Access users can explore courses and learning paths on the foundations of generative AI, ethical and privacy concerns with generative AI, and others.
For those looking to learn more about how generative AI can shape network engineering and operations, we just released two tutorials included in Cisco U. Free, Create a ChatGPT Client with Python and Chatting with Cisco IOS XE. These are great if you want to learn how to build your own integrations with generative AI.
Remember I said that Cisco has been folding AI into our products for a while now? You can start to get value from AI today in your networks using a variety of Cisco solutions.
Consider your WiFi network. I’m sure it shares characteristics similar to all WiFi networks, but I’m also sure it has unique aspects of how and where to deploy it. Instead of using static, manually tuned thresholds, you can use predictive AI in Catalyst Center Assurance to provide more accurate alerts when problems occur. In the image below, the green band is the predicted baseline, calculated over an extended time period, whereas the blue line is the actual value. The red bands signal anomalies, which will generate alerts.
Cisco U. has content you can use to learn more about Catalyst Center Assurance as well as ThousandEyes Internet Insights and Meraki Insight. On January 22, 2024, we announced a new CCNP concentration, Enterprise Network Assurance (ENNA) that covers skills you need to make use of these AI-enabled capabilities. The learning coming this spring for ENNA will delve deeper into these analytics and assurance solutions.
I started out talking about how, if you’re in this industry long enough, you’ll see your share of disruptive technologies. A past fear associated with automation was it would make network engineering obsolete. Instead, it’s made network engineering more scalable and exciting. Likewise, AI is a tool that helps push those excitement and capability boundaries farther. Now is your time to learn, experiment, and discuss.
When you’re ready, post in the comments what you want most from AI and networking.
Follow Cisco Learning & Certifications
Use #CiscoU and #CiscoCert to join the conversation.