The release of ChatGPT last year really launched the idea of generative Artificial Intelligence (AI) and large language models into the mainstream media. Artificial Neural Networks (ANN), Machine Learning (ML), and Large Language Models (LLM) are not new in the field of AI - I wrote my Bachelors dissertation on an application of ANNs and supervised ML back in the mid 90s - but this launch really demonstrated how far the field of research had come to people outside of it. Because of this, right now there is a lot of focus around these tools and their use in everything from Music to Fraud! Generative AI in particular seems to have captured the zeitgeist but there are many other uses for the foundational elements of such systems that have been in use for quite some time.
In my field of networking, ML - supervised and unsupervised - has been in use for some time, most commonly in security products. However, for a number of years now there have been a few options in the networking space to use AI/ML commonly referred to as AIOps. These solutions usually involve using the AI components to correlate events and provide the operator that leg up in resolving or identifying issues. But there is one solution that for several years now has been going a lot further, using an LLM to provide that operator interface so that they can ask diverse natural language questions of the platform in order to resolve issues. Furthermore the platform can use its advanced ML capabilities to automatically resolve incidents as they occur or again in response to natural language instructions.
That platform is Juniper Networks’ Mist AI, and it’s Marvis Virtual Network Assistant. In many ways, Mist AI and Marvis function similarly to ChatGPT. Mist AI's Marvis uses natural language queries to interpret and diagnose the network issues in real-time, allowing operators to generate quick and accurate remediation steps. What both of these AI-powered systems have in common is that they allow operators to interact with them in a conversational manner which greatly simplifies the troubleshooting process.
Imagine simply being able to check your ticket queue and being able to just ask the system “Why can’t Bob logon this morning?” and being given an answer or being able to ask “please troubleshoot the accounts switch” and having it troubleshoot the switch. Take that further and imagine the AI simply detecting a configuration problem and correcting it for you, and then raising a ticket to say what had happened and the action taken. This has always been the promise and power of automation but the addition of the LLM really enhances the simplicity and ease of operations. This is the key to achieving operational simplicity and efficiency and one of the true powers of AI that has been highlighted recently, Marvis can free network administrators from the dealing with the time consuming, but simple, problems and instead focus on what they need to do to develop the company’s networks in order to support business growth.
But how does Marvis bring GPT like abilities to the network environment?
Well GPT stands for "Generative Pre-trained Transformer," a deep learning model that's been trained on a wide range of text data to generate new text. Marvis matches that GPT responsiveness and uses it within the network environment and is trained with vast amounts of network telemetry data, so it can recognize and generate automated remediation steps based on conversational queries. This capability has been with us in the Wifi world for over 5 years now, long before the headline splash made by ChatGPT. Couple this with a microservices architecture that allows the central platform to be updated without disrupting service and you have a winning formula that has led the Gartner Magic Quadrant for Wired and Wireless network the last three years. And, speaking as a lifelong science fiction fan, if they were to add a speech recognition module we could simply talk to the computer and ask it to identify and solve the problems in true Star Trek fashion!
So whilst companies are rapidly linking their systems to GPT, Juniper have had to go to very little effort to do so with Mist AI and Marvis because they have already been doing this for several years! The concept of the AI-Driven Enterprise is something they have championed for the last 4 years as Marvis was introduced all the way back in 2018! In fact, Juniper have Integrated ChatGPT into Marvis allowing it to reach out and search documentation and other data to assist in troubleshooting more complex problems.
In summary, Juniper's Mist AI, with it’s Marvis virtual network assistant, is an AI-driven service that can simplify network management and operations, just as ChatGPT does for conversational interaction. Marvis brings GPT abilities to the network environment, allowing that same conversational interface for operators to quickly diagnose and remediate network issues. By bringing these capabilities to the network environment Mist can be a foundational building block in the journey towards becoming a fully AI-Driven Enterprise.
To learn more about why Juniper Networks is a leader in AI for IT: visit the Juniper Shop