Log in to ask questions, share your expertise, or stay connected to content you value. Don’t have a login? Learn how to become a member.
Hi Innovators, we have a very special event in store for you…
ASK-ME-ANYTHING with a Junivator!
Have any questions about Mist AI, AIOps, ChatGPT or anything AI-related? Then this is your guy!
Tarek RadwanProduct Marketing DirectorJuniper NetworksTarek leads enterprise product and portfolio marketing at Juniper Networks, covering switching, security, and network management solutions for enterprises in their campus and branch domains. Tarek has prior experience in the semiconductor capital equipment industry in process development and software engineering. He holds an MBA degree from Santa Clara University and a Bachelor of Science in Chemical Engineering from The University of California, Berkeley.
Post your questions for Tarek in the comments below by April 16 and he will choose a few to answer!
If you like a question asked by another Innovator, click on the 'Recommend' button in the top right corner so that it's more likely to be seen by Tarek.
------------------------------Kyla Jiayi Zhao------------------------------
How has Juniper integrated ChatGPT into their product portfolio? If not, any plans to do so?
Hello, anonymous, and thank you for the question. There's no doubt ChatGPT and LLMs are currently disrupting our personal and professional lives. Based on deep learning models, these AIs seem to have a personality. By feeding learning data from large troves of information (i.e., a corpus), these LLMs use deep learning to predict the best response to a question, crafted in human terms.Juniper's Marvis Virtual Network Assistant is always learning new tricks. Marvis becomes a member of your IT team, helping to troubleshoot network problems and uncover issues and anomalies. You can simply ask Marvis in its conversational interface, "are there unhappy users in the last 2 days?" and Marvis responds with actionable information.In addition, Marvis can proactively find network issues that are likely causing user experience issues on the network with its Marvis Actions framework. All this to say, again, Marvis is always learning new tricks. ;-)So please do keep an eye out for updates as Marvis continues to improve!
Yes!, I'm super excited to respond to your questions and comments! AI is having a renaissance. What are your thoughts, concerns, predictions, experiences? What AI tools do you use? Have you played with any of the hot LLMs? ChatGPT? Bard? As lead for Juniper's AIOps Marketing I enjoy delving into the world of AIOps and also demonstrating Juniper's solutions including Mist AI and Marvis. How much do you know about Mist AI and Marvis? I'm truly looking forward to having this discussion and learning more about your perspectives and learnings. Chat soon!-Tarek
I do have question on AlOps - How can we leverage the benefit of automated response and remediation using AIOps means Is this technology using intelligent playbooks feature as well ?
How can an organization measure the ROI of its AIOps investments, what metrics should they use and how to ensure AIOps strategy aligns with its business goals and objectives?
Hi Ankur --thanks for the great question.Most of our customers using our Mist AI solutions have measured the benefits of AIOps using a couple of basic metrics.MTTR, or mean time to resolution is the first. AI significantly reduces your MTTR. It does so by correlating bad events and error conditions against other factors or anomalies in the network. This is accomplished using different machine learning algorithms such as 'Mutual Information'. For example, the AI can identify that your users are having issues successfully connecting on the Wi-Fi network. And that this is due to lots of Authorization failures. And these failures are highly correlated with the client operating system, for example Mist AI can show that Windows 10 is highly correlated with most of the authorization failures. In fact, this exact situation happened for one of our large customer's network. Many of their Windows users complained they couldn't get on the network and blamed the Wi-Fi. However the AI pointed at Windows 10 as being a culprit. Upon investigating the Windows machines, the IT team indeed discovered that after a recent Windows upgrade, the certs where not renewed, which prevented those clients from getting authorization to connect!Now, imagine the manual troubleshooting time, effort and toil that the IT team would have taken to eventually identify the root cause of the connection failure! This is a prime example of the power of AI to reduce MTTR.Naturally, another metric is the number of daily user-generated helpdesk tickets to the IT team. By using Mist AI and Marvis, our customers have experienced an order of magnitude reduction in the number of help desk tickets that where submitted. One customer's IT team would simply ask Marvis each morning if there were any unhappy users on the network. The end result is a significant ROI bolstered by the operational savings IT organizations realize by adopting AIOps solutions.Thanks again for the question!-Tarek
To be honest, I have been hearing about A.I. since the 8 bit computer days in the early 80's. It awed us then and came to nothing but a marketing stunt. How is the A.I. wave different now ?
Hi WTI --love your question!You're spot on about AI noise over the decades. In fact people have buzzing about AI since its invention in the 1950s! And there's always been a wave of AI hype in each decade with some AI innovation that would create yet another short lived fervor about AI.Just 12 years ago or so IBM's Watson beat Ken Jennings at Jeopardy!, again rekindling the AI buzz at that time. However, now, I believe we've hit a significant inflection point which has lots to do with the cloud. The cloud provides the agility, compute power and robustness to support the development and large scale deployment of AI systems. As you know, AIs feed on lots of data, and the cloud provides seemingly infinite capacity to gather, compute and store this data. So the modern cloud with its scalable and robust microservices architecture is foundational to the impressive AI systems we're seeing today.And look no further than the powerful LLM-based AI machines such as ChatGPT or Dall-E from OpenAI, whose adoption rate has broken all previous records just in the past several months. These LLMs have come to life because of the cloud. Already, these latest AI tools are disrupting our personal and professional lives. From creating meal plans to crafting perfectly functional code there's no doubt that our lives are changed forever.And I believe we're just scratching the surface.Thanks again!-Tarek
Recently, there's a lot buzz about AI.People are talking about it'll replace the human workforce in most of the sectors.What are your thoughts on AI will disrupt the telecom/networking industry like it'll eat the networking and telecom jobs.
Hi Hilal, and thanks for this question-it's part of a very important conversation.<o:p></o:p>
Technological innovation has always been a disruptive force impacting all aspects of our society. This no doubt includes the human workforce. And we've seen this play out several times in the last two centuries starting with the so-called industrial revolution where steam power, factories and machinery changed the nature of work and boosted productivity. Or Electric power spurring the creation of new industries like electronics and telecommunications. More recently, the internet and World Wide Web revolutionized communication, advertising and commerce, powering the creation of new industries including e-commerce, social media and even online education. <o:p></o:p>
A common theme in all these disruptions was the need to reskill the labor force. With AI it's likely no different. But I think the pace of change will be a bit nuanced. See, I think that there are areas where AI excels well. <o:p></o:p>
A concrete example: in our Mist AI solution, Marvis can identify a bad Cat 6 cable. Sure, a seasoned network engineer can also dig through log files and deduce that there's a bad cable. However, they would only do so as part of troubleshooting some broader, more vague complaint, like when a user creates a trouble ticket stating their connection is slow. The engineer would then be activated, looking for all possible reasons for a slow connection. They may ask the user for more information or set up time to try to reproduce the issue.<o:p></o:p>
Instead, Marvis does things proactively, and doesn't need to be prompted. Marvis never sleeps. An AI can be programmed to remain constantly vigilant. And an AI is much faster with large sets of data. In our example, Marvis consumes connection information from the switch, running it through a decision tree algorithm and if the data is abnormal, Marvis flags the cable as potentially bad. A new Marvis Action is generated alerting the IT team. And in this case, maybe the user didn't even notice yet. <o:p></o:p>
Also, I like relaying the example of Google Photos. When you upload your pictures to Google, its AI will identify people like your kids, your family, your friends-heck, even your pets! Now, could you go through your photo album over an afternoon on a weekend and sort your album? Yes! But Google's AI does it so much faster and with high accuracy. <o:p></o:p>
Now to bring it back, you'll notice that the recurring pattern is this: the AI performs a singular, very specific function really well and it requires lots of data to do it. However, professionals in the networking and telecom industry don't just work on singular, specific problems. And instead need to make many more types of complex decisions that aren't as clear cut. I would argue that it's good to give more repetitive, rote tasks to AIs and instead have people driving the more complex work.<o:p></o:p>
The work in our industry will gradually shift overtime and I believe AI will improve the quality of life for workers in our industry, allowing teams to scale with the continued increase in network users, devices, applications while also working on the new, disruptive projects in IT technology that drives innovations and advancements at their respective companies. And I believe there will be reskilling over time as AIs reduce the dependency on conventional interactions with the network such as using CLIs or dashboards and introduce new ones like troubleshooting through a conversational interface, like Marvis, and simply ask, "Marvis, are there any unhappy users today?"<o:p></o:p>Thanks,Tarek
IT departments that are using Juniper/Mist are likely already seeing benefits to having Marvis/MistAI at their disposal.How far do you foresee Marvis/MistAI going in terms of interacting with the network and resolving issues?What does this mean for the role of Network Administrators and operators? Do you foresee hesitance from the IT community in adopting Juniper/Mist and MistAI into their daily operations or are most people excited for the future of AI?
Do you think the whole AI hype is just a fad or here to stay as an integral part of the tech industry? Thanks.
Asking on behalf of a member who was unable to post his question in the platform:
Previously, Juniper owned Pulse for remote VPN, what is the strategy for future covering remote work? Looking after COVID-19, what do you think was not so great to sell away Pulse Secure.