Whether administrators are ready or not, advances in AI and machine learning are beginning to seep into server management tool sets and permeate throughout the data center.
Microsoft artificial intelligence and machine learning efforts dovetail with the need for automation in the enterprise with technologies such as PowerShell and Kubernetes. Some administrators have thousands of systems to control and a flood of valuable data available from those machines. Machine learning algorithms can tap into that information to help stave off problems with critical workloads.
While a fully self-governing data center might never arrive, IT pros are seeing a glimmer of the future with some recent Microsoft releases. For example, System Insights in Windows Server 2019 uses a machine learning model to predict if a Windows Server instance might hit a capacity limit, then issues a warning to the IT staff so they can take action or fire off a script to make the necessary adjustments.
Jason Wynn, a technical architect for U.K.-based Computacenter, sees a lot of promise in this computer-assisted future. Wynn develops AI pieces to make the bots used in Teams. He spoke with SearchWindowsServer to discuss the future of Microsoft artificial intelligence and machine learning, how it might affect systems administrators, and what the evolution of this technology might look like.
Editor's note: This interview has been edited for brevity and clarity
What is the difference between AI and machine learning?
Wynn: AI can be intelligent or dumb. For example, Alexa is very much dumb. You ask it a question and then it queries what the answer should be and gives you a very dumb answer back.
If you look at the web services behind it, that machine learning is quite intelligent. We can take the analytics of what you're saying, how you're saying it and the speech recognition aspects and we'll make that more intelligent machine learning experience. The very dumb AI piece in the front is just a channel to the actual machine learning going on in the background.
What is the current state of AI and machine learning, and what are some examples in system administration?
Wynn: If you look at it like the doomsday clock, [then] noon is where we start, and we're going to end up back at noon again. We're probably somewhere about three [o'clock], where we're still very much young in how we're doing things, what we're learning, and the intelligence behind it is very much new. As the processing that we have gets bigger and with the onset of things that Microsoft is working on, like quantum computing, we'll see that being expedited quite quickly.
MyAnalytics is a piece within Office 365 that says how often Jason uses his PC to access email [and] how many conference calls was he on. It tries to help me be more efficient and understand what I was using and why I was using these things so I could drive efficiencies in my working pattern.
[In Microsoft Teams], you create channels to containerize thoughts and workspaces. As far as the AI pieces, we have applications within Teams that allow us to start pulling those together.
One of the original ones was called Chatbot. It gave us the ability to [identify who will work on a particular task]. We have something called AtBot, which ties into human resourcing things, and AttendanceBot, so who's doing what. There're loads of bots and loads of things that people are creating within it.
Are there any tasks that administrators do that AI and machine learning will actually replace? Is there anything AI and machine learning won't replace?
Wynn: If we look at purely the tooling that we use and the scripting that we do, [AI and machine learning will] be much more tied into what we're already doing. Microsoft offers a service called FastTrack where they do some migration tasks for sysadmins. That's very much AI-driven within Microsoft.
If you go into your Windows Server in the future and you tell it what you want it to be -- say you wanted to do certification so you wanted to do [public key infrastructure] or be a certificate authority -- it would go and harden that box to Microsoft standards to be what you want it to be.
We still have to have that design piece done. People still have to pick up a pen and paper or a whiteboard and draw it out and explain how it's going to work and tell people how things are going to pull together. That will always be there.
Likewise, any integration, particularly adoption and understanding [the new technology], will always be there, too. I can create the most amazing bot, but if nobody knows how to use it, no one will use it. You try to explain to them what's relevant to them, how it will save them time, how their life is going to be enriched at the end of it rather than a robot stealing my job.
How would system administrator skills and the role itself need to change as AI and machine learning continue to grow?
Wynn: It already falls into two camps. IT pros who do the server-side infrastructure still have to have the nuts and bolts and the processing and the hard drives and everything else that goes into it to run those bots. You still have that whether that's physically within your data center or located within a Microsoft data center somewhere in Azure or Amazon AWS. Somebody still has to feed and water that environment. Somebody still has to configure it, manage it and design it.
The other side outside of the IT pro is the dev piece. Microsoft in the last two to three years put a serious amount of focus on this. It's not just Microsoft; a lot of people have a rapid development side of things. How do we make sure that we have the ability to create those applications that people need?
What is the future of machine learning, and how quickly will these changes happen?
Wynn: One of the things we joked about on the podcast was The Jetsons. You'd come home and their little robot is vacuuming their rug and she's wearing a doily on her head. I don't think we'll ever have anything quite like that.
In the future, what we will have is integration into the tooling that we already use -- whether it's Windows Server or Azure platforms -- and we start ... enhancing how you work and bringing a more efficient and secure way of doing things for an organization.
We can already see intelligence behind AI, like Microsoft using Graph to tell us who's working on what and what's relevant to you. The clever thing in the future is when you start working on stuff and the AI actually realizes what you're working on and analyzes things across your organization and says, 'Someone's already worked on this. Do you want to team up with that document to make sure that there's not duplication of effort?'
How IT progression works, a lot of these things are interim steps where we take what we have and we build upon it, and then we're at our next version.
I almost liken it to what we see with phones. If you pick up a brand new iPhone XS and compare it to the previous models, we could see that the glass has changed, but the underlying IO that happens behind it and even the interface all remains the same. We're seeing interim steps where I think the likelihood of us having some sort of eureka moment and having a huge jump in what we see available to us is probably less likely.
What we'll see is just everything continuing to improve step by step, and the last piece is being built upon the previous version until we reach a point where I don't know what that place looks like, but hopefully it makes everybody's life a little bit easier.