Continuing on the theme of AI:
I can tell you what the MBAs are saying. Remember, the MBAs know more about network engineering than you, despite your training and experience, because, well, they’re MBAs! The went to Stanford! Or Kellogg! Or San Jose State!
The MBAs are sure you’re going to be replaced by AI. I’ve personally seen comments from “analysts” saying that’s the case. We’re not going to need network engineers anymore! So all that stuff you’re learning about BGP or OSPF or whatever. Don’t learn it! AI will do it for you. Poof! Away with network engineers! After all, they speak in a weird language we didn’t learn in MBA school! (But it’s ok to speak in our own weird language.)
I happen to live in a world called “reality”, a mythical fairy-tale land to B-school professors. In “reality”, life is a bit more complicated than B-school types believe it to be. And those of us who actually have a bit of experience building networks and making them work see AI for what it is: a tool. Could it replace some network engineers? Yes. All of them? Count me a skeptic. But then again, I don’t have an MBA.
I use a note-taking app called Obsidian. This is becoming more important to me the older I get, as my memory seems to be functioning less optimally. I was thinking I needed to see a doctor until I talked to other folks my age who have the same problem. Despite billionaires’ attempts to live forever, the human body has a shelf life and the clock starts ticking the moment you’re born.
Obsidian is great because it stores notes in text files, you write your notes in markdown, and you can generate tags. I’m terrible at filing things, so I hate programs like OneNote which place the burden on the user. You have to create notebooks and tabs, and so forth. With Obsidian, I just add tags where I need them and I can find anything later on.
Anyways, I had taken extensive notes on Hypershield after talking for quite a while with a TME who is working on it. In Obsidian, I work off of a “daily note” which is like a scratch pad for the day, and if something is important enough, I can select it and extract it to a new note. In this case, I already had a note on Hypershield and I wanted to extract the text to this exiting note. I couldn’t find a way to do it.
Enter ChatGPT. Obsidian is highly extensible, and ChatGPT wrote me a Javascript to extract and append to an existing note. Fantastic! I put it into a plug-in, selected the text, gave it the name of the note, and bam! My text vanished, nowhere to be found. Not on the daily note, not on the scratchpad, not anywhere in the directory with the text files. Since the deletion happened from a script, and not a keyboard/menu command, I couldn’t undo. The text was lost.
ChatGPT was quite apologetic about the mixup and offered several ways to recover the text, none of which worked.
Hallucinations and errors from GenAI are nothing new. We’re all aware of the problem, and I pointed it out in another context early on in the GenAI days. But this is a huge problem for the predicted future of networks being totally automated by AI. Despite the name, AI is not intelligent. It is a system which generates language, human or machine, by ingesting a lot of content and trying to predict what the correct output should be. And it can be wildly wrong.
Sure, we can implement checks and balances. It would have been wise to copy my text to the clipboard before running the script. Still, whenever we automate, we risk runway automation, and in environments where the network is absolutely critical, trusting AI is going to be difficult.
I’ve heard “analysts” say that their customers are complaining they cannot find qualified network engineers. In their incredible logic, then they reason, let’s replace them with AI! Brilliant. How about asking why it’s so hard to find network engineers? Is it a great career choice when you are young and reading that network engineers will be replaced by AI?
I’ve written about the “war on expertise” which has been going on in society in general, and in our field in particular. I got into networking (a) because it fascinated me and (b) because it was a very promising career field. When you have an interest and a passion for something, acquiring knowledge in that area becomes an end in and of itself. I remember reading an article about NBA basketball referees, and how, when they travel, they love to challenge each other on obscure rules. It was like this with network engineers. We were always trying to one-up each other, learn a new protocol or knob, figure out the mysteries of multicast, or new ways to design networks. When I go to Cisco Live (and for the second year in a row I will not be going to CL Europe), these are the sort of people who show up. They are hungry to learn, expand their skills, and up their game.
Now you go to a trade show and all you hear about is how you will be replaced by AI. Why the hell am I going to devote time to mastering a field if I’ll be obsolete in a couple years?
I’m worried for the future because we’re headed down a concerning path. Nobody will be entering network engineering anymore, and AI won’t be able to replace the humans who operate networks. Then what?
It would be nice to see networking vendors (and I mean all of them) encouraging people to learn networking, showing people this is a viable and lucrative career path, and that AI is a tool that they might use to learn networking and augment their capabilities. And it would be nice to see MBAs stick to finance. Or can we at least outsource the job of the MBAs to AI? Yeah, that’d be nice.