> Sysadmins are being left out of AI implementation
Expected to fix it anyway.
AI may be coming to help sysadmins, with the promise of alleviating repetitive tasks like log analysis, resource monitoring, vulnerability prioritization, and patch management, but few of them are put in charge of its implementation. According to a study by patch management platform vendor Action1, 82 percent of sysadmins said …
The "problem" with AI (LLMs et al) is that strategically it's got sod all to do with logs'n'all and everything to do with precision targeting of sales and is/will be driven by brand and marketing. IT just provides the plumbing. I'm not having a pop, it's just that IT is not commercial in its focus. You start spouting phrases like "sentiment analysis" at IT admin types and they think you're taking the piss.
> You start spouting phrases like "sentiment analysis" at IT admin types and they think you're taking the piss.
Why do I get the impression that you also don't see password management as being something that requires empathy? Although I suspect it may occasionally inspire a bit of creativity, eh!
《spouting phrases like "sentiment analysis" at IT admin types and they think you're taking the piss.》
More likely to start actually start pissing on the spouter with "analyse that sentiment" while reaching for baseball (or cricket) bat and rendering the coup de grace.
Then assisted out the window for the vertical trip to the car park.
Today, July 28, is System Administrator Appreciation Day, when...
when we inform our appreciated System Administrators that they will all be replaced by AI and are ordered to provide the learning materials for said AI and provide feedback on the said AI's responses for the remainder of their employment. Any under performers, deviations of expected work or lack of accuracy will translate in cut severance pay. Anyone showing bad performance may be sued for loss of profitability margins of their replacements.
"AI holds immense potential to free sysadmins from routine work and enhance their overall productivity"
I don't think these people understand the nature of A.I. First of all it isn't intelligent, it's trained on a large set of data and uses a predictive engine to generate output. Sometimes phrasing the same question differently generated different and contradictory responses. Going on current versions of ChatAIs these things don't have any common sense and will happily do the wrong thing if fed false data. May be useful to run what-ifs, but not to have direct control of the system.
Mentour Now! does interesting videos on what can happen when the computers on-board an aircraft disagree with the pilot. One such incident, the pilot put the landing gear down and set the auto-pilot to landing-mode. At an air-show so as he could do a fly-pass for the audience. So the plane duly flew past and did a landing in the forest next door. I'll leave the final word to ChatGPT.
ChatGPT: “The reason for giving different answers when a question is phrased differently lies in how language models like GPT-3.5 process and generate text. These models are based on probabilistic methods, which means they assign probabilities to different sequences of words based on the context they have learned during training. Consequently, minor changes in the input phrasing can lead to different probabilities and, thus, different answers.”
Of course they don't understand the nature of what is currently called AI. They don't understand Exchange either, but everyone has it so it must be had. Cloud ? Ditto.
The difference is that, with Cloud, they're starting to recognize the cold, hard reality of the bills and that it's hardly the promised land Marketing swore it was.
AI will be the same thing, except here there will be financial disasters that will occur and the bills will be as, if not more, important as Cloud.
At that point, after having costs a small fortune, AI will exit stage left and sanity will return.
It's just the beancounters that will be crying tears of blood in the meantime.
What "system admins" view as AI vs what "management" views as AI are different things. In most cases I suspect "AI" to be implemented in some kind of service or basic software that end users can understand, and that software will have proper names not generic "AI" (but "management" will more likely view said solution as "AI"). vs when a "system admin" may be asked about AI they will probably think the topic is much more in depth like setting up a LLM from scratch or something which is probably going to be a super rare occurrence more fit to a specialized set of skills.
I've been working with MySQL for 20 years and can set up the basics just fine but there's a whole different level of how you can set it up/understand how it's working beyond what I know for example, and I leave that to the DBAs(I can then take&duplicate the config easily enough, doesn't mean I fully grasp what the settings mean). Same for Oracle, I can setup an Oracle DB server, but if it's for a serious application I certainly won't volunteer myself to do that from scratch.
If it's a service providing the "AI", then likely the only stuff system admins may need to bother with is perhaps managing accounts, or other generic things that apply to any kind of service not specific to AI.
Everyone always forgets about us :(
Thank you for the link.
The section titles alone pretty much say it all.
Personally I would prefer the job sapping Mordor with a screwdriver.
By comparison Sauron was clearly more reasonable and rational than some of those I have encountered.
No way something could go wrong here.
So as a sysadmin training my replacement -
If I put something in the training set, say a “back door” then all systems built using that training set could be owned by me?
Just doing a BOFH here.
What does AI have to do with log analysis? Any half decent logs will be structured enough to be analyzed - with software, yes - without any need for AI, LLM, or whetever. The manglement and marketing bods who push AI for the purpose don't seem to say anything but "the logs are lousy and really all over the place, but rather than (invest a moderate effort in|push clueless vendors towards) improving them we should better deploy a hugely expensive and not very good or precise AI to try to figure them out".
[Disclaimer: I have developed (3rd party) log analysis SW a few times in my career, but only for internal use and generating detailed reports for customers, so I am not all that much of an expert.]
This post has been deleted by its author
"82 percent of sysadmins said their employers did not require AI implementation in their roles"
Is that because someone else is implementing it, or because "AI" in its current form is nearly useless to the company, so nobody at that employer is implementing it?
I saw a Reg article a while back saying something about companies missing the "4th wave" of AI. Eh, probably not; likely they didn't want the first 3 either, so of course they didn't "miss" having this one...