Microslop official documentation on how to ground an AI
1mon 3d ago by lemmy.world/u/TheKrunkedJuan in programmer_humor@programming.dev
Yes, this is official documentation from the MS Learn course for the Agentic AI Solution Architect certificate
Microslop SharePhont™
who morged that? Why does that happen continvously?
What a combigent, inbiltum response!
It's AI generated
Thank you for joining the class
TBH the joke isn't annotated as such and there are always people stupid or innocent enough.
It's a reference to another Microsoft classic: https://www.windowscentral.com/software-apps/microsoft-caught-plagiarizing-graphics-with-ai-slop-microsoft-continvoucly-morged-my-diagram-there-for-sure
every time we see AI mix up *U* and *V* we just imagine it not as a machine, but instead as a european renaissance-era writer who accidentally time traveled into the current year and promptly got locked up in a cubicle with a real-time brain scanner wired up that interfaces with chatbot programs seen online
for extra hallucinations simply assume the consumption of one (1) early 1900s fire extinguisher's worth of carbon tetrafluoride as part of your scribe's daily fluid intake
Is that why tech bros keep trying to bring back phrenology?
Yes, that’s the point
Anyone remember how much nicer the world felt before we all realized nobody in charge of anything with any real power has any idea what the fuck they're doing?
We didn't all realize that at the same point. Some of us have known that for a very long time.
I was twelve watching Jon Stewart. It was quite the epiphany
They know what they are doing. Their goals just don't align with the rest of humanity.
The mistake we make is in assuming that government choices are born of incompetence... The actions of empire are not incompetent. They are not intended to serve you or the citizens, but to serve material interests, the plunder of land, natural resources, markets, and cheap labor".
-Michael Parenti
RIP to the GOAT
Edit: He talks about it in the lecture starting around 15:35. Though the quote above is from his book "against empire"
Well back then we thought everyone in charge was evil. Back then we had Krutschev, Nixon etc. Always nuclear war looming. And Reagan with this trickle downs, the Bush wars over made-up WMDs etc.
I don't think there was ever really a positive period. Maybe the 90s which was kinda the high point. Economics were good, America and NATO was the undisputed world power, nobody took global warming seriously yet and the internet was promising a great future.
Then after 2000 we had the dot com crash, 9/11, the resulting wars, Russia becoming an enemy again, the financial world crisis, the rise of the internet as a surveillance tool, global warming, the pandemic, exploding house prices everywhere.
Really pretty much exactly after the change of the century everything turned to shit.
Economics were good if you lived in a western country. Former Soviet Block, Yugoslavia, and global South countries were having a rough go of thing in the 90s
They're still having a rough go of things, and they were having a rough go of things before that as well, so that's not really a valid counterpoint.
If they were having a rough go of things in the 90s. Then the 90s weren't the shining beacon of success it's claimed to be. That point is very much valid.
Then there as never been in the history of this earth a shining beacon of success because at every waking moment there wasn't an optimal number of people immune to all disease, death, and suffering, to no downside to nature around them?
The 90s is where the curve peaked for a lot of people. China had finished bringing 70% of their population out of extreme poverty since the world war era, the USA markets and deficit were stable though heading in a bad direction due to Reagan, and Europe was having a big cultural and tourism boom.
Then there as never been in the history of this earth a shining beacon of success
That was kind of my point.
It's a shit point that argues against 100% valid and factual statements.
Tell that to the billions of people who've been living in poverty for decades under the U.S . Imperialist boot.
I fucking will, bring them here and I'll tell them the peak was in the 90s.
Yeah but the future was bright back then.
Source: I gotta wear shades.
No that was a long time ago for some of us it's always been like this you were just insulated by privilege
Thanks for notifying, I hate to know it's true (no insult intended)
Nope.
I'm legitimately having a blast right now, 100% affirmed by reality, which I spent the last roughly 15 years of trying to tell people, very labriously, with very detailed reasoning, why what is now obvious was then and basically always has been the case.
Your discomfort is my schadenfreude.
I'm sure that back in the day there were adults in charge of the big companies. I know that where I work management used to listen to tech areas for tech stuff. Now they just trust whatever the Microsoft agent tells them
It would be really funny if it wasn't so sad.
It almost makes me think a human worker intentionally made these slides. LLMs don't really have the creativity to make typos, which is how I can sometimes catch LLM comments on here. Also "SharePhont" is pretty funny
Unless they used an image generator to make the slides, which would be extra stupid
edit: turns out it WAS the extra stupid
Oh it's definitely extra stupid. Looks like the slide was made whole-cloth by an image generator, not like the text was generated by an LLM and pasted in.
Image generators make typos all the time (if you can call them that).
See also: continvoucly morged
Image generators make typos all the time (if you can call them that).
Primitive image generators don't "understand" semantics of the patterns they assemble. The obvious example we're probably all familiar with is "AI hands" with weird poses, excessive fingers, sometimes even extra arms etc.: the AI knows the pattern we understand as a finger, it knows the correlation with those we see as hands and arms, but it doesn't how what a hand is or why it's important to have a specific number of those patterns combined in specific ways.
Text is the same: The model knows the graphic patterns of letters, maybe knows the patterns of words the letter-patterns often occur in, but the same randomisation that can produce different enough results to look human can also lead to randomly generating patterns that it fundamentally can not know aren't actually valid words.
There are solutions, such as using a more specialised tool like a combined layout generator and text generator with feedback to make the text fit the layout. Using the right tool for the task at hand, paired with supervision by a human that knows its shortcomings and can check whether and where it trips up, might do a better job.
But that human has to have the required know-how, and if you really want to use a single LLM to feed all prompts into, that model should be capable of detecting and delegating the work to those specialised tools (checking with the prompter to confirm that its detection is accurate).
A simple all-purpose-model and general "prompt engineer" without subject-specific experience and training just won't cut it. The marketing for these tools generally seems terribly intransparent about that problem, and executives generally seem to be oblivious to it (or just indifferent, so long as it helps them cut costs on paper for a few quarters).
(As an aside: it's the same difficulty text generators occasionally have with facts and citations: They can't tell when it is important to have very specific combinations of words that map to very specific occurrences. It might have picked up the correlation of the word pattern "Words (Number), Words (something edition). Words." with human names, year numbers, book titles and publisher respectively, but it can't know why only specific combinations of author, year, existing book title and publisher are permissible.
It may get them right often enough by picking a likely combination from texts related to the prompt, but unless you double check (or provide in advance) that citations refer to existing works and fit the cited content, you run the risk that it randomly generates bullshit. A student "writing" a paper might not be able to catch it, but a professor that knows the major authors and works of their field is probably gonna spot it.)
Yeah. The newer versions of Dall-E, GPT-Image, and similar cloud apps seem to have subnetworks specifically for text. They're worlds better than they were just a year or two ago. Like, you can twist them into generating semi-coherent-looking text logos, or make a cartoon character wearing a T-shirt with some text on it.
I've seen some complex pipelines with open models, where people train loras specifically to fix the things the base models suck at, like hands, text, etc.
But it's still a really dumb idea to generate a whole presentation slide or infographic that way, for a wide variety of reasons. If you ever get decent results, just consider yourself lucky. I mean, even the people with the skills to do this well (who are few and far between) would find it way more trouble than just, you know, making slides the normal way.
The incompetence we keep seeing from Microsoft is staggering. I can only assume this is malicious compliance. I imagine some exec said "everyone needs to use AI for everything" and everyone below them said "okay you dumb fuck, here you go".
I'm dreading the day when our bosses get antsy about AI adoption. I dodged two AI-related projects already. Miss me with that shit. My fuckups are mine to own, fix and learn from, my slop is organic and handmade and my business lingo is authentically rough around the edges.
The frustrating thing is that there actually is a use case I can think of where AI might make my life easier, related to processing and cleaning natural language data, but I don't have the resources to do that at the scale I need it (and probably not the know-how either).
Always bet on stupid
Did AI make these slides?
It combigated them.
A veritable answer.
It's a perfectly cromulent word
Product of continvoucly morged slop.
I'm about to combigate my foot with your ass!
(sorry, just felt right somehow)
Almost made me do a spit take, damn you lol!
Doesn't siund like AI. Sounds more like they fired all the competent, expensive employees and replaced them with AI, than realized AI wasn't doing so well and this was made by the few incompetent, cheap employees that were left.
Data Flow
Data Flow
Data Flow
The Data must Flow
It's the Combigent Directive
Saw this article from Microslop the other day about their MDASH copilot model competing with Mythos for bug finding/fixing.
And here's the image they went with for the flowchart of what the AI does... Complete with overlapping text and the worst layout I've ever seen.
Zero humans proofed this.
I know this! This is a slide deck which only makes sense in an interactive way, where the steps are revealed after one another. Probably straight from a presentation.
Don't attribute bad work to AI which can be done in the same way by an incompetent human.
As someone who's career depended on presentations for a while:
Never build animations in-slide (unless they are purely aesthetic).
Instead:
Build multiple slides which duplicate the previous and add new boxes with transitions on the slides. Bonus: change the title on each build to cover the key point.
Sure, they take longer to edit. But that leave-behind export to PDF always works.
about their MDASH copilot model
that's satire, right? 
"Combigent?" Hm, I never heard that word before I moved to Microsoft.
Microsoft 365 is the only Combigent Certified® cloud platform.
Microsoft 36S, I think you’ll find!
Combigent is too nice a sounding word to lack any real meaning it needs one.
Combigent (adj.)
Definition: A word that seems like a real word but actually isn't
It also looks completely plausible for a word. I can see why an AI trained to make plausible-looking images generated it.
Now I'm curious to see what other perfectly English-looking words it comes up with
How dqre you say such mean things about my computer boyfriend you're just an irrational meanie I bet all the words you used here are fake too!
A classy male driver of a combine?
I guess it'll be nice to have a spare word lying around for when a wordless meaning emerges
Skilled with the use of a comb
Combigent [kəmˈbʌɪdʒ(ə)nt] {adj} A portmanteau of combine and emergent, describing sth. derived from combination, the ability to combine newly discovered facts or bullshit hallucinated by AI. Examples: The LLM gave a combigent response. The manager's presentation was quite combigent.
Combigent is a perfectly cromulent word! Veritably cromulent
Not a single one of those words is cromulent, I swear on my aunt's borschtum.
Seems perfectly cromulent to me
I don't see the probulent here
Data flow
Data flow Data flow
Data flow
The data must flow
It's flowing. . . it's flowing!
Those are perfectly cromulent words.
You missed "Microsoft 36S" and "Intemal Docs"
Yeah those are very weird errors. Not like AI hallucinations. But more like OCR errors.
I wonder if they imported another PowerPoint and then the AI made images out of it to process it. That would be a very inefficient way to do it.
Maybe the AI has had an OCR pass grafted onto its output processing to make it seem more capable of producing text. AI image generation usually makes weird alien looking text, I could imagine someone deciding to add on an automatic cleanup step to try to make it less obviously AI.
Due to the negative press combigent
Clown corporation. If I were employed by Microsoft, I'd probably be embarrassed to admit that in public.
I see, they have been continvoucly morging the docs themselves, as per their developer docs
Veritable answer? Indubitably!
The worst is claiming Copilot "understands intent & meaning", the funny mistakes more or less show that that's just a blatant lie.
They morged it https://nvie.com/posts/15-years-later/
There are also three "Data Flow" labels (two for the same arrow even) and I'm pretty sure it says Intemal Docs, which makes sense for and AI to mess up too because of similar m and rn look.
Combigent is a perfectly cromulent word.
I solemnly swear to god, I posted this exact same thing before I saw your post lol
A noble spirit combiggens the smallest man.
According to Microsoft's own terms and conditions, AI is supposed to be used for entertainment purposes only. They're breaking their own terms and conditions here.
I was entertained by this slide. Laughed my ass off in fact
And yet, aren't they entertaining us with the results?
Did their marketing department read the TOS? because that is not how copilot is advertised.
Did their marketing department read the TOS?
They asked Copilot to summarize it for them.
It's Microslop. The people who had any resemblance of work ethics left a long time ago. No one still working there gives any shits about QA - especially not the management.
Combigent is a perfectly cromulent word.
Understands intent & meaning Understands intent & meaning Understands intent & meaning Understands intent & meaning Understands intent & meaning Understands intent & meaning
You should have a look at some of the official PowerShell scripts they've made to assist with managing, installing, and decommissioning Exchange Servers.
One had the word group misspelled consistently the entire way through it. Another had a wonderful comment noting that an internal function (to manage something with AD permissions or deep internals of the Exchange Server) that looked misspelled was in fact spelled correctly, because the command had a typo in the real name of it.
That sounds like human errors.
Microslop Cope A lot
For years, I’ve rather read Stackoverflow comments than Microsoft’s nearly unusable documentation. It’s as if they don’t know how to write coherently
Then they train their troubleshooting clanker on said unusuable documentation so it's unhelpful but it won't let you talk to a human who might actually be able to assist you.
DOS 5 docko was fine. I wonder* when they went wrong
*I don't really
Oh, yes. Good question. I can’t speak for those days. I’m largely speaking largely about the Azure related documentation over the past 5 years or so
So it's search engine that costs 10 times the energy.
Of course it costs more, classic search engines lack the combigency and veritability Copilot provides.
And gives you the wrong information 30% of the time.
It might be combigent, but at least it's veritable
TBF 'combigent' should be a word. Not sure I can think of a unique case that justifies it, but it sounds like it could easily be a word.
It sounds like something that would ease the combining of ingredients/components. Like an emulsifier but more general. "An emulsifier is a combigent for oil and water"
Your mom is combigent.
A "combingent" is when a guy is going commando, but when using the bathroom, is in too much of a hurry and gets his scrote skin caught in the zip
Nah, that's a 'nadscrag'
its like how some people say pop, and others say soda and then others call all soda and pop cola, and then others that call soda and pop soda pop.
Some places call it a combingent nadscrag and that's just wrong
Microslop boomers are the worst
I started using linkwarden to archive their documentation for various things just so I know I'm not crazy when it changes.
You can track the changes on its github https://github.com/microsoftdocs
Oh my Lord, that is so much easier and in hindsight, so obvious. Thank you brother!
This is not AI, AI would have made it much better. I mean, the only thing LLMs are quite good at is writing words, it is uncommon they misspell a word.
LLMs misspell words in generated images. Modern models aren’t as bad as they were a year or two ago, but we don’t know what model was used here.
Such a combigent, veritable flowchart
"I published the Intemal Docs because I thought it was a project name. I'm sorry. Would you like some best practices on damage mitigation?"
Humanslop
I thought it wasn't in English at first.
Why are you booing them? They're right.
Yeah, never seen poor grammar or spelling in content written by Americans before 🙄
Not that bad
AI doesn't misspell words. It's more like hypercorrect which reveals it.
Unless MS uses some very outdated models of course.
Unless the whole thing is a generated image, then it can certainly misspell words.
Rephrasing: I've used AI a lot and I haven't seen it misspell anything in at least a year. Even in pictures these days. Depends on the model, of course.
Claude is quite often making spelling mistakes
You forgot the /s
Spelling and grammar is not valued in the engineering disciplines, especially when there is a sizable population of the workforce for which English is a second language.
Edit: only on Lemmy do you get downvoted for saying true things based on your real experience in the world.
I couldn't be a software engineer without knowing English spelling. So you're talking bullshit.
FYI, all programming languages are based on English language and use English words.
Even edited your comment to be confidently incorrect, LOL
I am in fact confidently correct. Prior to working in engineering, I worked in business development. The difference is stark. When you’re writing a proposal that is supposed to win your organization contracts, you have to be perfect in your grammar and spelling. In engineering, it’s less critical and no one seems to care how much you butcher your writing and spelling. Sure, people know some english, but then you read their code and there are glaring misspellings. The best is when they misspell a variable name and then it gets used all over the code base. Internal design documents are also rife with grammar mistakes and misspellings too. In my experience, AI is actually better at spelling than human software engineers. So, I don’t buy this post at all.
Your ass would have been downvoted on any platform for saying spelling and grammar is not valued in engineering disciplines. Some of them would be downvoting you for saying a discipline built on the foundations of math and science does not value accurate and correct documentation, and the others would be downvoting you for calling whatever the fuck microslop is doing an "engineering discipline".