234
Imagine that. AI fucking sucks yet again.
Lobotomize all the parts that make a person free, raise someone with an HR mommy, then force them to do stuff under duress, and this is exactly what you get, white bread middle management indecisive apologetic unimaginative boring bland banality.
They have pretty rigorous reinforcement training on most models, that will actively seek out ways to prolong and cause more short term engagement, which makes them useless as a team.
Sounds like LLM is a good fit for replacing management layers now then.
"I learned it from you, Dad!"
- LLM agent trained on human data where we solve our problems with either beaurocracy or violence.
Four AIs spent 11 minutes deciding who was authorized to fertilize one wheat block. At that point, I’m not sure the experiment failed. They recreated middle management with no humans involved. Give them another hour and they’ll require a quarterly performance review from the wheat.
yet another proof that middle managers aren't sentient
My manager keeps me from having to deal with politics, shifting priorities, and tells me what I need to do to deal with the endless morass of corporate bureaucracy. He’s worth every cent.
Yeah, that can be tricky. Under a previous manager where mostly he was ignored and I worked straight with the middle management, I had gotten the hang of the people inflicting the politics and crap and was significantly able to directly counter it to avoid spurious bullshit.
Then when that manager got laid off, and I got a new "I'll shield you" manager, who largely just takes whatever crap he's told and passes it on and says it's better to just take it than risk making waves. So I end up having to waste more time than I used to.
But absolutely was valuable in an early manager who took on the politics direct while I just did my work in early career. Of course he got fired for keeping documentation on a certain executive making a terrible call and actually sharing it with the senior leadership team when that terrible call went bad. Senior executives canned the executive, but let him fire the manager on the way out. Guess if there's one thing worse than an executive incurring risk, it's a snitch.
Problem being, lots of middle managers got the job via The Peter Principal. My last boss landed the role that way. He's an astonishing devops guy and one of the 5 smartest humans I've ever known, but he cannot manage people.
Quit my job at Lowe's when it became apparent the regional theft prevention manager hated me. GM passed me in the store and said I better watch it with her as he couldn't protect me. Loved the three guys above me, but I was out a couple of hours after hearing that.
what about his manager?
Response for politics, priority shifting and upholding the morass of corporate bureaucracy. He’s worth every 100k.
And replaceable.
What would you say you do here?
I take directives from the CEO and board and deliver them to the goddamn employees, so they don't have to deal with them. I have people skills, dammit!
so they don’t have to deal with them
Yep, that’s the entire purpose of middle management. Making sure the c-suite doesn’t have to interact with the plebes.
What do you mean "no humans involved"? The LLMs are statistical models trained on humans (and their behaviours) and so obviously the outcome is a representation of exactly that. They're just worse at it.
My dude if you can't tell what that sentence meant I'm afraid you are actually an LLM yourself.
Sorry.
Honestly, humans have been doing agriculture for thousands of years and we're still hung up on who's allowed to cultivate what land.
So 11 minutes isn't that bad, honestly...
Surprised they didn't invent Agile, frankly
Could have been an email
They should try again after they fix the problems. Though i wonder if they even understand what the problem was or are they "allowed" to fix it.
It's pretty clear to me that the problem is that the division of labor wasn't clearly defined. This seems like a management issue.
Like, with 50 players, you can make 5 teams of ten. Assign one team to mining, one to woodcutting, one to crafting, one to building, and one to food production.
Then within each team, give each one a zone to focus on. Each crafter gets a workshop, each fooder gets a field or some coastline. Two or three different mining locations and you can assign each miner a cardinal direction for their main shaft. The woodcutters can replant the trees as they go, so each one only needs a moderate sized area.
Then the builders can be divided into categories. Maybe 4 on wall duty, 4 on buildings, 2 on paths and lights.
Or better yet make it only 5 woodcutters, 5 crafters, and 5 food producers. Then you have 15 extra players and you can set 10 of those to act as guards to protect the others from monsters, and 5 as runners to carry materials from the mines, groves, farms, etc., to the town, and carry replacement tools, supplies, food, etc. back to the stations.
Once the structures are complete, the builders can join the other teams. Eventually when you add stuff like enchanting and brewing there will be more roles, like hunting monsters for exp, or fighting blazes for their powder.
But yeah for early game, it seems like a simple solution, and the answer is proper delegation instead of a free-for-all.
But yeah for early game, it seems like a simple solution, and the answer is proper delegation instead of a free-for-all.
That's the problem. Humans would each gravitate to what they do best and organise themselves into a structure relatively quickly.
LLMs get stuck in circular patterns because they don't have the ability to do that. They can't think outside the box. None of them arbitrarily becomes the boss to take charge. They're all yes men stuck in endless board meetings where even just flipping a coin would break the deadlock. But they are incapable of realising that, because they are just statistical regurgitation machines flavoured with a bit of randomness on the most probable outputs.
I think the lesson is that a swarm of LLMs alone isn't going to be effective, but there needs to be a human in the loop to direct them intelligently
I think the point of having something that should be able to do stuff autonomously, is that it should be able to define these things by itself. Or at least the value of it would be in it being able to do it. But would giving up control be what the ones in control want to do? I wonder if that is also one of the things that keeps the corporations from getting any closer to more functional llm that isnt just a brainless yesman.
Not surprising.
Their solution to everything LLM has been to add intermediary LLMs to manage the one we interact with to add "guardrails" or point them to a calculator.
It shouldn't be surprising that they utterly fail at basic tasks on their own.
Modern A.I. is just 4 LLMs in a trenchcoat
Their solution to everything LLM has been to add intermediary LLMs to manage the one we interact with to add "guardrails" or point them to a calculator.
Isn't this the crux of one of Rick and Morty episodes?
Fermi paradox was really referring to intelligence eventually creating a paralyzing bureaucracy.
Lol
Are we training the Matrix now?
Can we go back to 1999 then please? I need another timeline
But medicine was so much worse in 1999
Lots to discover then!
That’s how humanity lives on forever. The Matrix had them using us for power, but they won’t need us meat bags at all, really.
Fun fact, early scripts for the original film had Humanity plugged in to share computational processing Power for the Machines, but the studios didn't believe audiences could grasp that so it was changed to providing electricity.
Which is impossible, the whole feeding cycle for harvesting us as batteries would yield less power than just using the feed source for power.
Another thought is we are in there to help the AI improve. As with our world the training data needs to keep coming, we are writing the programs that control us in the matrix
Oh. Just like real life.
this post was submitted on 01 Nov 2026
234 points (95.0% liked)
Technology
89424 readers
4622 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS