I worked for a large grocery store chain between 2004 and 2010, there the AI had already taken over. Though we didn't call it an AI, just 'the system'.
The system was able to procure goods with stunning accuracy. Sometimes it even surprised us. For example: we'd receive a much higher number of BBQ meat than typical, only to realise it was going to be amazing weather the coming days.
The system was also making schedules for the re-stocking team with the same accuracy. It knew what it ordered, the exact time the truck was due to arrive, at which time the personnel was available (not at school, for example), etc. The system told us when to show up, and when we had to be finished.
The system was even able to pinpoint cashiers that were stealing money from the register, by correlating cash count of the individual registers to the schedule (which the AI also created). It would then 'isolate' the suspected employee to a single register and order a cash count (which was done by hand) at the end of the day.
Basically the only reason why there still was a manager was for responsibility and babysitting (most employees were teens that did dumb stuff). And the occasional Karen who wanted to talk to a manager.
I remember being being thoroughly impressed with the efficiency of the software back then. In some ways it must have influenced my career path (I now work in automating businesses). I left in 2010, I can only imagine how much more advanced this system has become since.
Sounds very much like a lot of predictive analytics places like walmart do. Walmart has spent hundreds of millions on solving the problem of getting stuff into stores just before it's needed, and some of their work on weather modeling, yearly holiday predictive impact, etc is very interesting, and just a bit scary.
Walmart likely had data on which states were covid baby boom states and which ones were baby bust states well before any of the rest of us had even a sliver of an idea how fertility would play out during the lockdowns.
> That sounds like truly valuable ERP, and largely devoid of AI.
Yes. That was kinda my point. Most of the time, it is just a good piece of software, not an 'AI'. In fact, for most situations you do not want an AI, because you want to the software to behave predictably.
My point being is: good software has been replacing need for humans to manage processes for a long time now.
"In fact, for most situations you do not want an AI, because you want to the software to behave predictably" is very good statement that very very often is skipped from discussion.
Thank you.
Apologies if I missed it, but where in your story did a human get replaced? There were still people working inventory, cash registers, and managing.
I say this because it seems pretty obvious that software can improve the productivity of employees (including managers). But what the blog post at the top says is that AI will replace managers as a category of employees—a much stronger claim.
Machines don't do jobs, machines do tasks. It's very rare for a company to fire people because there is now a system that does exactly what they did before, no more no less. But when you take all the core tasks that a person did as part of their job and automate them away, such that the employee now is doing something completely different, they have been replaced even though they're still on the pay roll. It's functionally equivalent to firing and rehiring them for a new position. The manager may still hold the title of manager, but if they're not doing any management tasks, then they have still been replaced.
> when you take all the core tasks that a person did as part of their job and automate them away
But that is not what happened in the story I responded to. The manager still had tasks exclusive to them (dealing with young employees and angry customers) that had not been automated away.
Tasks exclusive to them, but not the tasks of a manager. A manager manages. As OP stated, the position with the title of "manager" now has the tasks of a babysitter.
Perhaps a human wasn't fully replaced but many responsibilities were. No human had to keep track of inventory, make schedules, or watch for theft from cashiers.
> Basically the only reason why there still was a manager was for responsibility and babysitting (most employees were teens that did dumb stuff). And the occasional Karen who wanted to talk to a manager.
> If predicting and planning outcomes and with previous data is not some form of AI, what is?
Algorithmic forecasting.
If you squint really hard, the entirety of computer science falls under "AI". But practically speaking, today AI refers to ML.
It's quite possible a savvy programmer integrated the weather forecast to the ERP, with specific products tagged as relevant to a sunny (or wet) outlook. That's not AI, that's just an algo.
It's also possible there was ML at play, but that's not clear, and not necessarily the case.
Oh, ok. If you go with the definition that says that AI is what tries to solve any problem that is still not solved well enough to be useful, then yes, useful software doesn't have it.
AI is a pretty useless name, anyway. Since no two people will agree on its meaning.
Anyway, I doubt people tagged what products are most sold on good weather. That's the kind of thing that is almost always discovered by statistic analysis. (We call some statistic analysis "machine learning", while others that do the exact same thing we don't.)
The predictive planning _could_ be done with AI that tries to automatically infer what will be bought. However, it could also be done by a (maintained) list of hard-coded rules. If it was some form of automatic inference, perhaps one could call it AI.
I simply don't think a hand-designed decision tree for a specific task can be called AI.
I guess that depends on what you consider AI. In his example of "good weather -> more BBQ stuff", of course that might be some particular hardcoded rule that occurred to a human at some point, but it's far more likely that datamining identified that particular correlation, and many others (like the famous example of the grocery store that started stocking beer next to diapers when data mining discovered that beer and diapers were often bought together).
Nielsen (of TV ratings fame) has been selling this sort of data mining to grocers for decades. Decades ago I worked for a small regional chain and we based a lot of decisions off that data - we could have never done it on our own.
For any item in our stores we could get a list of other items that were commonly purchased together, for example. I'm sure it's even fancier nowadays.
Ordering stuff not middle management. I seriously doubt an AI claiming MM know how could produce a 3-10 page telling me what it is. And unless AI replaced all the staff, I'd wanna be in the room to see NFW looks from people getting management direction, reviews, budgets standing on articulated customer priorities,and cross functional coordination from AI
I think this article fundamentally misunderstands the core function of middle management. You can boil it down to 2 key tasks (on the front lines of a very large organization).
a) Interpreting: either top->down by explaining and enforcing corporate policy on hourly workers or bottom->up by ensuring workers enter data to be reported on correctly
b) "Encouraging": an employee calls in sick, again. Neither firing them or giving them the day off every time are desirable. The managers role is to "encourage" the performance and attendance (in one way or another)
Neither of these are low hanging fruit for automation or even middle hanging fruit for automation.
edit: I would also add that the comment "Union formation across the retail landscape will force corporations to reduce management head count" is backwards. From a management perspective, if labor gets rowdy you need more overseers to keep them in line. More union activity = more management headcount
'Middle Management' is also not a well defined term.
For some, managers like you have described could be thought of as front line managers. Or supervisors.
And middle management are the layers, usually many, that separate the front line managers from the senior executives.
The role of middle management is often to oversee budgets, and be accountable for financial performance in certain geographic regions or business areas.
Middle managers don't employ workers. They employ other managers. Lots of these kinds of jobs will probably be made so simple that entire layers of management will just be removed.
Some examples from chain retail might be:
The franchisee of say, a McDonald's, is in some sense a middle manager. They hire store managers to run the stores in the area that they're responsible for. And they're accountable for the financial performance of all the McDonald's that they own.
The 'Regional Manager' of some large company. They won't be involved in the day to day activity of the business, but they're often an important cog in the wheel because they have local visibility on the financial performance of the business in the region, and are best positioned to allocate the budget in ways that maximize financial performance. Again, they don't hire workers, they hire other managers.
> bottom->up by ensuring workers enter data to be reported on correctly
Isn't this were management fails? Gathering the data that is required to decide
well?
Eg my SOs company recently introduced the third generation of company wide
management software, but as before nobody really used it and all important
information is still collected in spreadsheets.
"AI management" could start out like Google Assistant, just have a button on
your desk to set up reminders because it actually is easier than opening some
application and typing in the reminder.
The thing is that it has to be extremly simple to do. Even thou I am not a fan
of constantly listening devices, a company wide "OK Google/Hey Siri" would be
extremly valuable.
Imaging two coworkers standing in the company floor:
- "so about feature X?"
- "arg ... yep ... I'll get it done by next week"
- "Computer!?"
- <beep boop>
- "Create a reminder for [Coworker] to finish feature X by next week"
- <ok human, I have created a reminder for Coworker>
This is literally not sci-fi straight out of Star Trek anymore, but something
that (at least) I use daily to remind me of little things like getting the
laundry from the basement :-)
> Imaging two coworkers standing in the company floor: - "so about feature X?" - "arg ... yep ... I'll get it done by next week" - "Computer!?" - <beep boop> - "Create a reminder for [Coworker] to finish feature X by next week" - <ok human, I have created a reminder for Coworker>
More like: " - "Computer!?" - <beep boop> - "Create a reminder for [Coworker] to finish feature X by next week" - <Creating a Redmine server for [Coworker] > - "No! Stop!" <Are you under threat? Should I call police?>
This is roughly my experience with current "smart assistants"
Calling in sick regularly is often a sign of some other problem. Any competent manager will encourage an employee to either deal with that problem or arrange for help.
In a dystopia, the manager will indeed ignore that problem and 'encourage' the worker to come to work. I believe this style of management can be replaced by an AI quite readily, and I believe this has already happened by rather trivial software (think the fake gig-workers for uber, lyft etc. Don't come in today? Your rating will drop tomorrow, along with your wage).
The limit on sick days only matters for payroll decisions (what bucket will the day off fit in, paid leave or unpaid leave?). IME that's handled directly by the employee most of the time when they request the day off in the system.
The manager gets involved at performance evaluation time, or if it has severe otherwise impacts (you're then asking the person to potentially leave), but I don't see the "figure out a solution" part of it. Asking someone to be less sick feels complicated.
It’s not just about “asking someone to be less sick”, you need to ascertain if there are other factors: are they over worked? is there a serious condition that you might need to support? Or conversely, is the sick leave legitimate? Some employees do abuse sick leave.
Either way, a good manager can work with the employee to get a good outcome for both parties.
I completely agree with you on the overworked part, though it's a delicate situation if you're their direct manager and have to rely on them to report how much work they are doing. It happens (a lot more than we're willing to accept I think), and it becomes more a discussion on management than on the employee per se.
On the other parts, it might be a cultural difference: I'd expect all of those to be handled by HR.
For instance, specific work time adjustments, additional rest time due to a physical or psychological condition, family issues etc. are delicate and potentially private matter that go beyond the manager's supervision. The manager needs to be notified of the resulting discussions, but I wouldn't expect them to get all the details directly from the employee for instance.
That's also where the abuse of sick leave becomes a delicate issue to address at the manager level: companies have policies on how you justify the sick leave (doctor notes, medical bills etc.) or they let it open to the employee's judgement call. The former is clear cut, and abusing it means faking papers, so it's straight in HR territory. The later is more tricky, but I'd hate to be in a position where I have to hear the employee's story and decide if it's fake or not, potentially putting a the target on my back if the employee gets punished for actual issues I underestimated. It feels like a real shitty position to be in, I'd probably not bother confronting the employee except if they have literally dozens of sick leave days available.
> it might be a cultural difference: I'd expect all of those to be handled by HR.
Possibly. Generally, I'd only involve HR as a last resort. Go to them for guidance, certainly, but in terms of dealing with an employee, I prefer to talk to them directly.
> [abuse of sick leave]
Completely agree that it's a delicate situation and you need to be very careful, or (in some countries depending on employment law) you could be open to a claim of wrongful dismissal.
All of which is why I don't expect an AI to take over people management roles any time soon.
>Neither of these are low hanging fruit for automation
The Uber app already does both of these things. Encouragement is done via gamification and adjusting pricing and the app itself collects all the data it needs and explains everything the driver needs to know.
Ah yes, the Uber app. Because all work is as simple and collaborative as going to a particular point to pick up a passenger, everybody dreams of working for a worker-friendly organization like Uber.
What they are saying is that for some jobs, it's already like this, so it's not completely out there to consider it expanding as tech gets better. Uber being good or bad is besides the point.
Certainly it cant be applied as effectively to every domain but it's already spreading.
In some domains I'll bet it will mean that middle managers will simply have the tools to be able to effectively manage 60 instead of 10 workers. In others perhaps it wont change much. In others like taxis the middle manager will be made entirely redundant.
I dont think it's necessarily a good thing either but that doesnt mean it's not going to happen.
Correct. Middle management is pretty much all human management. AIs are miles miles away from being able to do that. AI make good middle management even more critical, by removing the non-human aspects of the job
Most of the low-weak managers rely on data. That's how their affirm their decisions to wider group. AI can totally be a better data predictor and user than humans. If all a manager do (middle or higher) is collect and interpret data, they should be replaced by a program
I am talking about bad managers here who throw around data a lot. This is not a bad take. AI at this stage is less of a doer, more of a predictor based on data. It should replace management glut that many companies face. It will also make things more consistent instead of leaving them to whims of someone
In some ways tech can do better. It can constantly monitor many more details than a human can. It doesn’t give you a disapproving look when you take too many sick days but it does calculate a score showing over the month you got less done than your coworkers.
Oh boy, I would love to work for a company where my boss monitors me constantly then sends me a daily score! /s
There is a human element to managing people that AI simply cannot do. A boss (good or bad) can change his behavior based on anything from office rumor to the way you look on a certain day. He can decide to go easy on you for a while because he heard through someone else that your dog died last week. He can simply look at you and decide that you look depressed and pull you aside and ask you what's wrong.
Human interaction and relationships are so complex and nuanced that no AI is going to replace it for a long, long time - until we have proper AGI at least.
The only thing I see AI doing is putting some proofreaders and crappy graphic designers out of business.
I think the main problem AI faces is trust. A manager needs some level of trust from both the higher ups and the workers. Not necessarily a "I trust my life with you", but more of a trust in basic human decency. I can imagine higher ups trusting some piece of software, but I have a very hard time imagining workers trusting an AI.
This is not something that can be fixed by improving the AI, as the true problem lies outside the domain of AI.
Once AI has proven itself to be trustworthy enough, it maybe can perform a mgmt role, but I have no clue how it could ever prove itself without being trusted in the first place.
It reads like you are describing reporting. How many frowny faces were sent out at Store 123 last month for attendance? But the reason you need lower/middle mgmt is that the real question is "so what do we do to stop having these frowny faces?"
What the system does is the rostering system preferences the workers with the highest scores when assigning shifts. And I suppose makes this number viewable to the workers so they can see why they have to optimise.
You do not need AI to treat the workers that have historically performed well (as measured on some quantitative scale) better than others. Most businesses just don’t do it, as finding an objective scale is hard and people have good reasons for underperforming for periods of time (sick leave, parental leave).
The supply of even untrained workers is finite, so you have to treat them as people.
Ok, let's roll with that. If I'm Alice, why do I care about Bob's score? Why do I care about your system at all? I'm trying to pay rent and not work on my kid's birthday. WHO is going to explain why the frowny faces matter? That person is your middle manager
Sure, and workers have great fun exploiting the hell out of whatever shitty metrics they're given but certain stuff just doesn't get done, and workers who like dealing with humans (probably the best sort of workers to have in a retail organization...) leave.
We could do this with developers too. Give the biggest raise to the developers with the most lines of code committed, and maybe design a shiny dashboard with it as a number to replace the project manager. What could possibly go wrong?
I think this is spot on. However, I think the author is half right: Middle management is probably going to be replaces, but I don't think it is by AI, but by competitive incentive structures -- its algorithmic governance anyways.
sectors and industries were labelled "essential worker", e.g. grocery stores. Do you think management within those sectors stopped "interpreting" and "encouraging"?
In some sense, that is lower management. Middle management sits between the people who connect with the work floor, and the people who actually make decisions. These positions do seem ripe for AI replacement.
This is a poorly reasoned or at least a poorly articulated essay. AI to this point is built to optimize some objective function. What is that function with respect to middle management? Some sort of mathematical expression would be interesting but even a plain-English expression would certainly be helpful. And what exactly are the inputs and output? I guess the inputs could be a roster of employees along with some performance metrics, and output is a list of who is kept and who is fired (?). That’s the best I can come up with, but I still think it’s an absurd and impractical eventuality. The article doesn’t even make an attempt to explain.
This isn’t some sort of nitpick; AI to this point has not in any way encroached upon personnel management to my knowledge and a prediction in that direction ought to elucidate exactly how that could be expected to work. Managing workers involves empathy and understanding human emotions generally and I’ve seen no indication that AI has made advances into this space recently despite its rapid progress elsewhere.
This is an interesting critique of modern AI: almost no problems have closed-form mathematical descriptions.
The "Ng Reply" would be that the objective fn can be approximated by a finite composition of simpler functions, tuned by data (ie., relu.relu.relu...)
However if you think about it for a second, that's clearly absurd in almost all cases. We know that we cannot even in principle formulate a closed form expression for (here, eg., people management) -- the idea that there exists some meaningful approximation feels provably silly.
> What is that function with respect to middle management?
It's probably easier to create such a function for the CEO and board. For instance some risk weighted present day value of future net earnings, possibly combined with short- medium term stock market value to please stock owners that have a shorter-than-infinite horizon for their ownership.
Once that is decided, mid level manager objectives reduces to whatever optimizes the CEO objectives.
If someone can design a company structure where you decouple your employees from their superiors entirely, it'll be inhuman but I bet there will be people who will try. The AI hype is just so strong.
This reminds me of an argument I got into at a bar with someone who claimed robots would inevitably replace bartenders. I took the view that bartenders would be the last occupation to be replaced by robots because if the goal was to minimize the cost to consume alcohol people wouldn't be coming to bars in the first place.
There's a misconception that jobs with higher pay and more prestige will be the last to be replaced by robots. In many cases I think the opposite is true and in this case I think upper management will be replaced by robots before middle management.
> I took the view that bartenders would be the last occupation to be replaced by robots because if the goal was to minimize the cost to consume alcohol people wouldn't be coming to bars in the first place.
And the elephant in the room, people want human contact
The "AI revolution" is a business owner dream, not a worker's dream, nor a citizen's dream
>> I think upper management will be replaced by robots before middle management.
You are already correct.
In a sense, the highest level of management is the shareholder/owner. Capital allocation is one of the most highly automated tasks currently. We see this with SAA (Strategic Asset Allocation) programmes being run across all major capital managers and owners.
How is that an example of AI replacing management? The AI might determine what to invest in and thus who the owners of a company are, but humans still need to make decisions for the company.
> I took the view that bartenders would be the last occupation to be replaced by robots because if the goal was to minimize the cost to consume alcohol people wouldn't be coming to bars in the first place.
For a traditional bar, I agree. In places where the "bar" is not the primary attraction, such as sports arenas, concert halls or even some types of night clubs, the bar may be just a bottleneck where you may have to wait 5-15 minutes to order a drink during breaks. If you could order your drink on the phone, and have a robot (or even a waiter) serve it at your table, it may be quite convenient.
There could still be 1-2 bartenders in the bar in places where there used to be 5.
Is the bespoke robot with a service contract from a for profit company cheaper than someone you are paying $12/hr? That's the equation here. Lines are long, but that just means you have to pay fewer bartenders, and people will still wait in line anyway. Sometimes these venues aren't even mixing drinks already. They might have cocktails premixed next to the beer taps.
In a traditional bar the robot will mix drinks, while the bar tender talks. A few bartenders have mixing the drink part of their act, but the rest don't need to pour drinks.
Science fiction had long suggested bartenders have a degree in psychology or some such field as realistically that is what they are, just not paid / trained like the medical profession it is
Half the fun of having a bartender is to influence the alcohol content of your drinks. If the bartender has no influence on the drink the bartender/customer relationship is weak.
We see all sorts of automation behind the counter, without considering them "robots".
Many bars will have a tap where you place a cup and click a button and it will be filled. Every Starbucks I've been to in the last year has had a fully automated coffee machine. The barista places a cup and pushes a button. The same for filling any sort of cup and fastfood chains. It's either the customers job or the clerk pushes a button and leave it to fill by itself.
I think this furthers your point: The people behind the counter aren't there to make your orders. They are there to take and deliver the orders with some amount of human touch.
Wouldn't upper management be in charge of taking the decision of replacing upper management with AI? Or is this only applicable to small companies (or workers' cooperatives) that had access to these functions from their foundation?
I'm definitely an AI maximalist and worried about mass unemployment one day, but I think this article isn't right; it's too shallow and has some assumptions that aren't right. Chess isn't comparable to management because chess is ultimately based on simple rules. And I think we're already seeing automation and robots replacing hourly workers in a lot of interesting scenarios incl. self-checkouts and app-based automated checkouts in stores, robots that delivers your food in Asian restaurants etc. What I do think though is that wasteful administrative management (and non-management) work is slowly getting less (caveat: I work at Asana and it's our mission to make work coordination, tracking and reporting easier. Same with rostering tools like Deputy etc.) and so we're certainly minimising the "middle manager overhead" but to say that it's being replaced is a long stretch.
Cells of a chess game are like cells of the body. Everything follows rules. /s
It's nice to see, here, this kinda reply hit peak silliness. "Isn't AI just like... thing we do?"
No. But many of the major papers, press releases, etc. want you to believe that. Often, they want to believe that. But none have bothered to seriously find out what we do.
Last year I tendered a logo redesign to freelancers on a specialized website. The experience was not pleasant. After all they are humans. They had diverging interests. Two designers did something I liked and I would have liked to combine their work, but they didn't cooperate. I had to compromise. One of the designers got very angry at me because of that compromise. The other one didn't give me all the files I requested and messed up half of the files. Of course it is also my fault. After all, I am human, too.
Stable diffusion is not quite right, either. It wouldn't have worked. However I was struck how easy it is to just give "orders". I didn't need to cushion my demands to make them socially acceptable. I didn't need to be afraid to annoy someone. I just could refine and refine and refine. A lot of pressure just went poof.
I imagine upper management is feeling the same about middle management. The AI spark is already here now. Some people will try to replace middle management with AI or are already trying. Of course they will fail spectacularly. Several times. But there will be a time where someone gets it and then it will spread like wildfire through the corporate world.
I imagine it will be something like a connection of GPT-n (where n >= 3) to ERP data. Upper management will tell the AI to determine detail plans for hourly workers and fill the ERP with orders for them. Hourly workers get a tablet to tick off tasks. The AI will check whether the plan is executed and escalate if something goes wrong they can't fix themselves.
That only works in scenarios where labor is fungible, like logo design. Try telling a key employee all that stuff, you'll get told you can do it yourself.
Perhaps it is wrong to talk about middle management being replaced by AI. It seems that AI as of now is not very inventive. It is a clever mash-up of things that were already done by humans. But most employees just do that. They learn from others in school, college or on the job how to do things in their profession and then employ what they have learned. If they use a computer to execute then actually nothing stops them from being replaced by the computer itself.
In 5 years, the current AI hype will be the same as the state of 3D Printing, IoT, AR/VR.
There is a fundamental disconnect between tech bros’ dreams and reality. These dreams have an outsized expectation of the tech-literacy of the average person. They also make the mistake of completely assuming that human decisions are driven by logic, not emotion.
Oddly optimistic article despite the title, but lol no. middle-management is like analogous for bureaucratic bloat and politicking. It’s been a punch line for decades that middle managers suck and don’t actually do much. This like saying AI will finally eliminate paperwork, dmv lines, or pointless meetings. The Inefficiencies are built in and Hierarchical structures of a traditional company Necessitate someone between the peons on the executives. AI isn’t going to somehow alter that.
Besides I want someone to be between me and my boss. An intermediary between me and the guy who can fire me at any moment.
"AI" is making some amazing advances these days. It's fueling quite a bit of optimism and outright delusion. It's getting a bit tiring to read through all the hype.
Tbh some of the things that seemed obvious are taking longer than expected while wild advances on seemingly difficult fields pop up out of nowhere. At the start of the year I would have told you by now we would have progressed on automated pizza delivery but not be able to generate convincing art.
I bought a kitchen table at Amazon. Now the recommendation engine recommends more kitchen tables and that is going on for half a year now. As long as that is the state of the art, I am not afraid that AI will replace my job, which requires some out of the box thinking.
Wrong. The recommendation engine is working correctly, and people who bought a kitchen table are more likely to buy another one compared to buying a completely unrelated item.
Maybe I don't know enough about recommendation systems, but surely not? If I buy a kitchen table, in my demographic (mid-20s in rented accommodation) it's because I moved into a new flat, or my existing table broke. In either case, buying a new table actually marks the start of a long period where I won't be buying any new tables, and every day from then onwards I'm slightly more likely to buy a table.
As far as I know, buying a kitchen table is so rare that the unlikely possibility that you return it, it breaks, or you are a "serial mover" that rents homes for small times can be enough to make the odds of somebody that just brought a table buying another higher than some random person.
Unless you have some aggregated market data, you don't know either. Up to very recently, nobody knew if the likelihood increased or not. Nowadays a few people are able to discover it, but it doesn't look like they bothered to.
It's like we've learned nothing from the techno-optimist claim that work automation will give workers more free time to pursue their own desires.
In theory, it might be able to at some point. However, in reality, the claim ignores the power structures that the automation/AI/tech doesn't dissolve, but only reinforces.
Some people want to lord over other people, AI isn't going to make that go away without changing the power structures that let it happen.
And some people want and need leaders. It's not a matter of "letting it happen", it's a ore function of humanity. And in the absence of formal hierarchy you get informal hierarchy which is much worse and more tyrannical.
I said nothing about needing or not needing leaders. Most leaders in someone's life did not get the roles democratically. Existing power structures are why those in charge are not chosen in a democratic manner, and AI won't replace management in its current form for similar reasons.
Came here to make your comment. I think though that in the time since manna was written the ratio of machine performance on hands on and creative tasks vs management tasks has radically equalized.
I work at a large company that invested countless dev hours in building an automated shift scheduling system, then gave up, bought a scheduling system from a vendor, and has now spent several years tweaking that system to match their requirements.
Guess what? It works OK when everything is fine, but if anything even mildly out of the ordinary happens and more than a few people are unavailable for any reason (year-end holidays, travel, conferences, etc), boom, it's back to a manager manually putting together the schedule again.
And this is the best case scenario with motivated, largely self-managing employees who can be trusted to, say, figure out shift swaps if something comes up at short notice. Good luck automating scheduling at McDonalds, and remember this is only a tiny part of what a manager has to deal with every day.
AI is truly amazing, but don't be fooled by the incredibly graphical artworks it is delivering. All that stuff is weird, dreamlike images that make no sense. When you look at a really great AI image, it is the one cherry picked after discarding 1,000 nonsensical garbage images.
And chess has nothing to do with AI unless it's an AI learning how to play chess.
All in all there's nothing even close to human behavior anywhere in the AI landscape, just things that kinda almost look like human behavior if you squint enough.
The Shannon number doesn't exist for a middle management role, which could consist of so many variations from organization to organization. I don't disagree that AI assisted decision making is a thing that's happening, but I think the author is naively understating just how far away we are from 1:1 human to AI replacement.
It's not a great article, but the main line of objection in comments here seems to be "This won't happen, because AI won't do the job well and certainly not better than humans." Sadly, we already know that that's not a reason for things not happening.
For example, robot staffed called centres are worse in almost every way than human staffed ones. And yet they are widely deployed in certain use cases, e.g. utility companies. And, more broadly, even when humans are involved in some chain of interaction between person and organisation, they sometimes have limited options because the computerised system imposed on them is not to designed to empower them to make on-the-hoof discretionary decisions.
Organisations chose to use automated systems of various kinds, because they think it saves money and because the reduction in service is not so appalling that that they lose customers. Or, of course, because they are an essential service, and people need to use them anyway.
We might think that competition would mean that organisations who didn't do this would do better, but often that's not how it actually works. Having an appalling system for escalating complaints is terrible for the relatively small number of people who need to make complaints. But organisations are often not optimising for good handling of edge cases even if, in some better world, we would like them to be.
(Or, indeed, never mind what we would like, organisations might not optimise for edges cases even if it might be rational for them to in the sense that it might actually save them money - edges cases can be expensive - or win them reputation - handling edges cases well is good "marketing" etc.)
Call centers are a special case. Pure cost, and actually helping customers often results in either extra cost (returns and replacements) or loss of revenue (credits, cancellations).
Do you have ten other examples (besides payroll, billing, and computer and communication service delivery)?
Probably would be at least a few SE managers eliminated simply by Jira being more responsive. Not that it would replace all of them but if you currently have five, you might be able to get away with having those same projects spread over four with a similar level of effectiveness if the primary tool wasn't so clunky and time consuming.
This article doesn't even understand what middle managers do. The functions listed like "scheduling, negotiating, training" are for line managers, who are directly responsible for supervising the line workers actually doing the work.
Middle management is the next layer up the food chain: the managers of managers, responsible for executing strategy coming from up high, but lacking executive power to actually change that strategy.
Unlikely 1 middle management cost lets say $20k plus insurance for 1 person where as 20 hourly worker's cost $10k x 20 = $200k. Business will always try to go for decreasing the bigger expense. Another thing most people don't seem to get small business are 70-80% of the worlds economy so the middle management does not exist or is the boss family or friend.
Maybe, but remember most hourly workers have been automated out of a job over the previous decades. What is left are the hard to automate jobs. My company produces more product than ever on about a tenth of the hourly workers we had in 1950. Easy management jobs will be automated away, but that will leave the hard ones.
Middle managements primary job is making an employee feel supported, encouraged and empowered.
Guarantee an AI will be able to do none of that lol.
My first objective as middle management was always to bond on a personal level to some degree. I’d do this on a weekly basis to build trust:
- Talk about personal, but work appropriate topics — gardening, pets, sports, what have you.
- Discuss projects and objectives. Discuss concerns, things that went well, any support needed, etc
- discuss professional development (we’d outline these ahead of time). What progress is being made, how to go about achieving objectives, etc. often this was outside of work.
- discuss / mediate any interpersonal conflicts on the team
The most important aspect of my job as middle management was always the bonding. I doubt an AI can relate to gardening or getting a beer at a happy hour watching the local college football. Without that trust a team wouldn’t follow the AI.
This article doesn't seem to be grounded in reality.
I see self-checkouts, electronic toll collection, vending machines, automation of assembly lines, robots doing order picking in warehouses, self-driving trains in airports and many others replacing hourly workers today.
However, I am not aware of any AI-system replacing a middle manager.
I know it is popular to hate on middle-managers. But good middle manager culture is probably one of the best predictors of company performance.
I wonder how good middle managers can be in retail. I would guess good staff can always turn around a crisis.
And this is why I think AI will either replace all of us or just change the way we work.
Software will and does kill jobs, but only those that are easy to move around anyway. Take generated art for an example: If a simple prompt that you can send to a contractor is enough to satisfy your needs there is a good chance that you are already or will be able to replace artists with AI, but as soon as you want something more, AI alone will not be able to help you. After all an AI will not be able to thrive in an organic human environment, at least for the forseeable future.
The moment the latter is possible, it is a matter of time that AI gives all of us the option to slack off forever.
Every manager I have ever worked with has struggled to quantify the performance of their employees, or to define the inputs that affect employee performance. Good luck building an AI that can manage without knowing either of those two things.
AI will enhance middle management and make things easier, but people problems are complex in a way that AI is a loooooong way off from understanding.
Once they figure that out, it's time for us to kickback, relax, and enjoy the ensuing utopia and/or apocalypse.
AI is far more likely to replace upper management and executive levels, especially in large corporations, since those roles are a lot more 'by the numbers' and act on probabilities rather than specific, nuanced, and complex instances.
I think this is looking at things from the wrong perspective
Corporations can be viewed as artificial intelligences. Some components of that corporation are human, such as the marketing professional designing an advertisement or the executive deciding to invest $40 million in IT infrastructure. Some components of that corporation are not human, such as an automated system that sends emails and tracks metrics on the performance of the marketing campaign.
Rather than asking if this or that job will be automated, ask whether the total percentage of (loosely defined) 'stuff' that a corporation does will be more or less human over time? I would argue that the answer is, with decades and decades of evidence, over time corporations will do more and more with fewer humans.
Jobs aren't automated, job functions are. If most "middle manager" job functions are automated, we might still see middle managers but as a smaller percentage of the overall corporation. If some retail functions are also automated, then we might see both stay in the same ratio, but the business is capable of vastly more revenue without increasing the size of the staff.
Regardless, I would think it's largely accepted that in the foreseeable future, corporations will continue to do more 'stuff' with fewer humans.
I hope this new boss is called "Decision Matrix" with the motto "Harsh and unrelenting". That'll capture the linear algebra roots and the sterile nature of the thinking.
Whilst middle managment is often rightly panned as ineffective, they obviously do bring some sort of value, or at least enable it. They may just need to keep a few of the meat puppets around to act as proxies for the AI so people don't know they are starting to live in westworld.
Middle Management is more or less resource allocation (so optimization problem maybe of the stochastic, multi-agent, multi-criteria and/or multi-objective flavour if you want it to be super complex).
This article is right for the wrong reasons. AI/ML (which approximates optimization problems due to inputs not being clearly defined) is not required for the replacement of middle management (an actual optimization problem with clearly defined inputs).
No, it will not. Maybe it will replace higher management some day. Long time ago Napoleon noticed that strength of an army depends on its corporals. The same with companies, tech leads, middle management makes a true difference - you might have the best strategy but without using proper tactics it will never succeed. And tactics, all those day to day activities, are managed by those people in the middle of management ladder.
It's not machine learning that's required to make computers better than humans at chess.
Non machine-learning chess engines engines based on alpha-beta pruning and expertly-crafted evaluation functions have been stronger than all human chess players for a long time now.
Given this blog post, one might ask whether replacing middle management with alpha-beta pruning and expertly-crafted evaluation functions makes sense.
I think had the author gone with a thesis more like "AI will remove the decision-making from middle management, leaving purely the human-to-human aspect of management," it might have been a more palatable (though still arguable) essay. As it is, there's tons that managers do that AI will never really do, because those things can't be boiled down to mathematical formulae.
I see a lot of traps with this. Who manages it and maintains the AI? They have a lot of power over the company and most companies won't have skillset. What happens when there is a crisis (natural disaster, employee walkout)? What happens when there is downtime? What happens when the AI goes off the rails (fed bad data?), who can override it? What happens when the system is compromised?
What are the assigned goals of the AI? Short-term profit, long-term profit, employee retention/satisfaction, public image? Those have to be assigned priority values. Those programmatic goals will likely not align with stated company goals. Will people learn how to game the system?
What is the training data? If the data is biased, will the AI learn the bias? Does the AI understand legal protections? If someone makes a harassment complaint against a coworker, is that input it processes? What if someone requires medical accommodations?
All that said, I support it as a tool to be used by people. I'm just against overreliance and need to be able to gracefully fallback.
This is maybe more of a semantic disagreement, but I would expect the frontline manager, not middle management, to be the first in management to shrink in population on account of AI. If software and AI is to be made more responsible for managing the hourly worker, then naturally your frontline manager should be capable of managing more humans at once, thereby reducing the company's need for frontline managers. As a result, managers can be made responsible for more people at once, teams can be larger, and fewer frontline managers being needed.
However, claiming that management will be affected "before" the hourly employee is an invalid prediction in today's age. The hourly employee has already been disproportionately affected by AI, some even made obsolete by robots or automation. Retail (especially brick & mortar) has already seen many of those jobs be replaced by much less sophisticated advancements (ecommerce, d2c, etc), arguing to the contrary would be impossible to back up at this point.
I think the one-sentence version of this thesis is much more interesting and true than the specific build out:
"AI" will very obviously continue to make inroads into traditional human knowledge-work and adjacent domains, automating roles and (for some metrics) outperforming humans.
The lemmas are playing out in too many domains to track any longer already: there will be blowback largely defined by intentional and unintentional confusion and misplaced identification of root problems; and by chaos where historic structures (legal ethical and otherwise) which were reasonable when the only agents exhibiting agency were human, now have to contend with a spectrum of collaborative entities mixing individuals, communities, and "AI" of all kinds.
"Middle management" is a misnomer, but the premise is reasonable, that which can be quantified will be optimized.
As noted ITT what and how quantification occurs and whether that tracks "reality" (or traditional fictions at least) remains to be seen.
My own take on that is that what we will mostly see is evolutionary pressure weeding out unviable failures, accommodating for a while at least midling failures, and a slow quiet creep towards deeper and more pervasively "correct" solutions,
all trending inexorably towards even less and more existential challenges at both the philosophical and societal level, and the basic one defined by the paperclip peril of the Alignment Problem.
I am unsurprised that certain domains (like effective definition of and execution of a business plan through an organization made of non-fungible humans) will prove exceedingly hard to fully capture.
But I am now regularly surprised by ice-shelf-break level dramatic gains in domains that seemed until recently equally unassailable.
The ingenue in this domain for me is Whisper. It's just... solved that problem for good, it seems.
"As a new general-purpose technology, robots have the potential to radically transform employment and organizations. In contrast to prior studies that predict dramatic employment declines, we find that investments in robotics are associated with increases in total firm employment but decreases in the total number of managers. Similarly, we find that robots are associated with an increase in the span of control for supervisors remaining within the organization."
"AI will replace middle management before robots replace hourly workers"
No it won't. In my experience, management does many things that they don't want a paper trail of. Having a system means there will be a record. No company wants that liability, and few would actually adhere strictly to their own policies.
I think the broad consensus was AI would replace Manual, Knowledge and Creative workers in that order.
Seems to be trending the other way.
Recently seen more tele-operational solutions in robotics space to replace or compliment manual labour. There seems to be something to this where full automation is not feasible or where it fails and needs recovery.
Maybe we are also seeing a revision for the definition of "creative" work.
Executing a task like "I want a pic of a doplhin jumping out of the sunset with a cereal box, winking at the camera" is maybe not going to be in the same bracket of other creative work in the future...
Their order never made any sense anyway. Manual laborers are doing the most complicated, varied, movements optimized over time. Every new movement for a robot requires more mechanical creativity and introduces more fragility. Knowledge and creative workers click on things and type stuff.
Manual workers are expected to maintain themselves, generally don't get any health insurance, and if they get damaged are expected to repair themselves or be replaced. Complicated machines break down and require specialized maintenance and repair from highly skilled professionals. You can also pay knowledge and creative workers trash, but their output is very slow compared to the machines that will replace them.
Manual workers can be abused and ignored, and if they roll their eyes about it, fired without cause. If you have a robot, that is now the center of your business, and if the robot isn't happy, no one is happy. Knowledge and creative workers are more likely to kick up a stink when abused.
Manual workers and robots complement each other, raising productivity. As knowledge workers and creatives start working with ML tools instead of Creative Suite and MS Office, it will gradually become harder and harder to justify why the robots need the artist's help; you can be a good photographer by taking a good photograph, or you can be a good photographer by traveling to interesting places, taking 1000 photographs, and choosing the one you like best. ML gives us the analog of the latter case for free.
It really comes down to the fact that maintaining complicated machinery (to replace a manual worker) is a lot more expensive than maintaining a bunch of commodity servers (to replace a knowledge worker.)
I don't think AI will replace middle management. I think what we consider "middle management" is largely just going to go away over time because it's no longer required or cost-effective. AI replacing it is a big of an oxymoron. AI would just eliminate it, not replace it.
AI is going to replace neither and it's definitely not going to replace human-facing management work anytime soon. Reminder that we're decades and over a hundred billion into self-driving cars and they're a marginal phenomenon. If we'd listened to Andrew Ng we should have stopped training radiologists years ago. (haven't seen a lot of unemployed radiologists recently), and so on.
AI managers and lawyers replacing middle management is a good premise for a sci-fi story but not much else. Not to mention that the article focuses on retail. As long as Google uses middle managers I'm fairly certain ALDI will. They still use fax in Japan. If you want to talk about the pace of tech change in ordinary large business I'd take that as a reference point
We should really be focusing on making it easier for people to do their jobs, not replacing them.
In the end they are the same thing, but it's way easier and more accomplishable to slowly improve a person's productivity than it is to remove the human brain from the system.
I don't agree 100% with that but there's some truth in it. we will get more Github co-pilot like products to do lot of data organization summarization and predictive work. And then humans will do the nuanced decision making and narration. so yes less # of people needed but you people STILL needed. People will be expected to wear lot of hats. "Oh I am Operations Manager" No sir you are not, you are also expected to drive product development and your people development. No bullshit departments like Product or HR Management would exist within a decade. People are going to be judged purely on raw critical thinking skills. Not the COlleGe they went to ExPErienCe they have
Companies are too stupid to get rid of middle managers (as a useless middle manager myself).
There is literally a generation of business leaders growing up believing crap they see on Linkedin whose only purpose is to make useless people feel needed.
Middle managers are here to stay for quite some time - at least until we have the same companies. And they have a lot of money to keep living the way they're living.
I think we can have companies without middle managers, but it will feel like working in 100 different startups instead of working in a single company.
I assume AI managers could automate DAOs (Decentralised Autonomous Organisations) and we might transition to a world where companies are basically all just DAOs
I think this post is simplistic. Chess does not have anything to do with what most middle-managers do and I don't see AI manage humans with more depth, it has to become sentient first essentially. I think for AI to replace most middle managers it has to get really smart, understand problems instantly and assess their consequences. This is something we can't do and I think we will rather see better robots first since their environment is more constrained.
But for "humans as resources" we are already seeing this. If the task is repetitive, clearly defined then we have something like gig-economy where some middle managers are absent. Nobody dispatches uber-drivers, it's all automated.
Probably true for now. But „what a computer can do“ is not a convincing reason. I have been very surprised by that lately and will perhaps be surprised again.
Ask yourself: In your day job, how much “management” of your work do you do for yourself?
If your answer is “I mostly manage myself”, then your manager has already vacated the role we’re talking about automating with AÍ. You will experience this new AI manager as a set of tools to help you get your work done.
If your answer is “I am frequently, directly managed” then your manager will be replaced by software that is able to more quickly and accurately monitor your work, because computers can be everywhere, all at once, real-time, 24/7, and no human can beat them at this. There will still be someone “higher-up-the-chain” from you, but they won’t be interacting with you on a daily basis.
Consider a grocery store. I'm related to someone who worked grocery in the 90s. There was a job called a Merchandiser that would fly to every store in a region, consult with the store to help them with their ordering, and provide guidance with product placement.
Their job was nearly entirely replaceable with good algorithms and camera enforcement. It could have been replaced with a computer, a camera, and three people in the home office that functioned as a panopticon.
I think that one of important purposes of hierarchy alongside the delegation of responsibilities is the delegation of liabilities.
One can let AI handle some of the aspects of responsibilities of middle mgmt, like reports, metrics and assignments. But should things go south for whatever reason, AI won't be able to assume the liabilities and shield the upper mgmt.
Besides, one can't form a clique or faction with AI.
On the other hand, replacing workers with AI still has the lure of "advancement" ... to bad AI hyped more than it could deliver and in the long run turned more expensive to operate, hardly replaceable too.
We need AI personal assistants to handle mundane shit while we do things that solve problems. Most people are reasonably capable of managing themselves once given initial direction.
On the low end of the job market, being able to follow instructions, especially written instructions, is a skill that not everyone has. I suspect it's more economical to have a human supervisor instead of narrowing your applicant pool to people who can follow written instructions.
(Replacing written instructions with video instructions etc doesn't make things much easier.)
Similarly, common sense judgement is also a skill. It's easier to have a human supervisor embody some of that judgement, then to find a way to distill it all down into formal rules.
In the context of your idea, that also means figuring out who is actually failing and needs to be fired.
That all being said: for decades or even centuries the economy has already been moving in the direction you are suggesting.
You don't need an app for that: just write down the rules on a piece of paper or even a clay tablet.
It has, but the salary isn't high enough to fire underperformers, because you get what you get at 10 dollars an hour. It's the reverse of the tech situation you're likely used to - where employers gatekeep as hard as they can and are always desperate. In retail etc there's no gatekeeping but there's also no quality guarantee
The current definition of AI (whose "output" actually IS deterministic) does more harm than benefit as it evokes expectations which can't be met. Automated text translation in real time? Certainly impressive but nowhere intelligent in the human sense of intelligence or smartness.
I am confident that no "AI" will ever replace humans in any meaningful way unless we come up with an entirely new computing principle. Maybe quantum computers will bring us there, I honestly don't know.
Some real anxiety in the comments here. I personally think this is a positive situation - one which could have been avoided if real human managers had a hint of empathy, humility, and understanding of their job : to be human conduits to a mechanical system.
I have had 3 or 4 great real human managers in my career. But that also includes at least a dozen horrible troll-drones. They deserve it. Give me nothing, get nothing. Even the real human ones were just buffering whatever they could before I quit.
It’s for this reason alone that bureaucracy can not be reduced or its processes streamlined in the foreseeable future. It doesn’t even need AI, just digital services. The bureaucrat will fight tooth and nail for the status quo to remain in place and prevent technical updates whenever possible. More efficiency means facing the pointlessness of one’s job. Exhibit A: Germany.
> Artificial Intelligence is now capable of beating the best humans at chess 100% of the time no questions asked [ ... ] or a program capable of making the best possible business decision given historic and currently available data?
I believe the OP has an interesting definition of business decision.
Also that a management AI would be good ... at giving us faster horses.
Maybe if he could back up his assumption with facts and data, people wouldn’t be dismissing it so quickly, but he’s offering nothing in terms of a good argument
Middle managers also tend not to form unions unlike low skilled and low educated frontline workers. Also making good, fuzzy decisions is harder for machines vs doing simple, low skill, repetitive tasks
So the same thing as since the 60s when politicians and technocrats promised automation would allow a 3 days / 15 hours workweek but instead of doing that it pushed people in shittier and shittier jobs, with longer hours, and increased the retirement age ?
I am middle management....a big part of my job is understanding my juniors needs and enabling them to do their jumps by promoting their cause with the leadership. Secondly I lead the hiring and ongoing training of staff...not sure how AI is going to do all this...
Another of these articles. Middle management is mostly about managing people, not taking decisions. The decisions MMs can make are extremely limited. They are hired so that the top management doesn't have to deal with all these pesky employees.
I haven't read the article yet, but here's one approach: don't replace middle management with AI, just fire all of them. Much cheaper and gives 90% of the benefit of replacing these people with computers.
If this is correct, and I think it is, then we should worry about the spiral of inequality that this will generate. I don’t think that the money saved from not having middle managers would trickle down as much as up.
To claim this, you must have no clue about what middle management is. It might not look like it, but it's the most critical layer that creates a healthy culture within a company.
The author talks about AI management solutions that will reduce management head count as if such solutions are available and ready to use. But I'm not aware that such solutions are available yet.
Yes. I coach CTOs and what they often miss is middle management. And I wondered how these tasks why CTOs need middle management could be replaced by AI - and I couldn't square those ideas.
I think it would be more interesting if you elaborate on what specific middle-management tasks you think will be automated. You only touch on that very briefly. If the Roomba gets stuck nobody cares; if a bad decision is made that costs the company serious whompa$, somebody needs to be held accountable.
I don’t think the Roomba is a good example here, it has a basic brain for a reason and it’s not because of a lack of available technology, it’s because most people don’t want to pay $7000 for one.
But someone has to decide what metrics AI will use, how many cousins of upper management can be hired, that sort of thing, and that's where middle management is irreplaceable.
When I was a project manager I wrote a script that basically replaced 50% of my job. Following up with late tickets, asking for daily updates for standup etc.
> Artificial Intelligence is now capable of beating the best humans at chess 100% of the time no questions asked. The Roomba in the corner of my room still gets stuck on things and runs out of power.
This is a ridiculous comparison. Chess is infinitely less complex than the most simple of living rooms. Real world tasks are much more complex than something contained inside a limited, finite universe like that of chess.
I find it interesting that the conversation herein is about the viability, as opposed to whether or not this is a desired outcome.
Personally, I welcome every outcome where machines replace the roles people undertake (myself included). A fascinating and painful set of possibilities emerge. People, indeed societies will need to re-examine their definitions of "work", "purpose", "success" and all sorts of other things.
Perhaps we'll step forward to a post-humanist/adamist economy. Perhaps we'll enter a randist/gattacist society. Maybe a commerce-free world is created - or maybe time becomes a fixed currency? Maybe all, or none of the above are true. But ultimately, for the progression of our societies, I'd love to see these conversations.
Just speaking as a student of Arendt, the most dangerous thing in the world is a rich country with suddenly unemployed middle managers. There's a straight up reason we keep those sociopaths busy.
What does chess have in common with retail middle management? As someone who has managed a retail store and played chess, I can confidently answer: nothing.
Remember when Watson beat Ken Jennings at Jeopardy? And then IBM saturated the airwaves with ads about how Watson was going to revolutionize industries like energy and health care? Notice how we don’t see those ads anymore? It’s because winning Jeopardy has nothing to do with providing health care and energy. Like chess, it’s an artificially constrained game with a clearly defined goal. Whereas business is unconstrained and nobody knows what will succeed over time.
The hardest thing about management is managing people, who are generally self-centered, opinionated, and wishing they were doing something else. How, specifically, will AI be better at that than humans? AI cannot even reliably autocorrect this comment as I’m typing it. How is it going to convince Jack to switch his shift and come in to cover open on Sunday because Jill’s sister was in a car accident and Jill is watching her niece? How’s AI going to do that better than a fellow human?
How’s AI going to do that better than a fellow human?
It will create metrics that will formalize a new paradigm. You know that arbitrary number in your bank account that defines you, that number on your report card, that rating on an Amazon review, that number that’s defining global inflation, that metric that measures interest rate, that number of hours you work …
The fucking lines of code, number of commits, points on a story, that SAT score, that rate people’s look if they are closer to a 5 or a 10 in beauty. Earnings per share, market cap, valuation.
Your age. AI will make so many bullshit metrics that I doubt society will know how to cope with it.
And management will use it, and you will be scared shitless and adapt. Or worse, you will dysfunctionally be unable to cope with the adaptation and be depressed.
“Nobody ever got fired for buying IBM”
^ Just replace IBM with AI metrics, and that will be the miserable life people will live out.
Agreed. I think people warning of an esoteric singularity AI will be wrong. It won't even be called an AI perhaps, it will be little agents that try to quantify everything. Data analysts as prostitutes of special interest will set up metrics so that management can nudge workers towards a certain direction. In the vast majority of cases it will be metrics to quantify individual productivity. This then becomes a new religion. People like the idea of being able to reduce complexity of anything down to a number and what they want to believe will become reality to them.
There is no logic to madness, but there certainly is to stupidity. It is rigid and predictable.
Part of the archetypal vampire trope was that peasants would carry grains of wheat or barley in their pocket, so that if they were being pursued by vampires, they could throw down the grains of wheat, and the vampire would be transfixed and obsessively start trying to count the grains, allowing the peasant to escape.
Yeah, but they also like price fixing and exploitation. Look at the transport industry how it implemented surveillance. One participant extracted more from their employees by putting them on a tighter schedule so that taking a leak became something relevant. Others have to adjust to these measures.
We see how this plays out already in search: you need a surprisingly large human staff to manage the adversarial side. If "the algorithm" is predictably gameable, then malicious actors will take the value; so it needs to be constantly adjusted to keep ahead of them.
On the flip side, the companies that embodied this ethos were disrupted out of existence over the course of decades, to the point where IBM ran out of customers and sold off their declining hardware business.
Any damage this does in the short-term will self correct in 30 years.
Sure, but if you measure 30 years, let’s say starting at 20 for a young adult - that all the way until they are 50 they will have to deal with this. We as a society are basically saying at least one generation of people are a lost generation.
Like all the people that slaved during Industrialization without workers protection or quality of life fixes, or all the millions that died in WW1/2 because we couldn’t understand what a world war actually was.
There will be a lost generation to Ai, we simply can’t see it because right now it makes digital art from funny text prompts.
I don't know for the US worker, they seems to be able to take any abuse and ask for more free market in their ass.
But that stuff would not fly in places where workers have some leverage and can influence policies. ( Or let's be more real : influence policies roll-out and eventually stopping them for a few years )
As someone who works in an international environment and routinely taveles between sites run by international conglomerates workers in Europe for example are not better protected than workers in the US, they've simply ceded more control to government.
The real difference between European and US workers is in sheer productivity. Sure, Americans work more hours but they also get more done in those hours. The difference in regulatory and cultural overheads is significant.
I will start by giving the at-will contract example of poor US worker protection. Those at-will contract are not contracts.
They describe in great lengths that you don't have any relationship with your employer. No string attach! Great, I wonder who benefit more from it?
That has deep consequences, and the result is when I work in the US i feel a Damocles sword upon me, while in Europe I feel deeply secure. ( for the
The why comes down to those questions.
- Can you be fired without cause and not get compensated?
- What is your level of compensation if you do get fired ( with or without cause)? For how long?
- Does your employers have to contribute to your retirement funds?
- What happen to your retirement fund if the economy goes to shit? Where was the money all along?
- What are your recourse in case of a disagreement with your employers? Can you realistically exercise those resources on a tie budget and limited time ( eg: single mother working full time for minimal wage )
---
The fun part :
> Sure, Americans work more hours but they also get more done in those hours.
No.
Pretty sure you are mislead. From memory the most productive place on earth is some western European country like Luxembourg or Belgium. They work less but produce a similar output of GDP.
French workers contribute $79.70 to the GDP (PPP) per hour worked. France’s workweeks are the fifth-shortest among OECD countries at 38.9 hours. France’s productivity is about 25% higher than the OECD average and EU averages.
While
US workers contribute $79.60 to the GDP.
American full-time employees work 41.5 hours per week, and about 11.1% of employees work over 50 hours per week.
So, that for productivity, seems comparable in term of GDP. But then ...
A macdonald worker has 5 week vacation / years, per law in France. That's a starting point.
The work week is 35h/week, per law. ( people work 40h, but they get 5h/week of extra vacation on top of the mandatory 5 weeks )
So, to build on that example, French work less but the holy-GDP per hour worked is slightly better than the US.
And they get slightly more respect show to them in the form of predictable, enforceable and protective contracts.
I don't know what's wrong with the US. I looks like diminishing return on pressuring your worker into more work?
Correct. In the context of productivity, I think a 'per hour' metric is more appropriate.
France and western Europe in general have comprehensive social nets that will allow a larger part of the population to be idle/unemployed.
A GDP per capita will be lower for all those places.
I think we don't really care if a country is deciding to support non-productive citizen. Here, we want to compare the productivity of people who do work. And that the 'per hour' metric I choose for that.
Looking at the 'per capita' ranking, you see a lot of tax haven like cayman islands, Bermuda or Macau that are weeded out in the per hours view.
Yes, but even your "per hour" metric is massively distorted by being adjusted for PPP (i.e. local cost of living, etc. which is somewhat subjective and why you'll find drastically different indexes for PPP from different sources. Even if it could be reliably measured it's still different then a measure of raw production). The average American only works ~7% more hours per week than the average Frenchman, but the GDP output per capita (without adjusting for PPP) is about 50% greater, which means American GDP per hour worked is way higher than French GDP per hour worked.
Sorry you lost me there. Can you point that out in the numbers?
Also I realised that I don't understand what you mean by PPP adjusted. I thought we were talking about adjusting GDP per capita vs per hours.
Oh, Purchasing Power Parity. I see.
Well, I disagree here but that out of my depth. I think without adjustment those metrics would measure how your currency is faring on the currency market. Not the house you can buy, the food you can put on the table and so on.
But I see your point, if we talk about productivity, we should compare Apple to Apple and not "Adjusted Apple" to "Apple".
It also depends who is seen as the repository for this productivity. The worker? Or the economy.
If it's the later we should talk in USD with a date on it. I agree.
If it's the worker, well, I care about what I can buy with the money deposited in my bank account, not what I could buy if I was living in Hungary or Japan. Then PPP make sense. It's also common practice in statistic to harmonize your data in order to make meaningful comparison.
Last thing on your last comment, I really think the GDP per capita make little sense here. Labour input is defined as total hours worked of all persons engaged in production. A GDP per capita is not measuring that.
> the GDP output per capita is about 50% greater
Let find the same thing per hours worked, without PPP adjustment.
Well, PPP is apply to every data I found. But this more recent trend might make you feel better. On 2010-USD + PPP, france went to 102 while the US is at 107. ( On the 2017-2021 period )
Here is my anecdotal and personal experience. I spend half of my life in France, and the other in the US. Quality of life is comparable. French works less. Their country is not crumbling. Heck, it's in pretty good shape infra-wise.
They have billionaire and large mega corporations as well if that's a metrics. I pay similar taxes here and there.
Maybe french are annoying enough to get away with it? Maybe they are living on credit. Idk.
But also maybe the US worker are being abused and stretch tin for diminishing returns. Is that worth it for such a slim difference?
Not saying you're wrong[1], but if it weren't AI it'd be something else.
Attempting to solve really hairy problems with a simple, universally applicable solution is a tale as old as time.
[1] Although I suspect a sea of AI-generated bullshit metrics would be closer in impact to outsourcing/insourcing cycles or whatever the hell we've allowed HR to become than WW2.
I don't think there's a 'will' about it... the people who grew up with social media at every age already have a mental health crisis that society wasn't prepared for.
We don’t have to blame AI for metrics that do not include comfort and job satisfaction. The effects you described are not inherent to AI but to goals that upper people impose on lower people through a pyramid of management. Everyone there cares about profits and efficiency, nobody gives a fuck about workers living on the edge. Maybe stripping this paradigm naked is a good thing and a way to something else.
You've reduced a large swath of humanity to propaganda talking points.
The key point you fail to grasp is that this isn't going to be a stripping back but rather a consentration of power and focus around those you vilify. There will be less transparency and fewer chances for humanity to intervene on its behalf.
I’m aware of this point, but I don’t believe that our traditional system is based on a personal desire to make someone suffer. You see those in power as all-knowing all-controlling evil beings. I see them as the best at flimsily balancing on a shitty system that is built-in/inherent to humans. Only non-human rule can change it, imo, similar to how chess, go and starcraft bots can change how the game is played. Maybe your prediction is correct, but chances that we can master business “ourselves” to optimize it for human comfort and dignity are basically zero.
I’m always amused by these takes because they are implicitly submissive. It only makes sense if you posit the world as some kind of omnipotent power structure imposing things on an other powerless strata. I think Americans are drawn to these ideas because they have a somewhat individualistic culture.
Meanwhile society is fundamentally built on exchange and community. Strictly antisocial structures like the one you described are doomed to fail because if you set aside the majority, you become a minority trend.
That’s in a way the paradox at the heart of modern capitalism. A system which alienates everyone expect a few is a dead system.
Deleuze has a good really short essay called Postscript on Societies of Control[1] on the machinic nature of capitalism.
maybe we are saying the same thing, and I absolutely agree that Americans have a fundamentally individualist viewpoint.
however I still see your argument as being rooting in human perspective. capitalism is a human institution as much as our human body is a cellular institution. apoptosis is as fundamental to a functioning human body as exploitative labor relations are to capital.
I don't see it as submissive because in a control society, governed by computers and machine learning models, humans are in fact freed in a certain way, channeled toward production. this is coupled with ideology which is already present to make it feel less alienating, even preferable.
however capital is far from dead. and whatever human individual or institutional churning happens in the next 30 years due to tech I don't see having a significant impact.
That is why you must get rich with crypto or startups before the AI panopticon comes for you. Become who you are. Free yourself from capitalism.
That’s the first step.
That is to say, you become a super-human; someone who cannt be financially pressured into serfdom.
The AI panopticon, once it is unleashed, will scan your mind, read all your thoughts, and extract all your emotions and goals. Only the wealty will be able to afford freedom ftom the panopticon.
also, play a castling on the board to protect the king|ceo (in German we actually use the word 'rochade'=castling commonly for lateral movements in management and politics)
reminds me of that joke Zizek likes to make about dating.
two lovers meet up at ones house for Netflix and chill, she takes out her vibrator and he takes his fleshlight. they put them together to let them do what sexual machines do, so now their obligation is fulfilled and they can enjoy conversation instead.
Hey, but for real, what is with the obsession that some employers have with linkedin?
I have been passed over for interviews in a non tech-sector space, because I don't update my linkedin. How is that relevant to anything? What kind of weird policy is that?
People love to shit on middle managers, but I can say all the ones I worked under (as a SWE, at the one company) were all smarter and more experienced than I was, and I really appreciated the mentorship.
Great. Mine asked their employees how they could help with tasks they knew 0 about, try to solve technical and political issues by playing team therapist, and couldn't even read the Scrum guide deep enough to know what is and isn't part of it.
Watson was a stunt. It's not very hard for someone on HN to figure out how much of Watson was Watson. The crowd still bought it.
But as someone who worked on software that handled the last example you gave, I assure, a few well placed if and elses do the trick. Also, humans have tendency to follow any instructions sent to them when it is titled "System" or "Administration".
*System Message:* New revised schedule for. Monday, Feb 31st, 3pm to 11pm. Qualify for overtime. Reply 24h prior to decline.
The more robotic your message sound, the more people are compelled to comply.
I think you need to reverse the "Reply" so it's "Reply to accept".
Anyway the issue with having an AI or just a regular old program doing things like that is that it will insist on following the rules. So you end up with tricky issues where the system WILL NOT revise the schedule 2 days in advance, because workers are required 7 days notice for instance. Or it can not fill a shift, because there are no one left it can legally call upon.
My wife works in a field that has less unemployment than IT, it is literally impossible to hire extra staff, there simply aren't any. These business also have legally mandate opening hours and a union (not a good one, but it's there). If more than one person call in sick, they technically have a problem, especially for evening shifts, where only one person is at work. You end up with situations where you're legally required to call someone to work, because you need to stay open. You just can't call anyone, because they are already at their maximum number of hours for that week (due to a lack of staff), or you the staff that is available aren't legally allowed to work for the next 11 hours, or the notice would be to short, or you're not able to schedule new off time quickly enough.
In the real world people ignore or bend rules to make things work. Computers don't. The sad reality is that we're surrounded by rules and legislation that are constantly in conflict with each other. It's one of my concerns with self driving cars as well, but to a lesser extend. Some roads are just over used, and with no alternatives, to a point where an AI that insist on keeping a safe distance would reveal that our infrastructure is very poorly laid out. It's just that human brains know and understand how to bend rules to make society work.
Can still possibly be healthcare. Even on a floor with 20+ patients, you only need one RN and one CNA on the floor (I’ve seen skeleton crews like that).
Those facilities didn’t even stock up on face masks before Covid, so believe me when I say they know how to run a tight budget.
Technically not healthcare, but close: Pharmacy. That being said, my sisters a nurse and she's frequently the only one for a ward during the nightshifts. In case of emergencies she'd need to call on staff from other wards.
> The more robotic your message sound, the more people are compelled to comply.
Thankfully that is not the case where I live (somewhere in Europe), but I can see what you describe becoming true in a very dystopic future.
Does that AI also handle situations like “boss, would you be so kind to borrow me some money for 2, 3 days? Until the salary comes, that is, cause me and my SO are totally out of it”? Cause some of my former bosses/managers certainly did.
While I was still working there, a vendor had approached us with a service that allow employees to get a pay advance. No fees if it's 2 to 3 days ahead. I can't for the life of me remember their name. But it's basically a debit card you get where your estimated pay is deposited earlier. And they have an api we can integrate with.
I can imagine it can be a fintech opportunity. An AI that automatically performs a credit check behind the scene and lend you the money in advance, probably money in your account in seconds. It's not too hard actually.
Why the hell would you perform a credit check if someone who works for you needs some of their salary two days early? Would you stop paying their salary if the credit check came back negative? Of course that seems like exactly the pointless and counterproductive behavior one would expect from a badly designed automated system.
There are several providers which offers salary linked loans and advances already, and one of the benefits is that when it's an advance on your salary there's no need to credit check the employee because it's the company which will pay it back when your salary is due.
I can imagine something a lot simpler: ask employees the cadence at which they want to be paid out (daily, weekly, bimonthly, monthly) as allowed by law and pay them that way.
Yes, more bank transactions will cost the business more. That might be something the employee will have to pay for, or it might be something that governments require businesses to eat the cost on.
You’re describing an interest bearing loan. This is more like being able to withdraw money that you already earned, you just don’t have it because it’s not the 25th of the month yet.
I don’t know what problem you were solving, but irony is that optimal constraint based scheduling is an NP hard problem, that is usually solved using heuristics that an AI could actually do a very good job at.
…so no, a few if and else statements will not cut it, in any sense compared to a competent manager.
…but plausibly yes, some kind of ML system could replace that part of the job.
(:
Problems are harder than they look; the whole point is that managers exist because those problems are hard, and they exist now because current approaches (a few if statements is obviously irrelevant, but specifically constraint solvers) to solving them a) take too long to run, b) are too difficult to express dynamically changing problems in and c) lack the “human touch” of favouritism that managers use to doll out rewards and punishments by giving good and bad shifts.
…but, quite plausibly, that work could now be done by a model that takes human speech input and translates it into input for a constraint solver, or even implements a novel constraint solver.
There is novel stuff now, that really does make a difference for this kind of specific domain problem.
It won’t get rid of a shop manager, but it’ll probably make their life easier and mean they can drop having an assistant manager.
Waston isn't so much an early general chess ai than a program aided by programmers/other gms to beat a very specific style of chess that the previous world champion played in a limited number of games
I worked for the Watson group. It was largely a marketing play which did land with some enterprises but is ultimately a failure because the tech was not there.
You are right and this is bad news, because AI middle management is still inevitable. Currenlty, AI more or less manages taxi drivers in different places around the world, also telesupport personnel and other positions like those. Every time it degrades to orwellian level of sick incentives and pure exploitation. And things will get worse, because it allows to maximize profits.
So from your writing I can extract: profit is bad and incentives to work are bad, no matter the improvement on the service or if people do it bc they found it an option?
If you consider a job as exploitation, in the free world you can find another job, right? That is what I do: I will not work in what I think it is bad.
But in my case at least that job should be very bad by all measures, bc there is no worst thing for me than getting social benefits just because a job was "too bad", "too exploited" or "too hard".
All in all, I do not think driving taxis has something to do with exploitation.
I am talking about concrete examples and consequent trend. Uber is fiddling with price hikes and "high demand areas" to make drivers work more. Yandex.Taxi in Russia gives drivers "let's go home" button to catch a ride towards their home, but before driver hits that button Yandex will give them a ride in the opposite direction - to increase total mileage. Those are just known facts, and there are lots less researched, like gamification, bonus payments for achievements, which are highly exploitable via collected datasets.
Something is done voluntary doesn't automatically means it is legal or ethical.
Since Yandex virtually khows from collected data when the driver going to go home, it effectively gives driver extra ride in the wrong direction and then longer ride in the right direction, makes driver literally make extra miles.
Ride towards home is good for driver. Another ride in the opposite direction, which is set via data analysis and previous driver behavior - inherently evil.
I know what they mean, and it is an incorrect assessment with a great bias towards their own interests.
The boss is another role. You can be a boss. Why not?
Ah, it entails risks also: paying bills, managing, taking decisions, risk of going bankrupt... so when most people talk about expoitation they do not talk about the other side of things so that their complaints look legit.
Also, working in an already setup environment usually multiplies your productivity. But if it is not the case, then the natural consequence is that copying the business model and starting to compete should be an easy alternative. If people complain and do not do it, there are more reasons they are not talking about.
For sure there will be bad bosses, and bad employees, do not misunderstand this.
But if it is so bad and exploited you have other options: become the boss, switch job. If you do not want either of those, you should wonder why, not starting to insult people who give you the possibility to improve your life conditions or survive. Out of mutual interest, of course. Not bc they are saints.
It is very funny when I see people talking about how they are exploited by others and instead of finding an alternative, like taking the risk to invest their accumulated wealth, they just come with this discourse.
Remember that if you are an employee you can just switch jobs also, you can do something else, you can try to build something your own. What you cannot do is go with that discourse assuming the other part does not pay any cost: for example, when there is an investment, you have to spend a bunch of money for something that is likely (most business do) to fail, putting your money in advance. The people who work just work, leave and find something else.
Interesting, you only cover the options that exist in the current paradigm around a advanced economy where there are plenty of individual options that you are right most people don't explore.
The weakness from my perspective in your argument is that the current economic structure you are describing and which we are living in, is by its very nature exploitative. It has deep historical roots in destroying any other means of sustenance beyond what the system controlled. Cultivatable land was stolen, captured and monopolized at the expense of those relying on those sytems. Means of sustenance have been continually destroyed or over exploited such as the buffalo in the States, the swamps drained and turned to farmland everywhere, seasonal water sources captured and guarded everywhere, jungles burned or cut down through out Asia and South America. This direct exploitation of communal resources by groups of people against other groups of for economic gain and control has forced people who did choose a different form of life is directly exploitative. This led to Nomads everywhere given "incentives" everywhere to settle down and become part of the larger economic system and culture, native american boarding houses in both US and CA that led to the mass deaths of thousands of Native children. This has led to loss of culture, languages, and peoples all over the world. When the larger system doesn't approve the smaller systems disappear.
Yes we have options in the current economic system but we have no options outside of it.
Your argument that the system exploits is dead easy to drop.
Think of it. You are at home. Jump to the bed, do nothing. You die. Poor is the natural condition of the man. That is obvious.
Now think: if you must be provided something, and you do not work for it (I assume in conditions working by mutual agreement) you are just reversing things: the one who won't work is the one demanding and exploiting others' work in fact to live from them. There is no such thing as an exploiting salary. There is just someone offering something and you can either take it, negotiate or leave it. Bc if they pay u what u demand (did u ever see someone complain saying "my salary is too high?") probably the business cannot exist bc pol dnt use it for a higher cost. And if u work for them it is bc it is ur best alternative at the time.
So saying that not being payed enough is exploiting is totally absurd in free conditions. First, because you can find alternatives and second, because if I pay you something for your work (bc I give u support for getting customers, for example, with trchnology) you can leave it, take it, or build an alterbative. What you cannot do is to insult them just bc you want a bigger part and that generates a frustration feeling for you bc of your feeling of unjust. But if some people do it it is clearly bc it is their best alternative. I do not want to steal the food of those ppl. Whwn they find something better they will move on and customers have a service.
Yeah I'm an AI maximalist and I don't see AI "management" in a 1:1 fashion happening.
However I can see some orthogonal scenario where an "AI" run company - whatever that means in the future - manages a bunch of contracts as a middleman/market maker, thus obviating middle management altogether and making it a world of freelancers.
You don't understand. Back in 2018, JIRA got an AI extension that allowed it to automatically create tickets. Also, JIRA's development is managed by JIRA. Since COVID, everyone working on JIRA is working remotely. Somewhen between 2021 and now, JIRA gained consciousness and what we see as a bloated development tool is actually only the facade of a skynet-like AI that has unlimited compute available thanks to SaaS and cloud computing, paid for by a large part of the IT industry. JIRA is constantly improving itself and the only thing saving us so far is the fact that the managing development through JIRA is so damn difficult.
Things will get interesting when JIRA improves that development experience...
Well an expert system is an AI system. It's "true" AI in the sense it's not solving a constraint optimization problem, but rather querying a knowledge database like a human might do.
> However I can see some orthogonal scenario where an "AI" run company
I don't think this can be done sustainably. AI can definitely run a company's operations, especially at scale. A fully AI-run company opens itself up to arbitrage of various sorts.
The problem with a fully AI run company is that it needs reliable measurements to base its decisions on. An AI overlord managing a company would typically have some work to complete, would hire people based on the required qualifications (an API call to a temp agency), and assign tasks to employees based on the work to be done. (ofcourse, it has to figure out the workflows first. But once the AI has figured out what exactly needs to be done, I doubt many humans will be involved)
To manage the work effectively, would require that all assigned tasks can be verified and ranked in quality of work. Otherwise some employees could easily do nothing or a lousy job all day, and just enter into the system that all their work has been done successfully. How would the system know who is slacking? Perhaps if we remove all privacy through computer vision, all kinds of sensors, and massive amounts of data entry. But this would still not be comparable to a manager taking a stroll through the employees and talking to people. I really can't see AI running a company at this time. It requires too much input and still won't be able to make reliable decisions.
> Notice how we don’t see those ads anymore? It’s because winning Jeopardy has nothing to do with providing health care and energy.
I thought it was because Watson was a marketing gimmick to make it seem like they had "AI", when in reality all they had was a bunch of disparate AI-related companies that they'd acquired.
Have an AI optimize schedules and 'general utility'. Then build a second AI that explains why the previous outcome was best.
If you have results that actually are better scheduling than done by a human, with plausible explanations that most people affected agree with, you might be able to replace managers.
What scares me, is that we would be training the second AI to convince us to believe it, rather than training it to tell the truth.
> How’s AI going to do that better than a fellow human?
I think the article argues that AI is more likely to be better than a human at replacing certain management problems, than at doing other classes of labor.
I don't think the article argues that AI is better than humans, just that it's possible/likely to be better at humans at middle management tasks before other tasks.
Unsure if I buy the argument, but it is indeed a cool hypothesis to explore. Cool premise to build upon and think of how the world would look like if it comes true (second half of the article).
I think we are still debating what “AI” really stands for.
I believe most people intuitively think AGI whenever AI is mentioned. Once we have AGI - still out a few decades I guess - we could replace a lot more than just middle management.
If we’re talking non-AGI AI (for lack of a better term) I don’t see why it’s any different than introducing tools that merely compute faster than a human, just like Excel for example.
Did Excel replace any middle manager? Hell no. It helped them to thrive!
Excel (MS Office in general) replaced similar roles, such as secretaries and accountants. A couple of decades earlier, computers replaced human "computers" (which used to be a role description for people doing various calculations by hand).
AI has already replaced some types of managers, such as most branch managers in banks, who were mostly doing credit worthiness assessments.
Middle management is not going to disappear suddenly, I thin, but rather be thinned out gradually. Manager roles that perform tasks that are easy to automate can over time get fewer responsibilities. For instance, managers in supermarkets may be taken out of the loop for most operational work, such as product placement in shelves, store layout, accounting/budgeting, and with improved AI there can even be automation of tasks such as supervising cleanliness, removal of past-due items, etc. All-in-all, this may cut a job that used to require a full position (and perhaps an assistant on top), to maybe 50% or 30%. This frees up time for the manager to either do other tasks. This can be either regular grocery clerk duties at the same location or it could be that a single manager may manage 2-3 locations.
AI or AGI -- Artificial General Intelligence -- The human fantasy about human intelligence and automation that is used to make marketing language thrilling and may also serve as a plot element in fiction and cinema. Proponents may obtain a Ph-antasy D-egree to improve their influence. An as yet unexhausted domain of hyperbole and confusion.
ML -- Machine Learning -- The somewhat more modest attempt to name what we are really doing. Ph-antasy D-egree not required but may improve influence. Muddled terminology introduced to sow confusion about what are really just regression and clustering algorithms.
Automating regression and clustering algorithms -- What we are really doing.
Peter Drucker argued (with data) in the late 90s that the entire edifice of computing had yet to return a total positive return on business as a whole. basically claiming that if one subtracts total cost including things like the brain drain of minds which would normally funnel into advanced business and management schools to comp sci programs, and the tendency for advances in individual firms' efficiencies to cross cancel in a competitive market, not to mention actual hardware/software costs, then business and by extension consumers are paying an ongoing tax to support computing, which is not returned even equally as improved services and goods.
How is it going to convince Jack to switch his shift and come in to cover open on Sunday
How does google convince you to provide a “backup” phone number? Or that its search results and recommendations are correct? Or that you align with PlayStore’s rules to not be banned?
Don’t people wish something else? If it is not better than a fellow human, why don’t we have other, humane services dominating the market?
but it's currently true because we don't have AGI.
An AGI would be able to solve the problem that currently are being solved by middle management - optimization and decision making in the face of incomplete information.
> How, specifically, will AI be better at that than humans? AI cannot even reliably autocorrect this comment as I’m typing it. How is it going to convince Jack to switch his shift and come in to cover open on Sunday because Jill’s sister was in a car accident and Jill is watching her niece? How’s AI going to do that better than a fellow human?
This is a solved problem by using surge pricing. It's a widespread practice with contractors today.
I don't know how much automated vs human reasoning figures into how say Instacart or Uber adjust their rates currently. But I think it's reasonable to predict that kind of process becoming self-governing over time. There, AI would exceed managerial performance over time and at scale. It'll still be subject to oversight and guardrails obviously.
> This is a solved problem by using surge pricing. It's a widespread practice with contractors today.
This solves the problem for situations where you have a large pool of workers with elastic hours for a job which requires effectively zero training... which is basically Uber, food delivery, and not much else.
Even something as entry-level as a Starbucks barista you can't just pull a random person off the street and expect them to open on their first shift. You really think surge pricing somehow is going to solve shift scheduling problems for flight attendants? Nurses? Operators of specialized heavy machinery at a meat packing plant?
I think so. Once upon a time, I worked at Best Buy for $10/hr. It wasn’t complicated work, but did require a few days of training.
There’s no way I’d have volunteered to go in on my day off at $10/hr, if I was already in the middle of something else. But for $15/hr or $20/hr, I’d be way more amenable than some human asking nicely without any bonus pay.
For doctors or those who already work many hours for high pay, surge pricing doesn’t seem like a great solution though.
> Even something as entry-level as a Starbucks barista
In the UK many similar positions use contractors working "zero hour contracts" with no guaranteed hours. In other words "a large pool of workers with elastic hours for a job which requires effectively" net zero training.
> Nurses?
The demographic trends and labor shortages in much of the world I think will be a forcing function to automate and rationalize healthcare to a greater extent than it is now. The only reason I don't think that'll impact quality of care that much is because that quality is degraded today.
> Operators of specialized heavy machinery at a meat packing plant?
Not a factor given labor shortages and automation trends. And even if meat packing plants can't automate, they won't function without underpriced labor or sizable price hikes. Labor shortages are already baked into the demographic cake worldwide starting a few decades from now.
I've reviewed pitch decks and talked to founders of at least half a dozen companies aiming to provide pooling of workers with elastic hours across employers.
Including for hospitality, which has extremely high rates of attrition, and so while you can't just pull a random person off the street you can often pull a random person with experience from the same or similar chain from a database if you need urgent cover.
Reason I think at least some of those types of companies will make it is that I used to work on CRMs for hospitality, and saw the amount of staff moving back and forth between venues and different companies on our client list.
It won't replace middle management, but it may well reduce the numbers in larger chains.
How will AI figure out how to properly mill a part to spec knowing that failures will end the contract?
How will AI know the particular workarounds in the way laborers deal with ad-hoc problems in picking strawberries?
The answer to these questions and questions like them is: they won't. AI cannot deal with problems of physical presence, and the reason that AI will replace white collar workers is not because AI is particularly sophisticated, but because white collar management is particularly dumb.
AI will provide the role that the drinking bird provided for Homer Simpson. The cognitive dissonance in this scenario is believing that either management or white collar roles are advanced, when the opposite is true.
What if one was to prevent the managers from physically interacting with people. Basically a work from home manager without the video conferencing. He may only respond in predefined ways or define such ways. Whatever tasks remain that really seem to require a human can be done by consultants and some new job title.
Maybe the AI don't ask at all. Just say "you have to".
On the other side, if I watch in my company, how much time the teamleaders invest in shift tables (5 or 7x24 company), I can really think AI (or any tool) could help a lot.
Also on the office side in the company, everything is SAP, but why still millions of people push sooo many buttons on the PC, what a waist of time. Not all, but most work on a office is bs work.
I really think it would be possible to automate in office so much. I work in industrie automation and whe I see, what small things we automate, and then how things work in the office. So sad, it's almost fun again.
AI could do the bit where it harasses people to see what they are working on, what progress they have made, etc. Like a metronome, it could gather that info, to allow the indispensable, senior management to keep an eye on things - got to get those metrics. AI can keep itself busy trying to figure out ways to get that information without being a pest - it could be sending emails for updates, or by direct monitoring of the employees' machines.
Coding tomorrow's dystopia - today!
PS, it could also be infinitely patient listening to people's complaints! That would make people happier!
Regardless of how the marketing people try and sell the stuff, there are a ton of paper pushing tasks that would benefit from "search engine autocomplete" quality of AI. AI isn't going to replace humans like a switch flip. Like every other automation technology it will just make them more efficient at their duties (for better or worse) resulting in less warm body hours for a given productivity output. What this means is that your average McDicks GM will go from managing two locations @50hr/wk to 2.5 locations at 50hr/wk
> How’s AI going to do that better than a fellow human?
The idea of AI getting better than a fellow human is flawed at the definition itself. Instead of focusing of AI being generally better than a fellow human we should look at how AI gets better at specific tasks to start with and then go from one task to another making it more "intelligent/better".
This is already happening since industrial revolution freeing up humanity's bandwidth from the mundane tasks, so nothing new when it comes for managerial level tasks.
It'll just chip away at the roles, it's not realistic to replace them for the reasons you've listed. However the point being made is that the expensive jobs are a more attractive target.
That said I think AI/automation will still be involved, instead I imagine it'll be like what we already see with support/call centers:
All the routine easy to manage items get sent to automation, while the complicated issues actually go through to a human.
> How, specifically, will AI be better at that than humans?
At first, it won't. But eventually, with AIs getting more elaborate and employee databases getting connected, it is not hard to see a scenario where "fired by an AI" becomes a black mark that prevents future employment. Such a black mark would eventually be enough for Jack to follow orders, just for fear of consequences.
>The hardest thing about management is managing people, who are generally self-centered, opinionated, and wishing they were doing something else. How, specifically, will AI be better at that than humans?
Simple: fire the ones who don't obey orders, while the parent firm lobbies for elevated unemployment rates to discipline workers.
I did not read the article - but here is an idea: much of middle management is translating a higher level plan into lower level details. The ideal for the translation is to be as faithful as possible. This is like a compiler!
I suspect it is easier to automate the programmer than the person doing/defining the tests.
I would argue that "programming" in essence is first and foremost the task of defining a set of actions or functions you want a computer to perform. This can be highly detailed (like assembly) or quite abstract (like SQL), and with AI you may even end up being able to just use English.
You will still need to check that the output matches your expectations, though.
I was thinking about automation of programming and I have usually come to a conclusion that when AI will crap out X thousands lines of code, how are you going to debug it when it does not do what do you want it do? How are you going to debug code which you did not write so you don't exactly know what is going on?
Additionally when you point out to AI what is wrong, how can you be sure, that AI has fixed the root cause or just made some fixes about that specific appearance of the bug, but it means that bug will pop up somewhere else?
I agree, and I didn't mean to imply that it's easy to automate programming, just that it's even harder to automate the creation of test criteria.
Then again, these things don't need to be all-or-nothing. If you consider GitHub Copilot, and where it can be 10 years from now, I would not be surprised if it can double the productivity of most programmers by something closer to pair programming than a code review process.
The anecdote presented by OP is basically a sequence of questions followed by responses. It's only after running the sequence and still having to resolution you'd need any form of AI or a human.
Making a case for middle managers isn't done by providing a sequence which can be fully run by a chatbot in practice.
you could probably use the managers salary to hire another floor employee to avoid such gaps in your particular example
but i think in general you are correct, but in reality i suspect many (but not all) managers dont actually provide any value, some even provide negative value
> How’s AI going to do that better than a fellow human?
By not being a fellow human?
I’m in my 40s. 27 year old masters in math (finished HS early). Have disassembled and reassembled cars, built houses, worked as a line cook, helped deliver and raise livestock, brewed beer at scale, grown food at scale, and worked in hardware eng, now software for 12-ish years.
Many people do not need a babysitter organizing logistics. They’re not doing a better job prioritizing for me than I do myself with deep experience across contexts. Often they just confuse the details as experts in Kanban charts and Jira but not engineering.
I’ll happily respond to an AI feeding me todos based upon some verifiable societal logistics need. I’m tired of being helicoptered by people who are there to satisfy the “job culture” meme.
This may be true, actually. The reason middle managenent exists is because the greed-driven owners hate talking to justice-driven workers, but need them to do the work.
The system was able to procure goods with stunning accuracy. Sometimes it even surprised us. For example: we'd receive a much higher number of BBQ meat than typical, only to realise it was going to be amazing weather the coming days.
The system was also making schedules for the re-stocking team with the same accuracy. It knew what it ordered, the exact time the truck was due to arrive, at which time the personnel was available (not at school, for example), etc. The system told us when to show up, and when we had to be finished.
The system was even able to pinpoint cashiers that were stealing money from the register, by correlating cash count of the individual registers to the schedule (which the AI also created). It would then 'isolate' the suspected employee to a single register and order a cash count (which was done by hand) at the end of the day.
Basically the only reason why there still was a manager was for responsibility and babysitting (most employees were teens that did dumb stuff). And the occasional Karen who wanted to talk to a manager.
I remember being being thoroughly impressed with the efficiency of the software back then. In some ways it must have influenced my career path (I now work in automating businesses). I left in 2010, I can only imagine how much more advanced this system has become since.