Humans Still Cheaper Than AI in Vast Majority of Jobs, MIT Finds (bloomberg.com) 47
AI can't replace the majority of jobs right now in cost-effective ways, the Massachusetts Institute of Technology found in a study that sought to address fears about AI replacing humans in a swath of industries. From a report: In one of the first in-depth probes of the viability of AI displacing labor, researchers modeled the cost attractiveness of automating various tasks in the US, concentrating on jobs where computer vision was employed -- for instance, teachers and property appraisers. They found only 23% of workers, measured in terms of dollar wages, could be effectively supplanted. In other cases, because AI-assisted visual recognition is expensive to install and operate, humans did the job more economically. [...] The cost-benefit ratio of computer vision is most favorable in segments like retail, transportation and warehousing, all areas where Walmart and Amazon are prominent. It's also feasible in the health-care context, MIT's paper said. A more aggressive AI rollout, especially via AI-as-a-service subscription offerings, could scale up other uses and make them more viable, the authors said.
Only 23% of workers? (Score:4, Insightful)
That shouldn't be hugely disruptive /s
Also study is only looking at LLMs (Score:2)
The real disruption will be the massive automation boom we're about to see, as big if not bigger as the one that caused the recessions in the 80s [businessinsider.com] only this time there won't be a
I think the study is not limited to LLMS (Score:2)
AI-assisted visual recognition[...] The cost-benefit ratio of computer vision is most favorable in segments like retail, transportation and warehousing,
I think if anything the most viable usage scenarios highlighted seem to be machine vision. Also, a bit of "old news" since those technologies were aggressively being deployed before LLM fad set in.
Instead there's just more automation, few jobs.
Should the day come where that holds true, it'll either be disastrous or fantastic, depending on how we collectively handle a reality where we don't have enough to "do" to keep folks busy. Sadly, the most believable hypotheticals I've seen have all been describing a "disaster" scenario.
Re: (Score:2)
The real disruption will be the massive automation boom we're about to see
Have you seen this [mercurynews.com]? It just shouts "replace me with a computer". My local Mickey D's already has those ordering computers, all the workers do these days is flipping burgers.
At $20/hr, computers will become even more attractive.
Re: Also study is only looking at LLMs (Score:1)
The problem with automation is that there are more repairs, especially with 1st and 2nd generation products, in the âoework fastâ crowd.
If Teslas âoefull automatic drivingâ actually worked, weâ(TM)d have exponentially less drivers on the road. It doesnâ(TM)t, so gas cars are still selling.
We have a local Caliburger, which uses a robot to make and bag fries. Mechanical arm takes them from the fridge, weighs and drops them in the fryer, then dumps them in a hopper, bagging them.
Re: (Score:3)
My local Mickey D's already has those ordering computers, all the workers do these days is flipping burgers.
I think the actual flipping was optimized out quite a while ago. Now they use the high-tech equivalent of George Foreman grills to cook both sides at the same time.
Re: (Score:2)
I would like to add that even with current technology, even without LLM, there are a lot of tasks that can be automated with traditional classifier AIs. Few actual implemented examples:
- Sorting cucumbers
- Sorting legos
- Counting fish
- Locating illegal slavery from satellite image
These are not perfect, but they are easy to make, low risk use cases and much more accurate and much more cheaper than humans.
Re:Only 23% of workers? (Score:5, Insightful)
The summary kind of alludes to it but somehow misses this key point.
Of course, whether these percentages mean anything at all is the bigger question. It's all massively uncertain. For one thing you never really know the cost nor the efficacy of something until you pour a lot of money into developing it. For another, the cost can shift dramatically with one breakthrough product.
Getting even an accurate time frame for the replacement of one specific job vision-intensive job - driving a car or taxi - is still elusive, despite billions invested in doing precisely that.
Re: Only 23% of workers? (Score:5, Interesting)
If any business really automated 20% of their labor... most of us on this site will wonder what government assistance that company had; to be so inefficient for even a year before going bankrupt.
From 2006-2010, I spent a good amount of my time reviewing, documenting, and walking through business processes in many types of businesses. And although I was fresh to the workforce, I was confident that 30-50% of the tasks being done can be automated!
But no one does. Now going on 13 more years doing similar stuff but exposed to a global environment, I see things automated in the US that isn't in the EU, and EU that isn't in S.Africa, India, China, etc... And yes some vice verse (ie: voting, healthcare, mobile payments).
What I learned is, the capability to automate plays very little in its applicability. Everything automated needs to be supported. Automation may replace 100 people and now plays a part in producing revenues of 50 of those people.
When that automation breaks down or needs to adapt, you need a set of people with skills valued at the cost of many of the people you fired. And you don't have the funds to do so because the market forces lowered the pricing you used to charge. You gained productivity that is expected of you with a lower unit cost and higher availability expectations.
This is what keeps automation from replacing "skilled workers"; a term that varies place to place. Humans are just highly adaptive and accommodating and innovative. All these traits are extremely expensive in automation. Additionally, at some point you must have an automation-to-human handoff/receive interface... and these are extremely expensive just to reach a compromise. Automation has always been easy, it's the details that are expensive and maintenance much more so.
AI is just a marketing term for automation. It seems to alleviate some of the upfront discovery costs and ongoing adaptability costs. But no one seems to be able to properly quantify the delivery costs. Right now they appear to be heavily subsidized to create "successful examples" with the assumption the cost of hardware and power will go down over time making for a good return. It almost feels like bitcoin. But no matter how low the cost of calculating 2+2 goes, the requests for 2+2 seem to be ever increasing; far outpacing the unit cost savings with unnecessary and wasted volume.
Re: (Score:2)
Good analysis. Cory Doctorow arrives to many of the same conclusions here:
https://locusmag.com/2023/12/c... [locusmag.com]
The Real Cost of AI (Score:2)
Re:The Real Cost of AI (Score:4)
Ugh, the prompt approval meeting was *humans only*. Now we have to throw out the training data and start over. The project is going to be delayed for weeks.
Re: (Score:2)
without accidentally starting an HR incident
Fewer humans mean fewer HR incidents.
The administrative and management overhead of dealing with humans is a big reason robots are being adopted.
Robots don't need HR, don't have sick days, don't get overtime pay, don't file complaints with the union, don't waste time gossiping, etc.
Re: (Score:2)
I'm not so sure about that. There are more than a few incidents where someone accidentally pointed their LLM at 4Chan and created a racist AI chatbot.
Re: (Score:2)
Microsoft should offer Tay as an AI personality.
Hey Tay, take this telemarketing call for me!
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Don't worry though, apparently, AI is still no match for human stupidity.
Re: (Score:2)
Wrong point (Score:5, Interesting)
Re: (Score:2)
The risk isn’t “AI” replaces workers wholesale, it’s that many workers have increased productivity using those tool sets and thus can do substantially more. This eliminates the need for as many workers faster than any expansion of new jobs created and thus displaces a percentage of workers.
This happens when the CEO is short-sighted (which most are). The better approach is to lower prices and expand the market, because the same company with the same workers can now produce X times (say 2-3x) more product than before, and hence able to sell at, say, 30-50% cheaper and still makes much more profit.
AI is a meaningless term (Score:2, Insightful)
sad future for humans (Score:1)
Already being used by shady cheap co's (Score:1)
The first cars were more expensive than horses. They were mostly a toy for the well-to-do who didn't mind tinkering. The first AI use will probably be for cheesy companies who already had lousy human help desks to begin with, and don't care about quality. As the bots get better, they'll slowly crawl up the price and quality ladder.
I already get telemarketer bots that try to pass off as human. They get a little better each year.
CEOs are too expensive (Score:5, Funny)
I think even a low-quality AI could accomplish 50% of the work of a CEO at 1% of the price. Seems like a really good value proposition to me.
AI sadly cannot replace politicians because the AI has no need for bribes or creative donations. It lacks the basic feature used to communicate in our democracy, a bank account.
Re: (Score:2)
Re: (Score:2)
This is an excellent point. Lots of people think they could do a better job at being a CEO, than the CEO they work for. But if they really knew how to do the job, they would be doing the job.
you're saying one of us could get Tim Cook's job? (Score:5, Informative)
This is an excellent point. Lots of people think they could do a better job at being a CEO, than the CEO they work for. But if they really knew how to do the job, they would be doing the job.
So simply by knowing how to do Tim Cook's job better, we could have his job? So no human being in the world is better than Tim Cook at being Apple's CEO? Now I actually think Tim Cook is a good CEO. However, Sundar Pichai is definitely not. You're saying there's not a single human being that could be a better CEO than Sundar...or if they exist, they already have the job of their dreams?
Your statement assumes leadership is based on merit. This is flawed for many obvious reasons. This isn't a sport that can be objectively measured, like running or powerlifting, or even a popular team sport where we can measure an overall record. There's no way to determine who is the most qualified CEO.
Secondly, people who lament about representation in leadership forget that leadership isn't doled out based on merit. The CEO of a company is usually a relative of the founder, like the Waltons. In Tech, it's sometimes a favored insider, like Tim Cook or Sundar Pinchai. If you're just a rando with exceptional skills, you're not going to get the job...no matter how qualified you are. There's no way to objectively determine if you're qualified and the boards are not giving an open casting call...they only pick their friends. It is a corrupt nepotistic and incestual practice.
So with all due respect, kindly fuck off with the synchphancy. Our overlords are not oppressing us because they are simply superior to us. They are oppressing society because of a mixture of luck and talent...in some unknown ratio, but I'd wager it favors luck over talent.
Spend some time listening to the business elite. They're fucking dumbasses. Have you ever seen Steve Balmer interviewed?..the fucker is dumber than a bag of hammers. I can't believe that man graduated from college, let alone went on to lead one of the most valuable companies in recorded history. Ever listen to Carly Fiorina?...she's almost as dumb as Balmer, she's just self-aware enough not to say as much...because the more she talks, the more you realize she's dumb, lazy, and in over her head. It was so funny that she thought her "business talents" could translate into politics. The voters heard her speak and were quite underwhelmed.
Founders?...OK, they're usually pretty sharp (what the exception of Larry Ellison)...the people that replace founders?...most of them are underwhelming, at best...in the case of Ballmer, a big fat ugly bimbo...Fiorina...someone who clearly had no clue what she was doing, but lacked the confidence of Steve Ballmer to be so open about her stupidity...or Pinchai?...well...I honestly don't know if he's smart or dumb, but I can tell he's a SHIT SHIT SHIT CEO and running Google into the ground.
Leadership is not a meritocracy. You get opportunities based on who you know and who you blow. So can I be a better CEO than these folks? I don't know...never thought about it, but I am confident nearly any good CEO could do a better job than Pinchai, Fiorina, or Ballmer.
It's fair game to criticize them. It's also an equally dumb standard to compare my specific talent to everyone else on the planet. Am I qualified to perform surgery?...no...but just because I can't give someone open heart surgery doesn't mean I don't have the right to think a doctor is a shit doctor when he kills his patient by sewing up the patient with medical clamps still in his chest. Just as I don't need to be a superior surgeon to know that killing your patient due to stupidity is bad, I don't need to be a better CEO to know that a CEO sucks at his or her job.
Re:you're saying one of us could get Tim Cook's jo (Score:5, Interesting)
You assumed a lot from my two sentences, most of which was not at all what I said.
I don't believe the CEO job is received based on merit. I believe that people who *know* how to be a CEO, get the job by *taking* the job, by making it happen for themselves. They don't wait for it to come to them, they go after it.
At heart, the job of a CEO is chief salesperson. They first have to sell themselves to the company they want to lead, or to its board, or to its investors, or some mixture of the above. A good salesperson knows how to get you to agree with them, to buy what they are selling. A CEO has to sell his ability to lead, and sell it well.
You don't have to be "smart" to be a CEO (in the traditional sense), but you do have to be ambitious.
Yes, it's fair game to criticize CEOs. They usually deserve all of that criticism. But they also perform (sales) tasks that few others can do. So again I say, if you can perform the sales tasks that a CEO has to perform, then you certainly *would* become a CEO, by whatever means you had at your disposal.
So they could become CEO of WalMart? (Score:2)
I don't believe the CEO job is received based on merit. I believe that people who *know* how to be a CEO, get the job by *taking* the job, by making it happen for themselves. They don't wait for it to come to them, they go after it.
At heart, the job of a CEO is chief salesperson. They first have to sell themselves to the company they want to lead, or to its board, or to its investors, or some mixture of the above. A good salesperson knows how to get you to agree with them, to buy what they are selling. A CEO has to sell his ability to lead, and sell it well.
So if someone was a good enough salesperson, they could become CEO of WalMart?...the Trump Foundation?...so every corporation where control is handed to a biological heir was because they didn't meet a good enough salesperson yet?
Sorry, gotta disagree. There's no open casting call for CEOs. There's no auditions. You cannot become the CEO of a chosen company by sheer talent...in sales or any other field. The CEO could have gotten the job because they were shrewd, but in most cases, they were simply l
Re: (Score:2)
So if someone was a good enough salesperson, they could become CEO of WalMart?...the Trump Foundation?
That's like suggesting that if someone is good enough to be a junior developer, they could become a CTO. No, certainly not. A CEO is, as I said, *chief* salesperson. There are a lot of steps between "salesperson" and "chief salesperson." And just as 99% of junior developers never become a CTO, 99% of salespeople never become a CTO.
I think you completely ignored what I posted, because your rebuttal simply repeats what I already agreed to.
There's no open casting call for CEOs
Yep, we agree. CEOs don't wait for a casting call, they make the job ha
Re: (Score:3)
Just because I can't personally can't do something doesn't mean I have no right to criticize. I am certainly no worse than a film critic or an art critic. But ultimately my complaint is not of individual CEOs but with the public's lack of trust in the institution of modern business itself and of the all too frequent behavior of those in C-suite positions.
I am in investor in a small business (of a dozen people, below what is considered small-cap) and in jointly held real estate. Yet I've never tried to vote
Re: (Score:1)
Just because I can't personally can't do something doesn't mean I have no right to criticize. .
Just because it is your opinion and a criticism which you are completely entitled to doesn't mean others aren't within their rights to call it out for the uneducated bullshit it is.
Re: CEOs are too expensive (Score:4, Insightful)
Inflammatory remarks and profanity are certainly not a constructive contribution to the discussion.
What about: Steve Ballmer...Carly Fiorina? (Score:2)
yeah everyone hates CEO's, the reality is though most of them work there guts out 7 days a week 12-18 hours a day. They tend to be highly ambitious and success driven. I doubt an AI is even close to being able to replace even a bad CEO at this point.
Hmm, I am not confident an AI can't do a better job than Steve Ballmer, Carly Fiorina...watch them speak...if they had drive ambition, intelligence, and talent, it drastically faded by the time they became famous enough to give interviews to the press, or in Fiorina's case, run for office.
Depends on the cost assumptions (Score:2)
The paper doesn't show the actual numbers for estimated AI costs, even though those are the key numbers. AI training is the big cost, and the paper assumes that ongoing training will be needed. That's a big assumption, and the final results are likely very sensitive to that assumption.
Re: (Score:2)
Re: (Score:2)
machine only getting cheaper (Score:1)
Douglas Adams was wrong (Score:2)
Telephone sanitizers are irreplaceable!
Re: (Score:2)
Telephone sanitizers are irreplaceable!
Congratulations, you missed the joke.
After sending all the Telephone sanitisers (UK spelling please) and other useless professions away on the B Ark, the Golgafrinchans died from a virulent disease contracted from dirty telephone.