This is unbelievably ironic considering Luddites literally opposed the advancement and use of new technologies because they believed that it would put them out of business. Those screaming that AI must be stopped or it'll get rid of all of our jobs have far more in common with Luddites, and in fact the quote you're trying to compare to Luddite beliefs is in direct contradiction with them.
Just like any technological improvement in all of human history, AI will replace some jobs and open up economic incentive for brand new ones to exist.
> considering Luddites literally opposed the advancement and use of new technologies because they believed that it would put them out of business.
I don't think you really understand the story of the luddites that well.
So, lets say I replace your job with AI today, also I replace the job you were going to get after this job with AI. Ok, you need to eat today, and need to pay your bills tomorrow. How exactly are you going to do that. I mean you can start retraining, but that's going to take months to years and you have days before you're kicked to the street. This is what the Luddites experienced.
In the US we really have no, to nearly no social safety net. If there are mass layoffs because of AI it will cause deep issues in the employment world as everyone adjusts.
>I don't think you really understand the story of the luddites that well
Nothing you said was evidence that I don't understand the story well. You should leave your baseless opinion on my background with the topic out of it or provide evidence that what I said was incorrect.
Additionally, nothing else that you said was relevant to my message. I'm not arguing with anyone about "the US social safety net". Calling someone that argues AI won't take your job anytime soon a "Luddite" is the exact opposite of what Luddites were. If I'm wrong, I'm happy to hear more about that.
Well you see, the government will print money and give it to the companies, the companies will buy up assets creating further asset inflation. The government will print more money for austerity and give it to the companies that will then use it to buy up other companies. In the end one large company will own just about everything, and the government won't do anything about it because they 'are too big to fail'.
I mean, you could say that's not possible, but watching the markets over the last 15 years should give one pause to realize it's not completely improbable either.
I'm not a smart man. But I just don't see how any kind of economics work at this point. This sounds analogous to the universe eventually reaching the same temperature and pressure everywhere, and no events can occur ever again after that.
EDIT: Oh, we're being pedantic (I guess I kicked it off). "They were right", as in a group of individuals lost the employment they had at a particular time, reasonably believed to be a result of automation. Not that these people were unable to find new employment or that the total number of jobs available was reduced.
Just look up any history of the luddites. There are plenty of resources online. I don’t see why I should have to cite sources when I’m pointing out something utterly uncontroversial for anyone knowledgeable on the subject.
Many of them were weaving workshop owners who lost their livelihood to automated factories. The transition to automation in the early 19th century was painful. That doesn’t mean I support what they did or think automation was wrong. In the grand scheme of things it was a necessary transition. That doesn’t mean nobody was hurt by it though.
EDIT I’m not being pedantic. I just made a statement that was factually correct that I’m surprised I have to defend. You seem to be inferring a whole lot of baggage that just wasn’t there.
BTW for context like the articles author I see LLMs as just being another productivity tool. They will become part of our work flows and will have effects similar to those of other productivity tools.
Luddites were conducting industrial action. Their activity correlated strongly with union busting activity by the English elite. It has nothing to do with 'stopping progress' or trying to reverse automation.
You can’t do much, unless willing to take extreme actions. Elites will decide whether to kill off rest of the human population or allow them to scrape by via UBI.
Though buying couple of commercial GPUs and building your own company via LLM based agents might be an option for some.
>Prompt: Someone said "Models are still expensive to run, hard to use, and frequently wrong" Does this also apply to human worker?
>GPT4: This statement, while referring to AI models, can also apply to human workers in certain contexts. Here's a breakdown of each aspect mentioned:
> Expensive to run: Human workers may have higher costs associated with hiring, salaries, benefits, and overhead expenses. However, the costs of human workers can be justified in many cases due to their adaptability, creativity, and ability to learn.
> Hard to use: Human workers can be more difficult to manage compared to AI models, due to factors like communication, collaboration, and personal differences. However, humans are capable of adapting to new situations and understanding complex, nuanced scenarios.
> Frequently wrong: Humans, like AI models, are not infallible and can make mistakes or errors in judgment. However, humans possess the ability to learn from their mistakes, improve over time, and apply critical thinking to avoid making the same errors in the future.
> So, while there are similarities between the statement's implications for AI models and human workers, it is important to remember that both have their unique strengths and weaknesses. The best approach often involves combining the capabilities of AI and human workers to achieve optimal results.
ok, i guess this isn't something we are supposed to admit, but there are many things i have learned over the years which have been distilled to thought-bites, right? i often want to comment just short things, slogans almost. these AI have been invaluable in beefing up my slogans into longer statements with the un-distilled facts laid out in bullet points. it's pretty helpful as an argument de-compression tool.
but that has pretty serious epistemological concerns. and to the people who are concerned about that - well i think that's funny... maybe take a walk down the science of how poor memory really is.
i guess i just have a very attenuated view of human (including my own) rationality. it's waaaaay more inflated for the average human than we're willing to admit.
i made a comment early on in December about the oncoming epistemological crisis and damn has it been really fun to watch unfold the last few months.
An advanced specialized Model could probably create fleshed out arguments from your thought-bites and subtly manipulate you by weaving other positions you don't/didn't have into the argument. "You're right, because xyz is correct and that makes your argument correct", making you more likely to accept xyz as correct, because we like being right and if accepting xyz makes you right in this very moment, that's a small price to pay.
it feels like that would require much more sophistication, but then again who hasn't been amazed how intelligent ChatGPT & friends feel?
the RLHF has indeed made them quite sycophantic. i don't believe in any of the 'latent space wario' stuff, but i really enjoy using it explicitly to get clarification and explanation from opposing viewpoints.
i would love to see some experiment where a debate is held/moderated by an LLM which can help find common ground in highly contentious viewpoints. the post-mortem from each side would be a fascinating read.
Why people arguing agains AI importance are using some cherry picked examples from currently available models while being blind to rate it improves? It's like saying wikipedia is shit because it currently has only 50 articles at some point in the past.
> It's like saying wikipedia is shit because it currently has only 50 articles at some point in the past.
On the other hand, the people drawing inferences from the exponential curve of Wikipedia's editor base growth about its potential to be a resource of unmatched accuracy were even more wrong, and it hit its quality ceiling pretty early...
Well Wikipedia is both clearly above the sustainability threshold and failed to use its critical mass of editors to exponentially improve its accuracy as opposed to creating more maintenance tasks and arguments about which version is better. Does this sound like a plausible future for LLMs once the low hanging fruit is picked off? I'd say yes. Certainly anyone predicting that the exponential growth of Wikipedia would replace actual human researchers would have been miles off, and not because people stopped trying to improve it...
Are you seriously saying wikipedia was a failed project?
Unlike wikipedia, with AI there is a threshold where it can start improving itself entering self improvement loop.
Similar to how computers can improve next generations of computers (with a lot of human effort) AI can improve AI at some point (with minimal or no human effort).
No, I'm seriously saying that Wikipedia is moderately useful provided you don't trust its answers too much, and stopped improving in quality fairly early on and is now merely bigger than it was. So as an analogy, it strongly favours the people arguing in favour of diminishing returns to improvements.
And, at least at this time AI isn't going to ask for rights, time off for family, or even ask for a break after 340 hours of working.
We are having modern John Henry moment where the machines still stuck just enough that humans are ahead, but even small improvements in machines lead to a rapid replacements of parts of the workforce.
Much like we didn't stop improving drills, and no one in their right mind would drill a commercial tunnel without machine labor, our current AI models based in LLMs are still in their infancy, no one has any plans on stopping here.
We should also note that John Henry dies in the story from overworking, so we may want to take that in mind too.
> at least at this time AI isn't going to ask for rights, time off for family, or even ask for a break after 340 hours of working.
People said the same stuff with the advent of computing. We’d all be living lives of leisure while computers did our jobs for us. It definitely didn’t work out that way.
You can blame a worker it something goes wrong and you trusted their result. But you can't go to YOUR boss and blame ChatGPT if YOU used the wrong prompt and trusted the result of the AI.
Salary isn't for delivering an untrustworthy result, its for being responsible for the result. AI doesn't take responsibility.
You've provided the answer yourself. The person who signs off on using the AI output bears the responsibility, unless the AI company provides some kind of guarantee, in which case the AI company bears responsibility.
These aren't new issues. That's why we have things like malpractice insurance, liability insurance, and so on.
Also humans are very expensive too, and in many places starting to become a more limited resource. If we look at a huge number of countries currently, many are at risk of population collapses. Japan being the poster child here. If you want more human brainpower you have to convince humans to breed, then dump a huge amount of resources in to that child for nearly two decades. After that you hope and pray they actually want to work at solving the problems in the world instead of playing online all day with their Waifu Replica.
From a historical point of view, computing was expensive and rare, the idea we'd have massive warehouses full of servers connected to a global network was but a dream when I was a child. A single computer (large) computer would have cost as a much as a house and had less power then what you're surfing instagram on in your pocket.
This is definitely being achieved in some respects today. We certainly have the metrology needed to work with this scale now.
I would also add quantum computers to your list. I realize the tortured path this idea has gone down, but even non-exponential speedups like Grover's algorithm could have a major impact on the economics of large-scale optimization & search. IBM is releasing a computer this year that allegedly has 16k+ qubit support.
Yes, quantum computing will replace a subset of classical computing at a cost that is 1/1000th of traditional computing. I have added quantum computing to the list.
> Even researchers can’t keep up with AI innovation
So things are improving so fast that researchers can't keep up... and therefore nothing will change?
This whole article smells like wishful thinking.
You don't need workers to understand how their jobs are being automated, you just need a few experts who can drop in, turn half your payroll (or more) into a redundancy, and then move on to the next project.
It becomes way cheaper to start a new business, so the number of jobs for people who know how to use AI increases, but all other job-markets with AI-exposure tighten.
<< You don't need workers to understand how their jobs are being automated, you just need a few experts who can drop in, turn half your payroll (or more) into a redundancy, and then move on to the next project.
Someone has to maintain it when experts drive away. I personally foresee something of a clusterfuck based on interactions with vendors and consultants, because that is kinda where it is heading. At some point, someone has to know what they are doing. It can't all be magic blackboxes.
We have good enough robots to get warehouse workers out of the jobs - it did not happen overnight and is still in progress. I was working on warehouse automation 8 or so years ago robots were already insanely efficient at that time.
I switched area to more compliance stuff. Guess what - loads of companies still doing pen and paper.
I don’t even know if we had in history such an event where bunch of people would be cut out.
Interesting points in the article, but it seems to assume current state of the art will continue as-is, and frankly, shows little to no imagination of what "might be". It reads like the personal computer "ok, so what?" responses from the 80's.
The creativity of a large group of people will find new and interesting ways to leverage this technology. We will see more AI-assisted IDEs, debuggers, assistants, document "interpreters" (e.g. converting contracts into litigation, specifications into running code, rough ideas into whole new literary or film worlds), and so much more.
And, as far as I can tell, we are terrible at gauging the speed of AI advancement. See [1], [2]. These may be cherry-picked examples, but some of these are orders-of-magnitude off.
"Training to use AI will be a job in itself"... until it isn't. Adversarial AIs are probably already in the works to teach these LLMs right from wrong. Such training will take time, for sure, but the next leap is pretty clear in my mind: human obsolescence given our current (at least here in US) sensibilities. I'm not saying new sensibilities won't come, but historically, it has taken a long time for such base-line cultural shifts to take effect. Even after armed conflicts.
In my opinion, the question isn't, “[can] we automate away all the jobs, including the fulfilling ones?” The question is, "What will you do when your job becomes redundant?" The best AI researchers may be way off on their predictions here... it could happen in the next 12 months.
Thankfully, to some extent, this is just the output of LLMs and predictive analysis. But I'd bet someone is already tweaking a prompt/chat along the lines of "Build me an AGI which can make decisions on its own."
> "Training to use AI will be a job in itself"... until it isn't. Adversarial AIs are probably already in the works to teach these LLMs right from wrong.
If you have a project that uses 2 AIs to train themselves and nothing else, and a project with 2AIs training themselves, aided by a human team, who will win? I'll place my bet on the human+AI team.
> Build me an AGI which can make decisions on its own.
The best models still just output what they’ve indexed from the training data. They can’t build an AGI because no one’s built one yet for them to index.
In our current high internet rate environment, we’ve already seen perverse incentives around firing workers to juice stock prices. Even if everything is the post is true, I don’t see what’s preventing c-suite’s from doing more of the same with AI. Sure, maybe it will all blow-up in their faces when an unchecked AI engages in wildly irresponsible or illegal behavior, but they’ll just golden parachute away to their next gig replacing people with LLMs.
What you’re describing is the commandment to “follow incentives”. We will adopt it to the hilt no matter who we are because the incentives dictate we must. What governments or people will do when many are out on the street because their job has been made obselete, that’s the real question.
In the US, we already have the answer to that in the decline in life expectancy as a result of overdoses, suicide, and interpersonal violence in areas where industry moved overseas. Those “out on the street” will be left to rot.
"Everyone" is doing a lot of implied work here. AI doesn't have to take over everyone's job. Even 10% would have a disastrous impact on our workforce, especially if there are insufficient safety nets.
That being said, I feel fairly certain that our economy will come up with (if not outright demand) more Bullshit Jobs† for the workforce to fill to keep all the appropriate graph lines going up.
100 years ago a whole village had to help with crops, now a single air conditioned combine harvester can do the job (hell, doesn’t even need a driver, it works solely through GPS). We somewhat earned a bit of time with many of the bullshit jobs, but I think that bubble is bursting — we will have to undergo a huge societal change (UBI or something like that), otherwise we will get to Gotham city levels.
We'll go from robots supporting humans to humans supporting robots.
My Roomba still can't vacuum the house without needing some form of "help" at least three times. Robots can do a lot, but it will probably be cheaper to pay human custodians to manually intervene as needed than to program AI and bots to account for every conceivable failure case in every possible scenario.
I'm with you, I don't see where everyone will go. We've got a few sinks or de facto jobs programs (military; healthcare system, in the US) for "excess" workers, but I don't know how any of those can expand enough to absorb a shock of the sort that LLMs and related tech may soon usher in—especially when those programs or semi-deliberate inefficiencies aren't actually sold as being for that purpose, how much waste will we collectively tolerate?
UBI definitely seems like a more-honest approach, and probably healthier all around.
[EDIT] Just after posting this I had a mini-nightmare about how LLMs might enable a much, much higher volume of lawsuit activity with the same number of highly-skilled workers, driving an expansion of low-skilled ancillary jobs in that sector, at the cost of an absolutely huge drag on the overall economy and probably increased overall stress, to basically no actual benefit to society. God, chills up and down my spine.
Assistant administrator to the prompt engineer? Director of Phone Polishing? Air Circulation Specialist?
In many high population countries, people are used where other countries would use automation or robots. Automation is relatively expensive when compared to people desperate for any job that lets them meet their basic needs.
Bullshit Jobs are an important part of every economy. If the criteria for work was "only jobs absolutely needed", unemployment would be at 30% and society would be unstable.
Even the low-touch, mind-numbing jobs seem safe. I was at a 7-11 yesterday and I thought about how bored the clerk looked... but there's no way AI could take over the job of "being the human who sits in the store and makes sure nothing gets stolen". The task of physically carrying out the checkout and taking the money is pretty much already automated, but there still has to be a human there, for human-related reasons.
The only reason there is a human and the store isn't one big vending machine is the large upfront capital cost of doing it that way. Once AI+Robotics is good enough, those costs will come down significantly. Hard to imagine those humans not being replaced as well.
Humans like to browse. And setup of modern stores is rather space efficient for storing and displaying all the items. I don't think stores will be replaced. Any level of automation is unlikely to drive prices low enough for that to happen.
Did you ask the cashier if they owned a house and had any children?
We're not exactly solving anything at a society level if we replace all the high paying jobs and let corporations pocket that money directly, while only leaving lots of shitty low paid jobs for the starving masses to fight over.
This actually the biggest negative I bring up in these AI conversations.
AI will have nothing but negative impacts for society as a whole. It's benefits will be experienced almost exclusively by the corporations and their C suites. If the creators of GPT and it's successors are doing it for profit, I guess I can understand. However, if they are doing it to improve civilization and revolutionize things for the better, they are either extremely idealistic or completely ignorant.
> there's no way AI could take over the job of "being the human who sits in the store and makes sure nothing gets stolen"
Humans aren't there to stop theft, most are specifically trained not to interfere. Everything in a 7-11 could be sold from a vending machine.
Service Merchandise was most of the way there. No human was really needed except to pick items and load them onto the conveyor belt at checkout. Robots can do that now.
The irony is, in the new world, nobody will have enough money to buy anything from these places to begin with.
The way stores used to work, when goods were relatively more expensive than labor, would be that you would go up to the store counter and ask for what you wanted and the clerk would fetch it from in back rather than the customer having to find goods on the shelves. With cheap robots you could go back to that model.
I'm assuming this is not in Europe, where automated cash registries in grocery stores is more common every day. In Estonia lots of grocery stores have ~10 automatic registries and just 1 real human one. The automated one needs no human at all, and I'm sure soon even that 1 real human will no longer be needed.
As for stealing stuff, well, that's cultural. Seems northern Europe has managed to be just fine automating all this stuff away, I'm sure the rest of the world will catch up eventually.
We already have food delivery by robots (https://www.starship.xyz) here in Estonia, currently testing package delivery by robots as well, combine that with automated grocery store experiences, kiosks where buying things is also entirely automated, well, I really don't think low-touch, mind-numbing jobs are safe at all. Unless in places where the culture is holding the advancement back, but it surely isn't technology that does.
Amazon has its convenience stores in which you scan your phone at the entry and walk out with whatever you want, it charges you for whatever you take. No human necessary
They don't have to take jobs completely. Every job these tools make 5x, 10x, 100x more efficient, that'll be fewer workers needed for the same output. Those tasks will be rolled into other jobs (I expect a lot of that—consider how we're all secretaries now), department headcount reduced, whatever. Unless demand also rises tremendously because costs plummet, which it may for some things, but surely won't for all.
Everyone's job? Yeah, sure, that's probably some time off or never gonna happen, depending on how things shake out.
Yep, one of the worst possible outcomes of AI currently (skipping the entire AGI kills us thing) is by drastically increasing informational efficiency, hence replacing most of our high paid jobs, while not significantly increasing resource efficiency. You would have a massive decrease in wages while not having deflation in any other metrics.
This is when things like 4 day work weeks and expanded vacation time via regulation actually make sense. We've become so much more efficient? Reduce the amount of work allowed to achieve full time employment to force the efficiency into more leisure time. Of course, the conundrum there becomes outsourcing and nations being incentivized not to do so to build their economies. We may be trapped in a tragedy of the commons leading us to slavery.
None of the required changes will be made in our current corporate kleptocracy. We are going to see massive unemployment that will cripple the economy as a whole. Not the utopia we should have. The pause for me is to figure out the redistribution from the spoils of efficiency.
> Unless demand also rises tremendously because costs plummet, which it may for some things, but surely won't for all.
Maybe, maybe not. Demand has always risen in the past. Historically 90+% of us were manual agricultural labors, generally unfree at that (i.e., serfs or slaves). Nowadays machines do almost all that work, yet we don't have anything like 90% unemployment (yeah, there are still farmers, but not 90%... plus the farmer isn't grubbing in the ground with a stick in the blazing sun, Mesopotamia-style. He's riding in an air-conditioned combine harvester with a fridge full of tasty beverages and an internet connection).
We've absorbed a lot of workers into lengthier education and various jobs programs (the MIC and the military itself; in the US, the horribly-inefficient healthcare sector) at enormous cost, following suspiciously well with the timeline of eliminating many of those other jobs. Would the US have seen such a great postwar boom if we'd disarmed again, or would unemployment have shot back up? What if we'd not had the GI bill to stick a bunch of 22-year-old ex-GIs in college for a while, keeping them out of the workforce? If a four-year degree hadn't become the new high school diploma, and wasteful healthcare spending not expanded and funded tens of thousands of jobs that aren't a net-benefit to society, right around the time computers began automating away a lot of jobs? I wonder.
IMO we're paying through the nose to keep all this automation from being a disaster, we're just doing it in ways that aren't as obvious as an outright, robust welfare program.
> This is one of several questions posed by the Future of Life Institute’s recent call for a pause on “giant AI experiments,” which now has over 10,000 signatories including Elon Musk.....
In other news, Elon Musk paints over the W in twitter in petty dispute with landlord and changes logo to dodge coin logo.
yup all quality people that know what's going on signing that petition.
Let's take a historical analogy. Here's a cool video about how to make money trading. If you're under 40 you might only have seen this form of trading in the Eddie Murphy/Dan Aykroyd movie Trading Places:
There's two kinds of advice in this video. There's the timeless advice: be informed, understand why people are trading, know the rules. There's the incidental advice: be physically dominant, stand in the right place in the pit, look around you.
Not long after this talk, trading started getting done by computers. This is still ongoing amazingly, but it started around when this video was recorded. People were already known to be bringing in machines to calculate prices, and a friend of mine was one of the first to type in old commodity prices into a computer in order to find long running trends.
Some of the guys who were successful on the floor were also successful "upstairs" sitting in front of a screen. Others went on to driving a black cab.
What happened? Two things, one is that fundamentals remain fundamentals. If you knew the rules and understood the customers, you could still make money on a computer. You also had to learn how computers worked. This often meant that new hires were of a different culture entirely to the Essex boys who'd previously ruled the market. My first boss traded options without the equivalent of a high school diploma. He (wisely!) hired an Oxford engineering graduate.
What also happened was that a new incidental skill-set became very important, namely programming. Suddenly opportunity opened up if you understood how microwave links worked, how lock-free queues worked, how exchange APIs worked. This was perhaps the equivalent of that story about the Rothschilds and the Battle of Waterloo, where the economic fundamentals were the same but the winner was someone who could organize the mechanics best.
I suppose AI will be the same. Various businesses will have their fundamentals, but AI may entirely change the face and the culture according to who can figure out how to use it. Some people will survive having retooled, others will have to leave, having discovered their position was actually only held up by incidental factors.
The arguments being made in this article are remarkably poor, and the entire premise is poorly defined.
What do you mean by "take over", "everyone", and especially "soon"?
If AI changes the world as much as the internet, personal computer, and/or automobile, does it really matter if it takes 6 months or 5 years? That's an extraordinary amount of change over such a period. That's worth the hype.
> how, exactly, would AI be used to automate all jobs? Setting aside whether that’s even desirable—is it even possible?
Obviously we can't automate literally every job at this time. You're still going to need a human around to feed questions or instructions into the machine and review or test the results, but automation was never going to eliminate all human labor simultaneously. Machinery reduces the number of humans you need. So far, this has always been the case. The printing press doesn't eliminate the need for people working on duplicating books. It means one person can do the work of a whole monastery. Stores still need cashiers to watch the automated checkouts, but now the machines can take care of 4 or 8 people at a time instead of 1.
> OpenAI has touted GPT-4’s ability to pass numerous standardized tests—but did the model genuinely understand the tests, or simply train to reproduce the correct answers?
It doesn't matter if the machine can think or feel. What matters is the result.
> Still, it’s clear that a future in which “we automate away all the jobs, including the fulfilling ones,” is more than six months away, as the Future of Life Institute’s letter frets.
Is this supposed to be comforting? I don't think anyone reasonable is under the misapprehension that they're going to be fired and replaced by GPT literally tomorrow.
> Open AI’s ChatGPT-3.5 Turbo, for example, prices API access at roughly US $0.002 per 750 English words
If you make $80k/yr, you're being paid something on the order of $0.64/min. You'd have to type pretty fast to overcome the two orders of magnitude between those numbers.
And this is with current hardware. I know Moore's law isn't holding anymore, but computer hardware is still rapidly accelerating.
It's also with the current trajectory of AI development. Nobody can say that won't accelerate. Similarly, nobody can say it won't slow down.
We also don't yet know how or if humans will be able to react fast enough to the change on the horizon. Is the herald of a much better world or is it the antecedent to catastrophe. I don't know, and neither does the author of this article.
It's refreshing to see an article acknowledging the limitations of AI technology in the workforce, as opposed to sensationalizing the idea of a future where humans are completely replaced by machines. The article highlights valid concerns around the high cost of running AI models, their complexity, and frequent errors, reminding us of the indispensable value of human empathy and emotional intelligence in certain industries.
Dismissing the potential of AI is naive but so is the mindless optimism that surrounds it. In a lot of tech circles it’s being treated as an absolute certainty, like when the internet was invented and everyone said the office was dead. The reality was a whole lot more complicated than that.
I'll risk a top-level comment without reading the article at all, not even clicking, with my first opinion-response: humans are _very cheap_, _mindbogglingly cheap_ if you were born in Cambodia or Gods Forbid a place even worse off (no offense to Cambodians, we're all _born somewhere_).
Before I even finished high-school I worked on a factory floor - via a temp agency no less - and was just like "why doesn't the machine also do that too, it's so dumb?". Went to a decent university and... understood. So. Much. Cheaper.
Now I'm life-settings further by a bunch and it still bothers me "Machines will free us all to be the best we humans can be!". No. For many reasons though!
So yeah, the Żabka / 7-11 employees will be there for a long time to come...
... Page me when Japan declares "no more convenience store human employees".
101 comments
[ 4.5 ms ] story [ 101 ms ] threadLuddites - circa 2023
Just like any technological improvement in all of human history, AI will replace some jobs and open up economic incentive for brand new ones to exist.
I don't think you really understand the story of the luddites that well.
So, lets say I replace your job with AI today, also I replace the job you were going to get after this job with AI. Ok, you need to eat today, and need to pay your bills tomorrow. How exactly are you going to do that. I mean you can start retraining, but that's going to take months to years and you have days before you're kicked to the street. This is what the Luddites experienced.
In the US we really have no, to nearly no social safety net. If there are mass layoffs because of AI it will cause deep issues in the employment world as everyone adjusts.
Nothing you said was evidence that I don't understand the story well. You should leave your baseless opinion on my background with the topic out of it or provide evidence that what I said was incorrect.
Additionally, nothing else that you said was relevant to my message. I'm not arguing with anyone about "the US social safety net". Calling someone that argues AI won't take your job anytime soon a "Luddite" is the exact opposite of what Luddites were. If I'm wrong, I'm happy to hear more about that.
I mean, you could say that's not possible, but watching the markets over the last 15 years should give one pause to realize it's not completely improbable either.
It was the Luddites that complained automation destroyed their jobs. They were right.
People now are complaining AI will take away their jobs. Others disagree with them. Which of these two groups takes the modern role of the Luddites?
[Citation needed]
EDIT: Oh, we're being pedantic (I guess I kicked it off). "They were right", as in a group of individuals lost the employment they had at a particular time, reasonably believed to be a result of automation. Not that these people were unable to find new employment or that the total number of jobs available was reduced.
Many of them were weaving workshop owners who lost their livelihood to automated factories. The transition to automation in the early 19th century was painful. That doesn’t mean I support what they did or think automation was wrong. In the grand scheme of things it was a necessary transition. That doesn’t mean nobody was hurt by it though.
EDIT I’m not being pedantic. I just made a statement that was factually correct that I’m surprised I have to defend. You seem to be inferring a whole lot of baggage that just wasn’t there.
BTW for context like the articles author I see LLMs as just being another productivity tool. They will become part of our work flows and will have effects similar to those of other productivity tools.
Though buying couple of commercial GPUs and building your own company via LLM based agents might be an option for some.
The quote is true for human workers.
>Prompt: Someone said "Models are still expensive to run, hard to use, and frequently wrong" Does this also apply to human worker?
>GPT4: This statement, while referring to AI models, can also apply to human workers in certain contexts. Here's a breakdown of each aspect mentioned:
> Expensive to run: Human workers may have higher costs associated with hiring, salaries, benefits, and overhead expenses. However, the costs of human workers can be justified in many cases due to their adaptability, creativity, and ability to learn.
> Hard to use: Human workers can be more difficult to manage compared to AI models, due to factors like communication, collaboration, and personal differences. However, humans are capable of adapting to new situations and understanding complex, nuanced scenarios.
> Frequently wrong: Humans, like AI models, are not infallible and can make mistakes or errors in judgment. However, humans possess the ability to learn from their mistakes, improve over time, and apply critical thinking to avoid making the same errors in the future.
> So, while there are similarities between the statement's implications for AI models and human workers, it is important to remember that both have their unique strengths and weaknesses. The best approach often involves combining the capabilities of AI and human workers to achieve optimal results.
but that has pretty serious epistemological concerns. and to the people who are concerned about that - well i think that's funny... maybe take a walk down the science of how poor memory really is.
i guess i just have a very attenuated view of human (including my own) rationality. it's waaaaay more inflated for the average human than we're willing to admit.
i made a comment early on in December about the oncoming epistemological crisis and damn has it been really fun to watch unfold the last few months.
it feels like that would require much more sophistication, but then again who hasn't been amazed how intelligent ChatGPT & friends feel?
i would love to see some experiment where a debate is held/moderated by an LLM which can help find common ground in highly contentious viewpoints. the post-mortem from each side would be a fascinating read.
On the other hand, the people drawing inferences from the exponential curve of Wikipedia's editor base growth about its potential to be a resource of unmatched accuracy were even more wrong, and it hit its quality ceiling pretty early...
Unlike wikipedia, with AI there is a threshold where it can start improving itself entering self improvement loop.
Similar to how computers can improve next generations of computers (with a lot of human effort) AI can improve AI at some point (with minimal or no human effort).
We are having modern John Henry moment where the machines still stuck just enough that humans are ahead, but even small improvements in machines lead to a rapid replacements of parts of the workforce.
Much like we didn't stop improving drills, and no one in their right mind would drill a commercial tunnel without machine labor, our current AI models based in LLMs are still in their infancy, no one has any plans on stopping here.
We should also note that John Henry dies in the story from overworking, so we may want to take that in mind too.
People said the same stuff with the advent of computing. We’d all be living lives of leisure while computers did our jobs for us. It definitely didn’t work out that way.
Computers have in theory massively automated the world, but it also consolidates capital.
Salary isn't for delivering an untrustworthy result, its for being responsible for the result. AI doesn't take responsibility.
You've provided the answer yourself. The person who signs off on using the AI output bears the responsibility, unless the AI company provides some kind of guarantee, in which case the AI company bears responsibility.
These aren't new issues. That's why we have things like malpractice insurance, liability insurance, and so on.
Edit: added quantum computers
From a historical point of view, computing was expensive and rare, the idea we'd have massive warehouses full of servers connected to a global network was but a dream when I was a child. A single computer (large) computer would have cost as a much as a house and had less power then what you're surfing instagram on in your pocket.
This is definitely being achieved in some respects today. We certainly have the metrology needed to work with this scale now.
I would also add quantum computers to your list. I realize the tortured path this idea has gone down, but even non-exponential speedups like Grover's algorithm could have a major impact on the economics of large-scale optimization & search. IBM is releasing a computer this year that allegedly has 16k+ qubit support.
So things are improving so fast that researchers can't keep up... and therefore nothing will change?
This whole article smells like wishful thinking.
You don't need workers to understand how their jobs are being automated, you just need a few experts who can drop in, turn half your payroll (or more) into a redundancy, and then move on to the next project.
It becomes way cheaper to start a new business, so the number of jobs for people who know how to use AI increases, but all other job-markets with AI-exposure tighten.
Also anything AI adjacent that hasn't been automated (yet?), but I don't see any guarantee of a favorable ratio.
Someone has to maintain it when experts drive away. I personally foresee something of a clusterfuck based on interactions with vendors and consultants, because that is kinda where it is heading. At some point, someone has to know what they are doing. It can't all be magic blackboxes.
Or more accurately, it can..until it breaks.
Maybe you can ask ChatGPT to summarize the article for you, if you don’t have the time to read it.
We have good enough robots to get warehouse workers out of the jobs - it did not happen overnight and is still in progress. I was working on warehouse automation 8 or so years ago robots were already insanely efficient at that time.
I switched area to more compliance stuff. Guess what - loads of companies still doing pen and paper.
I don’t even know if we had in history such an event where bunch of people would be cut out.
It happens gradually anyway.
The creativity of a large group of people will find new and interesting ways to leverage this technology. We will see more AI-assisted IDEs, debuggers, assistants, document "interpreters" (e.g. converting contracts into litigation, specifications into running code, rough ideas into whole new literary or film worlds), and so much more.
And, as far as I can tell, we are terrible at gauging the speed of AI advancement. See [1], [2]. These may be cherry-picked examples, but some of these are orders-of-magnitude off.
"Training to use AI will be a job in itself"... until it isn't. Adversarial AIs are probably already in the works to teach these LLMs right from wrong. Such training will take time, for sure, but the next leap is pretty clear in my mind: human obsolescence given our current (at least here in US) sensibilities. I'm not saying new sensibilities won't come, but historically, it has taken a long time for such base-line cultural shifts to take effect. Even after armed conflicts.
In my opinion, the question isn't, “[can] we automate away all the jobs, including the fulfilling ones?” The question is, "What will you do when your job becomes redundant?" The best AI researchers may be way off on their predictions here... it could happen in the next 12 months.
Thankfully, to some extent, this is just the output of LLMs and predictive analysis. But I'd bet someone is already tweaking a prompt/chat along the lines of "Build me an AGI which can make decisions on its own."
[1] https://www.weforum.org/agenda/2017/07/how-long-before-a-rob...
[2] https://www.forbes.com/sites/forbestechcouncil/2021/05/06/ar...
I honestly think we are getting there. the world will be a completely different place in 3 years or less (in terms of human activities/work/jobs)
If you have a project that uses 2 AIs to train themselves and nothing else, and a project with 2AIs training themselves, aided by a human team, who will win? I'll place my bet on the human+AI team.
The best models still just output what they’ve indexed from the training data. They can’t build an AGI because no one’s built one yet for them to index.
"Everyone" is doing a lot of implied work here. AI doesn't have to take over everyone's job. Even 10% would have a disastrous impact on our workforce, especially if there are insufficient safety nets.
That being said, I feel fairly certain that our economy will come up with (if not outright demand) more Bullshit Jobs† for the workforce to fill to keep all the appropriate graph lines going up.
† https://en.wikipedia.org/wiki/Bullshit_Jobs
100 years ago a whole village had to help with crops, now a single air conditioned combine harvester can do the job (hell, doesn’t even need a driver, it works solely through GPS). We somewhat earned a bit of time with many of the bullshit jobs, but I think that bubble is bursting — we will have to undergo a huge societal change (UBI or something like that), otherwise we will get to Gotham city levels.
Babysitter.
We'll go from robots supporting humans to humans supporting robots.
My Roomba still can't vacuum the house without needing some form of "help" at least three times. Robots can do a lot, but it will probably be cheaper to pay human custodians to manually intervene as needed than to program AI and bots to account for every conceivable failure case in every possible scenario.
UBI definitely seems like a more-honest approach, and probably healthier all around.
[EDIT] Just after posting this I had a mini-nightmare about how LLMs might enable a much, much higher volume of lawsuit activity with the same number of highly-skilled workers, driving an expansion of low-skilled ancillary jobs in that sector, at the cost of an absolutely huge drag on the overall economy and probably increased overall stress, to basically no actual benefit to society. God, chills up and down my spine.
In many high population countries, people are used where other countries would use automation or robots. Automation is relatively expensive when compared to people desperate for any job that lets them meet their basic needs.
Automatic gun turrets?
Did you ask the cashier if they owned a house and had any children?
We're not exactly solving anything at a society level if we replace all the high paying jobs and let corporations pocket that money directly, while only leaving lots of shitty low paid jobs for the starving masses to fight over.
AI will have nothing but negative impacts for society as a whole. It's benefits will be experienced almost exclusively by the corporations and their C suites. If the creators of GPT and it's successors are doing it for profit, I guess I can understand. However, if they are doing it to improve civilization and revolutionize things for the better, they are either extremely idealistic or completely ignorant.
There will be a need for HR, but will a company need all of them, when fewer could do the work with the help of an AI?
Humans aren't there to stop theft, most are specifically trained not to interfere. Everything in a 7-11 could be sold from a vending machine.
Service Merchandise was most of the way there. No human was really needed except to pick items and load them onto the conveyor belt at checkout. Robots can do that now.
The irony is, in the new world, nobody will have enough money to buy anything from these places to begin with.
As for stealing stuff, well, that's cultural. Seems northern Europe has managed to be just fine automating all this stuff away, I'm sure the rest of the world will catch up eventually.
We already have food delivery by robots (https://www.starship.xyz) here in Estonia, currently testing package delivery by robots as well, combine that with automated grocery store experiences, kiosks where buying things is also entirely automated, well, I really don't think low-touch, mind-numbing jobs are safe at all. Unless in places where the culture is holding the advancement back, but it surely isn't technology that does.
Everyone's job? Yeah, sure, that's probably some time off or never gonna happen, depending on how things shake out.
Maybe, maybe not. Demand has always risen in the past. Historically 90+% of us were manual agricultural labors, generally unfree at that (i.e., serfs or slaves). Nowadays machines do almost all that work, yet we don't have anything like 90% unemployment (yeah, there are still farmers, but not 90%... plus the farmer isn't grubbing in the ground with a stick in the blazing sun, Mesopotamia-style. He's riding in an air-conditioned combine harvester with a fridge full of tasty beverages and an internet connection).
IMO we're paying through the nose to keep all this automation from being a disaster, we're just doing it in ways that aren't as obvious as an outright, robust welfare program.
In other news, Elon Musk paints over the W in twitter in petty dispute with landlord and changes logo to dodge coin logo.
yup all quality people that know what's going on signing that petition.
https://www.youtube.com/watch?v=ffEPqLIBnkM
There's two kinds of advice in this video. There's the timeless advice: be informed, understand why people are trading, know the rules. There's the incidental advice: be physically dominant, stand in the right place in the pit, look around you.
Not long after this talk, trading started getting done by computers. This is still ongoing amazingly, but it started around when this video was recorded. People were already known to be bringing in machines to calculate prices, and a friend of mine was one of the first to type in old commodity prices into a computer in order to find long running trends.
Some of the guys who were successful on the floor were also successful "upstairs" sitting in front of a screen. Others went on to driving a black cab.
What happened? Two things, one is that fundamentals remain fundamentals. If you knew the rules and understood the customers, you could still make money on a computer. You also had to learn how computers worked. This often meant that new hires were of a different culture entirely to the Essex boys who'd previously ruled the market. My first boss traded options without the equivalent of a high school diploma. He (wisely!) hired an Oxford engineering graduate.
What also happened was that a new incidental skill-set became very important, namely programming. Suddenly opportunity opened up if you understood how microwave links worked, how lock-free queues worked, how exchange APIs worked. This was perhaps the equivalent of that story about the Rothschilds and the Battle of Waterloo, where the economic fundamentals were the same but the winner was someone who could organize the mechanics best.
I suppose AI will be the same. Various businesses will have their fundamentals, but AI may entirely change the face and the culture according to who can figure out how to use it. Some people will survive having retooled, others will have to leave, having discovered their position was actually only held up by incidental factors.
What do you mean by "take over", "everyone", and especially "soon"?
If AI changes the world as much as the internet, personal computer, and/or automobile, does it really matter if it takes 6 months or 5 years? That's an extraordinary amount of change over such a period. That's worth the hype.
> how, exactly, would AI be used to automate all jobs? Setting aside whether that’s even desirable—is it even possible?
Obviously we can't automate literally every job at this time. You're still going to need a human around to feed questions or instructions into the machine and review or test the results, but automation was never going to eliminate all human labor simultaneously. Machinery reduces the number of humans you need. So far, this has always been the case. The printing press doesn't eliminate the need for people working on duplicating books. It means one person can do the work of a whole monastery. Stores still need cashiers to watch the automated checkouts, but now the machines can take care of 4 or 8 people at a time instead of 1.
> OpenAI has touted GPT-4’s ability to pass numerous standardized tests—but did the model genuinely understand the tests, or simply train to reproduce the correct answers?
It doesn't matter if the machine can think or feel. What matters is the result.
> Still, it’s clear that a future in which “we automate away all the jobs, including the fulfilling ones,” is more than six months away, as the Future of Life Institute’s letter frets.
Is this supposed to be comforting? I don't think anyone reasonable is under the misapprehension that they're going to be fired and replaced by GPT literally tomorrow.
> Open AI’s ChatGPT-3.5 Turbo, for example, prices API access at roughly US $0.002 per 750 English words
If you make $80k/yr, you're being paid something on the order of $0.64/min. You'd have to type pretty fast to overcome the two orders of magnitude between those numbers.
And this is with current hardware. I know Moore's law isn't holding anymore, but computer hardware is still rapidly accelerating.
It's also with the current trajectory of AI development. Nobody can say that won't accelerate. Similarly, nobody can say it won't slow down.
We also don't yet know how or if humans will be able to react fast enough to the change on the horizon. Is the herald of a much better world or is it the antecedent to catastrophe. I don't know, and neither does the author of this article.
When jobs are going to be taken, it's not by AI, but by the willful decision of a handful of people.
The issue is with those who pocket the benefits and leave the consequences to everyone else.
Privatize the benefits, socialize the losses.
Before I even finished high-school I worked on a factory floor - via a temp agency no less - and was just like "why doesn't the machine also do that too, it's so dumb?". Went to a decent university and... understood. So. Much. Cheaper.
Now I'm life-settings further by a bunch and it still bothers me "Machines will free us all to be the best we humans can be!". No. For many reasons though!
So yeah, the Żabka / 7-11 employees will be there for a long time to come...
... Page me when Japan declares "no more convenience store human employees".