Will >97% of all jobs be fully automated before the year 2075?
➕
Plus
105
Ṁ16k
2075
62%
chance

Up to now, increasing automation has resulted in more jobs being created than lost or overtaken by the machines.

However, there are an increasing number of reports which state that, with the advent of AI, this time will be different. Recently, Goldman Sachs has estimate that 25% of all jobs could be entirely automated by currently existing AI technologies. And there was a report by McKinsey, published in 2017, which claimed that by the year 2055, nearly 50% of all jobs could be fully automated. Recent advances in generative AI have only added to the discourse with the supposition that it is the white collar work that will be automated away first...

Will resolve "Yes" if at any time before January 1st 2075, all jobs that could be held by a human are instead given over to an AI, and everyone everywhere finds themselves both unemployed, and unemployable. Will also resolve as "Yes" if it is found that an AI could perform any given job instead of a human.

Else, if it can be proven that there are some lines of work that can only be performed by a human by 2075, or an over abundance of such, it will resolve as "No".

Update 08/30/2023: Question reworded to "Will >97% of all jobs be fully automated before 2075?", in order to account for the tiny minority of jobs that will never be performed by an AI, such as "Human Pet" 😂

Get
Ṁ1,000
and
S3.00
Sort by:

This is ridiculous. Because as soon as you automate it, it ceases to be a job. Think about any basic task which a computer does. Every mathematical operation could have been made by a human, but is now automated. As I type this paragraph the computer is doing the work that hundreds of humans would previously have done. As we live right now 99.99999% of all jobs have been automated. Probably more. Because it takes a human days to run the hash function for bitcoin but a single gpu does this in some ungodly order of magnitude. So just in the field of cryptography the number of jobs which have been automated is basically infinite.

predictedNO

@ChadwickMiller same line of thinking i had. as soon as a human’s job gets automated, they go and get a new one. we’ll keep making new jobs

@ChadwickMiller This is why I have rigidly non-moving goalposts for when we have AGI. We have AGI when a company hires an AI for a job. It can be any job; even a checkout clerk, or a help desk.

But it cannot just do the stuff, for precisely the reason you say. The company has to actually hire the AI, not make an agreement with some AI company. The AI has to get paid, not some AI company. The AI gets to spend the money on whatever it wants. It does not go to the AI company. Once this happens, namely once people are actually making deals with AIs, they are clearly on their level or above.

People will say this is unfair, because it is far more demanding than simply being intelligent, and because there are legal and other obstacles in the way. Indeed; the same is true of the Turing Test, which not only requires intelligence but faking being human. And in both cases, the reason for the higher burden is to ensure that you don't get something that passes the test, but is not generally intelligent. To be sure that something is the case, you don't have to get barely past a line; you have to get well past it.

@DavidBolin This seems weird to me. Why would it matter if the AI do the exact same thing, but isn’t pay and the company who build the AI get pay instead ?

@DavidBolin with the AGI that can enter legal contracts as a subject it would be more likely that it would just build its own shop.

@dionisos It would matter because it means the people paying trust the company, not the AI. If the AI does something wrong, it is the company's responsibility to fix it.

It is a test of AGI, so all the trust has to be in the AI.

@DavidBolin so your definition of AGI is based on the trust involved rather than capabilities? That seems like a really imperfect definition that might never be reached even if AGI is reached.

Happy to bet anyone 1-10 btc on this at 10% odds.

@MaybeNotDepends I'd take No.

predictedYES

@MaybeNotDepends Fair enough, you’re on!

Once upon a time, most jobs were farming jobs. Most of those have been automated away. Most humans still have jobs.

How would these resolution criteria handle that event?

predictedYES

@EvanDaniel That is true, and I don’t necessarily rule out that the next wave of automation will create other newer kinds of jobs instead of replacing them altogether. The resolution criteria I have in mind for “job” is as follows:

  • Isn’t superfluous or purely recreational (i.e. the job is needed and an integral part of keeping society running. We don’t need marathon runners or celebrities, or horse and buggy riders, strictly speaking…).

  • Requires direct human input or supervision. The key word here being “required” i.e. the AI cannot do it on its own.

That’s why I’m not counting things like “human servant”, and in much the same way that I don’t consider “pet” as a job occupation currently filled by cats and dogs, or “farm animal” for cows. The 97% thing is a little tongue and cheek in that regard.

In previous technological revolutions, a large part of that process was that they created new needs within the economy. Someone needs to maintain cars, roads, transmission lines, computer networks, etc. It also helps that up until now, all of the machines that have been built have no intelligence or autonomy whatsoever. Most cars still can’t self drive, for instance, and as of 2023, the ones that do currently suck at it.

The same process may well apply to full automation with AI. I have seen arguments that, with full automation, lots of new jobs may be created in education, especially in “teaching” AI’s. Maintaining data quality and pipelines might also be the domain of humans for a long time to come as well. I will concede that it is unknowable if they’ll be able to do that themselves before 2075, and it is always possible that the space of possible new jobs is far larger than we imagine.

Another note is that I suppose entertainment, broadly speaking, could count as a need. But historically that has always been a the domain of a very tiny minority of all people who have ever lived, and this will likely be the case under full automation. Not everyone can be a celebrity, and one really needs to stretch the definition of “jobs” if you want to make the argument that we will all be entertainers (and entertaining each other) while the AI’s run and manage everything for us. And, there is also the fact that an AI could equally do all of the entertaining too.

How many jobs are automated now? A few hundred years ago well over a majority of humanity worked in agriculture. These days it’s about a quarter of people, and in developed countries single digit percents. Most people still work we just do different/new jobs.

People will surely still work, it will just be vastly different than what we do today. What that means for this vague question? Hard to say.

predictedYES

@WillH See my response to @EvanDaniel above, I clarify things there. To summarize, I don’t discount the possibility of new/different jobs which the vast majority of future humans will likely perform.

< 0.001%

Simply put, what will your numerator and denominator be in 2075? If you cannot answer that then the question is too vague. And if it's #of autonomous 'jobs' and # of jobs, then it makes no sense.

How do you count autonomous jobs? To count jobs the government asks people if they have a job. They then add up the number of people that say yes. Counting people is easy. There is no counterpart.

predictedYES

@sziner To answer the questions on both of your posts:

  • The main criteria is a job which is not superfluous or purely recreational (ie a job which exists purely for its own sake, or a “bullshit job”). You can see my response to @EvanDaniel above for a greater clarification.

  • I make no assumptions as to the perceived quality of those future jobs, should there still be plenty to go around by then.

  • Yes you are correct that it is possible that humans might still be cheaper in many domains even if an AI can do it. In which case, one can also ask whether that will always be the case into the far future.

Another thing I was thinking about as well is that there might be environments for which an AI would be ill suited to operate in. Current shielding methods from things like EMPs are very expensive and will likely always be. Humans don’t have that problem broadly speaking though!

@Meta_C you did not answer the question.

You've said what you think constitutes a job but not how you actually *count* them. I can agree that checking people out at a cash register is a job. I can agree that working a lumber yard is a job. I can agree that doing valet parking is a job.

What I have no clue is how you count the amount of "automated jobs".

Is 10 self-check outs 10 automated jobs? What if the self checkouts only have the speed of 5 former workers? And then, they create the new job of the person helping people the self-checkout. How many jobs are automated here. Then there's all the grocery stores and restaurants that opened with self-checkout. They never replaced anyone per se. How are you counting these jobs?

How many jobs is a single conveyor belt that does the work of 100 people carrying lumber? Then there's entire industries that are only economical because of conveyor belts. How many automated jobs is this? How many jobs were needed to build pyramids that are now 'automated' because we have cranes and trucks?

How many jobs is self-driving car technology?

And it goes on and on.

Tell me how you'll count automated jobs. There is no other question that needs to be answered.

predictedYES

@sziner

A couple ways of we could count the number of automated jobs vs non-automated. We could either:

  • Do it same way that we would count the number of non-automated jobs, if we want to count any job that an AI would perform as one (so a self-driving car is 1 “job”).

  • By estimating the number of (human) jobs that are displaced due to automation as a proxy. See here for example:

https://news.mit.edu/2020/how-many-jobs-robots-replace-0504

This way, via displacement, is probably a better and more consistent method. But then, as you point out, there is the question of whether the “labor” performed by a conveyor belt or a crane would count as one job vs 100 jobs. There are some estimates out there for past jobs that have been displaced/automated away, though they are inconsistent. I won’t pretend to have an answer for that.

Of course, if new ones are created in place that replace the old ones, then it matters little and we likely wouldn’t care to count them.

predictedYES

@sziner

Another method, also by proxy, is perhaps counting the number of humans who are employed, and then comparing those who can be meaningfully considered to be “permanently unemployed”.

I haven’t found any statistics regarding that; so far the convention is to either consider them long-term unemployed, or to remove those who have given up looking for work from employment statistics altogether. If the worst case scenarios typically imagined in the media do start playing out, this practice might change and we may well count those who can’t find work, or have given up, as permanent unemployment.

predictedNO

@Meta_C seems you acknowledge there is not one way to define what you mean.

This is way too vague. By some interpretations already >97% "jobs that could be held by a human are instead given over to an AI". It depends where you draw the line on what a job is and what an AI is.

We used to pay people to connect phone lines. Is connecting phone lines still a job?

Is AI only LLMs? What about neural networks? What about simple linear regression?

Adding up and multiplying numbers is a job that could be held by humans. Right now my computer does it. AI is doing it all the time. If you consider that a job, then it's already the case.

If you mean, "will the unemployment rate ever hit >97% because of AI", then the answer is no. The reason is simply supply and demand. For the equilibrium quantity of labor to be that low, literally anything humans could do would have to be more economical to do by AI. Problem is, the available supply of human labor, would be so large and therefore it's price so cheap, that it's absurd to think AI could be economical.

We already could replace every job by AI if we wanted to. Throw a few hundred billion dollars at most jobs and it could be done. It's simply not worth it. It's not economical.

predictedNO

Can you define 97%? Is it 97% of current job titles, or 97% of current job positions, either of which would actually fit the original “NO” position? Or simply a 97% drop in employment rate from now till 2075, which is more in line with the original “YES” position?

I apologize if it isn’t your intention but I can’t help feeling a lot of people are getting rugpulled by the rewording.

The job of "human servant" cannot be replaced by AI. Already in the present a big (if not the main) point of having a servant for people without disabilities is a display of wealth. So if automation makes everything cheaper, there will be even more incentive for the upper classes to employ humans, just to show off. AI cannot fulfill this job by definition. More riffs on this idea at:
https://worldbuilding.stackexchange.com/questions/234416/in-a-capitalist-society-where-automation-has-made-the-value-of-human-labour-wort/234421#234421

predictedYES

@MartinModrak That is a good point. Though I would imagine that such jobs, if we want to call it that, would be entirely superfluous and may well no longer be defined as “employment” (in which case they might as well just be slaves or pets).

predictedNO

@Meta_C Why wouldn't you call it employment? Those would (or at least could) be people voluntarily entering into a contract to perform assigned tasks for money... You can definitely redefine "employment" or "job" in a way that would make your prediction more likely to be true, but I think the common sense meaning includes jobs like "human servant" or "making X by human hand" which cannot in principle be replaced by AI and which are extremely likely to remain valuable in the future (as already noted by Phoebe).

As another contemporary example, every year, my mom buys a set of Christmas postcards with reproductions of paintings made by people with severe physical disabilities (e.g. holding brushes by feet or mouth). She would not buy the very same postcards had they been made by able-bodied people or AI (and would require money back if it turned out the images were drawn in a different manner than claimed). So the job of "painting with mouth" exists, despite hand being much more suited to painting and so technically "outcompeting" those artists.

Also, professional sports exists. Despite e.g. cars or bicycles "outcompeting" human runners on a marathon, you can make money as a marathon runner. Why? Because people value weird things and you just cannot reduce this to a one-dimensional competition for "better" or "cheaper".

predictedYES

@MartinModrak Is it still voluntary if, functionally speaking, the only job that will be available to you is “human servant”?

I grant that people value “weird things”, and so there still may be a tiny minority of people who will make a living off of professions such as professional sports, or painting, under conditions of full automation. But I seriously doubt that this is going to be the primary engine of economic activity.

Then again, perhaps I worded the question badly. Maybe it was better to ask “Will >97% of all jobs be fully automated by 2075…”

predictedNO

@Meta_C I agree that's a more sensible wording (although it potentially makes my bet and those of others less valuable as those were made for the original wording).

Just note that this may totally be voluntary - let's say there's universal basic income in place. At least some people are expected to work for various reasons (boredom, fulfillment, getting a better standard of living than the one supplied by UBI)

predictedYES

@MartinModrak Yes agreed!

@Meta_C there are real jobs right now called travel influencer. These are attractive people who take luxury vacations and film themselves doing it. They publish the videos to YouTube and tik tok etc.

It pays enough to pay for the underlying vacation so the influencer never stops vacationing. While dealing with a camera and editing a video is work it obviously could be automated and some YouTube stars make enough to pay someone else to do this part of the work.

Is this a "job". How about Let's Play videos where someone speedruns saving the Titanic in virtual reality. Is this work?

This question is only valid if the employment rate for humans is under 3 percent.

predictedYES

@GeraldMonroe

The 3% thing is largely semantics, the spirit of it remains the same; would anybody need to do any work to make a living… if the current wording of it annoys enough people on here I can change it back 🙂.

On that note, is that what you think everyone will be doing with their time in the future? And will everyone be required to become a source of entertainment to justify their existence?

I suppose it could happen, but I personally find this line of thought extremely sketchy.

@Meta_C Taking the question more seriously, if humans don't want to be helpless pets or consumed by AI, they will have to remain working. They will have to build AI systems in such a way that real human review is required and an AI system will be unable to take action without it for more than a short period of time. And other safeguards to ensure the humans are paying attention. (one way is to have the system insert extremely hostile plans by the AI into the review queue, like "and then build a fission bomb.." to ensure the human reviewers are checking). Human activity would also be vastly larger in scope, someone would need to inspect and manage each O'neil habitat and antimatter factory near the sun and starship construction and testing and so on.

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules