What happens when the robots (peacefully) take over?

[k9bfriender]

@SlackerInc

I would call it very optimistic to think that administrators would respond to this by saying “we’ll just keep the same number of teachers but give the students more value”. If they can effectively teach with a smaller payroll, they will do it. (This comes from experience, as my wife is a teacher.) At the very least, they will cut the number of “paras” (what used to be called “teacher aides”), which I assume is not included in that four million number but must be millions more. Or maybe they will use the lower-paid paras more for classroom management and sack the expensive teachers.

Administrators are not the ones in charge of education. The board of education, elected by the community is. Administrators just carry out the will of the elected officials.

So, what I am saying is that the community will respond to better technology as teaching aides by saying that they will increase the value given to students. This is exactly what my local school district has done. They have rebuilt or refurbished their schools with better technology and reduced class sizes. I think our taxes went up a bit in the process, but that’s fine with me.

There may be other school districts out there with communities that don’t care about the education that their students are receiving, and in those cases, the students will get what little resources that the community chooses to give. They may not benefit from technology. Many of them barely benefit from books.

It’ll be the same as it is now. If a community values putting resources into teaching their children, then they will have resources to teach. If they don’t, then they won’t. Technology won’t change that either way.

[SlackerInc]

In my wife’s experience, the superintendent has great power over how money is allocated. For instance, to her frustration, cutting resources for special education while spending big on an Astroturf field for the high school football team. Teachers also sent a representative to school board meetings, but with a weak union it was understood that they would oppose the superintendent’s agenda at their own peril. In places with a stronger union that may well be different.

[Lemur866]

Right, but it’s the school board that hires and fires the superintendent, and it’s the voters that elect the board. If opposing the superintendent gets the teachers in trouble that’s only because the board sides with the superintendent.

If the voters want better books instead of a better sports field, they’re going to have to elect a board that supports books over sports, and the board will have to hire a superintendent who supports books over sports, and the superintendent will have to hire staff who support books over sports.

[k9bfriender]

@SlackerInc

In my wife’s experience, the superintendent has great power over how money is allocated. For instance, to her frustration, cutting resources for special education while spending big on an Astroturf field for the high school football team. Teachers also sent a representative to school board meetings, but with a weak union it was understood that they would oppose the superintendent’s agenda at their own peril. In places with a stronger union that may well be different.

I don’t know how it works in all places, but at my old high school, sports actually paid for themselves. The field and all of that was paid for by the sports (mostly football) teams, through tickets, concessions, donations and so forth. No taxpayer money was spent on it. That was a big part of building the new schools, that not a dime of taxpayer money would go towards extra curricular activities.

In any case, your superintendent is not a monarch above reproach. They are answerable to the school board, who is answerable to the voters.

[SlackerInc]

Special education funding is not supposed to be subject to the will of local voters (which does admittedly make it a partially unfunded mandate). There are federal laws and regulations which stipulate what resources must be brought to bear. So in theory, the school board and administrators must spend sufficient money for that and other mandates to begin with, and then use the funds that remain as they see fit and as the voters want them to.

In practice (at least in her old school district: I have the impression that it is much better in the one she works in now, presumably in part because it is in a much more affluent suburban area, but also because it is in a region which puts much more stock in public education), the administration looks for every corner they can get away with cutting, unless and until the parent of a special education student knows the law and their child’s rights, and threatens to sue. At which point that child, and that child alone, will get the resources they are entitlled to, while other students get even fewer because the remaining resources are spread more thinly rather than being increased overall. Hence my skepticism.

But I’m afraid this is getting quite off topic!

ETA: Also in theory, the faculty and staff of the special education department could report the lack of compliance directly to the feds. But this was seen as a radical, even suicidal, option not even remotely worth considering. Whistleblowers, unfortunately, are almost always punished. Even if they aren’t fired their lives are made into hell until they quit.

[k9bfriender]

@SlackerInc

Special education funding is not supposed to be subject to the will of local voters (which does admittedly make it a partially unfunded mandate). There are federal laws and regulations which stipulate what resources must be brought to bear. So in theory, the school board and administrators must spend sufficient money for that and other mandates to begin with, and then use the funds that remain as they see fit and as the voters want them to.

In practice (at least in her old school district: I have the impression that it is much better in the one she works in now, presumably in part because it is in a much more affluent suburban area, but also because it is in a region which puts much more stock in public education), the administration looks for every corner they can get away with cutting, unless and until the parent of a special education student knows the law and their child’s rights, and threatens to sue. At which point that child, and that child alone, will get the resources they are entitlled to, while other students get even fewer because the remaining resources are spread more thinly rather than being increased overall. Hence my skepticism.

But I’m afraid this is getting quite off topic!

ETA: Also in theory, the faculty and staff of the special education department could report the lack of compliance directly to the feds. But this was seen as a radical, even suicidal, option not even remotely worth considering. Whistleblowers, unfortunately, are almost always punished. Even if they aren’t fired their lives are made into hell until they quit.

I don’t disagree that our educational system is lacking in many ways.

But, when the robots take over, that won’t be a concern anymore.

I do see computer aided learning as being something that will increase the education that pupil gets, while also reducing the cost. Whether your community values education enough to use those cost savings to further increase educational opportunities, build athletic fields, or simply save the taxpayer money has nothing to do with the technology being developed.

[SlackerInc]

@k9bfriender

I do see computer aided learning as being something that will increase the education that pupil gets, while also reducing the cost. Whether your community values education enough to use those cost savings to further increase educational opportunities, build athletic fields, or simply save the taxpayer money has nothing to do with the technology being developed.

I don’t disagree with that at all. As long as we move toward a “mincome” setup or something similar, I’m fine with developments like this that serve people better, even when they slash jobs.

[SamuelA]

@Mijin

We’re not “raving” about human workers. The point is just that sentient beings can very easy switch tasks. There is no kind of AI on the horizon that is sentient.
So that is an advantage that human workers have over AI, and it’s a big advantage in the messy real world (even if, of course, there are also some relative *disadvantages *of using human labor).

The economy. The size of the economy is not fixed, and indeed a higher-productivity AI replacing a human worker is an example of growth.

First of all, AIs don’t have to switch. You build a modified AI, using root libraries and techniques that are shared between all AI and periodically updated, to do any specific job*. At first it will only be economically feasible to replace a job that is both well defined, simple, and something that millions of people are employed doing. This is why autonomous cars are one of the first big applications for serious AI.

But as the tools get better and better, that number will shrink. Sooner or later it will be feasible to build an AI specifically for a job only 100k people are employed doing. Then 50k. Then 25k. Then eventually all the way down to maybe 100. (below that point, admittedly you will need self designing AIs)

Second, humans take years to learn a new skill to a level of competence worth paying for, and for older workers, it takes even longer. I feel like there’s going to be an era where people are losing their jobs, going back to school or starting at a low rank for that hot career the AIs haven’t gotten to yet (from truck driving to warehouse worker! Yeah!) and getting replaced before they make a dime.

*Honestly all our sci fi conceptions are just wrong. Imagine an object recognition library that all AIs use, or a planning library, etc. And some instances of an AI learn something new and add it to that library, and the updates get pushed back to the central repository, somehow filtered for quality, and then all AIs everywhere get better at that task.

So in a way every AI is this “gestalt” being sharing basic strategems with all the other AIs.

[SlackerInc]

Fascinating, Samuel. Are you a coder? It sounds like you really know this stuff.

[Snowboarder_Bo]

Not to cast aspersions on SamuelA, but what he’s described there is just the common sense outcome of the system that is being built right now.

AIs don’t forget something once they learn it unless they are told to. That means that once we train an AI, for instance, to be able to recognize a bird when it sees one, we no longer have to train an AI to do that ever again. AIs can just have the skill uploaded into them from a central data center/repository.

But that’s a pretty limited skill set, so we need to find ways to integrate skills into AIs. Once we can integrate 2 skill sets so they work together, it won’t take long to increase that number.

And the more algorithms that we can get AIs to run simultaneously, so they cooperate with each other to produce a desired result, we can start giving them more complex tasks to do.

Every success builds in a way that it can’t with people. And like he indicates, once we have a warehouse AI that’s smart enough to do the heavy lifting, it won’t take long to integrate a sorting algorithm, and then a scheduling algorithm etc, until the AIs are doing complex multi-task jobs.

If it takes two weeks to train an AI, it only takes a few seconds to teach that same thing to a new AI, to thousands of new AIs if necessary.

[XT]

Thought I’d link to this series on AJ+ that delves into this subject. I was going to post this in IMHO or another forum but saw this thread pop back up again on the radar so figured I’d post it here. Basically, this series talks about the future of automation and what jobs are likely to be automated in the near future and what the impact of those may mean. There are 4 videos in the series that discuss different aspects, I think they are all about 15 mins or so. It’s interesting stuff even if you know a bit about this. If you don’t’ know that much about it it will really help.

My overall take is to remind people that we have had these sorts of job displacements in the past. When they talk about 50% of jobs (or whatever) being automated, that certainly sounds scary, and in the short term for those people in those sectors it certainly is, but recall that at the turn of the last century 90+% of people were in the agricultural sector, while today it’s less than 3%. And just to understand, I personally believe that IT (what I do IOW) is moving towards more automation as well, and the job I do could certainly be automated or done by a good expert system or weak AI…and, in fact, I expect it WILL be in the next 10-20 years, tops.

Anyway, if anyone is interested in the videos feel free and post what you think.

[Mijin]

@SamuelA

First of all, AIs don’t have to switch. You build a modified AI, using root libraries and techniques that are shared between all AI and periodically updated, to do any specific job*. At first it will only be economically feasible to replace a job that is both well defined, simple, and something that millions of people are employed doing. This is why autonomous cars are one of the first big applications for serious AI.

But as the tools get better and better, that number will shrink. Sooner or later it will be feasible to build an AI specifically for a job only 100k people are employed doing. Then 50k. Then 25k. Then eventually all the way down to maybe 100. (below that point, admittedly you will need self designing AIs)

Second, humans take years to learn a new skill to a level of competence worth paying for, and for older workers, it takes even longer. I feel like there’s going to be an era where people are losing their jobs, going back to school or starting at a low rank for that hot career the AIs haven’t gotten to yet (from truck driving to warehouse worker! Yeah!) and getting replaced before they make a dime.

*Honestly all our sci fi conceptions are just wrong. Imagine an object recognition library that all AIs use, or a planning library, etc. And some instances of an AI learn something new and add it to that library, and the updates get pushed back to the central repository, somehow filtered for quality, and then all AIs everywhere get better at that task.

So in a way every AI is this “gestalt” being sharing basic strategems with all the other AIs.

No, I still disagree.
Let me expand upon what I was saying.

I’m saying that the typical office-worker (say) has three kinds of job:

1. Well-defined tasks that are similar day-to-day
2. Tasks that might be repeated, but have unique features and need a level of understanding e.g. fix the following bugs, write a risk analysis for a project proposal
3. Miscellaneous stuff the boss might ask you to do that might just be one-offs (e.g. See if our old installer works with .NET 4.5 installed, investigate whether Azure might be a better platform for our server etc)

Now tasks of type 1 are going to get steadily chipped away at by AI, and that’s a good thing. We’ll need fewer people to achieve the same productivity.
Tasks of type 2 and 3 will probably be assisted by AI at some level, but can’t really be taken over by it. Unless and until it’s essentially sentient, and can understand things like what our objectives are and form plans on how to achieve them.

Perhaps there are some jobs that are almost entirely type 1 tasks, and I would agree these are vulnerable to being automated out completely. But I think many here are underestimating how many jobs involve plenty of type 2 and 3 tasks.

[k9bfriender]

There will be jobs that do not get automated. High end restaurants and other services where people are willing to pay extra for the prestige of having it made by people, not robots.

There are some other jobs that do not really make themselves eas to automate. Hair salons, barbers, and of course, dog grooming, are some industries that, even if we managed to make the technology to do it, people would still be very uncomfortable with it.

Until or unless AI starts competing with us on creativity, those are the jobs that are most secure: Writers, artists, even performers.

The benefit of AI is that labor is usually the highest cost of anything you purchase. AI lowers labor costs, lowering costs of goods and services, making them more available to more people.

[XT]

I think some of you are thinking in strict either/or terms. Jobs will be automated and that means we don’t need humans, or jobs will be too tough to automate and thus will require humans only. My own take is that this is the wrong way to look at what’s coming. Basically, we’ve seen the automation routine already for years…what we are starting to see is automation coupled with AI. So, we aren’t talking strictly about some blue collar manufacturer jobs, but, say, doctors or lawyers whose jobs are potentially up for serious automation. Does that mean we won’t have doctors and lawyers anymore? What about my own job as an IT engineer…does automation mean we won’t need IT engineers anymore?

I think the answer is no, we will still need people…in fact, people might become more valuable. The automation and AI, I think, will actually be enhanced by having a human in the loop, both doing the parts that are to their strengths and filling the gaps in the weakness, basically giving us the best of both worlds.

But this won’t be a 1 for 1. As we’ve seen with automation throughout history, you have a task that takes 10,000 workers done by one that takes a thousand, done by one that takes 100 until you have a task being done by 10 that once took an entire workforce. So, what do we do with the 9,990 workers laid off or displaced because we don’t need them anymore? As I’ve said in other threads, my WAG is that they will be doing work that we don’t even know about today or things that we don’t consider to BE ‘work’. The best example I have for that is something like Twitch streamers or YouTube content providers. We pay people to play games or make interesting content that we can’t see or get from the traditional channels. There are all sorts of possibilities with this, especially with the new technologies. Perhaps, in the not so distant future, we will see people producing their own movies or shows that are better than what we get from Hollywood or other traditional entertainment venues. Or maybe the big thing will be the environment…perhaps people will have, as their job, to undo the damage to our environment. Or maybe it will be some other thing we haven’t even seen yet. Heck, perhaps it will be building space-based mega-structures and colonizing the solar system. Ever other time in history where you have a surplus of human labor we have eventually found a use for it, and I have no doubt we will this time as well. What that will be is hard to say, but to me, the possibilities are endless with the expansion of the very technologies that are placing those jobs at risk. Ironically, exactly the same thing that has happened with all the other automation displacements…the tools that render some jobs obsolete enable people to do things that weren’t even thought of before.

I do think we need to be thinking about the gaps that are surely coming soon. There are going to be some fairly significant displacements in the 10 to 20-year time frame, IMHO, with millions of people potentially affected. We should be considering our options today to what we will do about that when it happens.

[SamuelA]

@Mijin

I’m saying that the typical office-worker (say) has three kinds of job:

1. Well-defined tasks that are similar day-to-day
2. Tasks that might be repeated, but have unique features and need a level of understanding e.g. fix the following bugs, write a risk analysis for a project proposal
3. Miscellaneous stuff the boss might ask you to do that might just be one-offs (e.g. See if our old installer works with .NET 4.5 installed, investigate whether Azure might be a better platform for our server etc)

How about you name a type 2 or 3 task, and I’ll answer if I think the current methods can automate it, roughly how, and link you to a description of that method.

I agree there are such tasks. Generally where the AI is going to have trouble is if the “happiness of a human” is the desired outcome, or some other vaguely defined metric, or “sounding like another human” is part of the task.

[SlackerInc]

@XT

I think some of you are thinking in strict either/or terms. Jobs will be automated and that means we don’t need humans, or jobs will be too tough to automate and thus will require humans only. My own take is that this is the wrong way to look at what’s coming. Basically, we’ve seen the automation routine already for years…what we are starting to see is automation coupled with AI. So, we aren’t talking strictly about some blue collar manufacturer jobs, but, say, doctors or lawyers whose jobs are potentially up for serious automation. Does that mean we won’t have doctors and lawyers anymore? What about my own job as an IT engineer…does automation mean we won’t need IT engineers anymore?

I think the answer is no, we will still need people…in fact, people might become more valuable. The automation and AI, I think, will actually be enhanced by having a human in the loop, both doing the parts that are to their strengths and filling the gaps in the weakness, basically giving us the best of both worlds.

But this won’t be a 1 for 1. As we’ve seen with automation throughout history, you have a task that takes 10,000 workers done by one that takes a thousand, done by one that takes 100 until you have a task being done by 10 that once took an entire workforce. So, what do we do with the 9,990 workers laid off or displaced because we don’t need them anymore? As I’ve said in other threads, my WAG is that they will be doing work that we don’t even know about today or things that we don’t consider to BE ‘work’. The best example I have for that is something like Twitch streamers or YouTube content providers. We pay people to play games or make interesting content that we can’t see or get from the traditional channels. There are all sorts of possibilities with this, especially with the new technologies. Perhaps, in the not so distant future, we will see people producing their own movies or shows that are better than what we get from Hollywood or other traditional entertainment venues. Or maybe the big thing will be the environment…perhaps people will have, as their job, to undo the damage to our environment. Or maybe it will be some other thing we haven’t even seen yet. Heck, perhaps it will be building space-based mega-structures and colonizing the solar system. Ever other time in history where you have a surplus of human labor we have eventually found a use for it, and I have no doubt we will this time as well. What that will be is hard to say, but to me, the possibilities are endless with the expansion of the very technologies that are placing those jobs at risk. Ironically, exactly the same thing that has happened with all the other automation displacements…the tools that render some jobs obsolete enable people to do things that weren’t even thought of before.

I do think we need to be thinking about the gaps that are surely coming soon. There are going to be some fairly significant displacements in the 10 to 20-year time frame, IMHO, with millions of people potentially affected. We should be considering our options today to what we will do about that when it happens.

I strongly agree with the first part of this post. But the second part I think is flawed in essentially imagining that humans will have an endless ability to expand their desire for products of human labor without running into a tightfisted capitalist downsizing buzzsaw. If what you envision does come to pass and keeps employment at the same level it is now, I think that will be because government has come up with more and more makework jobs for people because they were unable to summon the political will (or lack the ideology) to go all the way to a mincome.

I would also point out that in the past transitions you refer to, there were always new jobs, at least eventually, for what we might unkindly call the “great unwashed masses”. Instead of having mass numbers of peasants working the fields, we had mass numbers of urban blue-collar workers on assembly lines. As those jobs have in turn diminished, we have created more work in shipping and truck driving and other transport type stuff, as well as washing dishes in restaurant kitchens, mopping floors, and so on. Those kinds of jobs will also be gone soon, while even the intellectual jobs we have now will require far few humans to do them when augmented with AI.

What you’re envisioning jobs consisting of in the future would only be suitable for creative intellectual or artistic types. The only way I can see that as being feasible is with widespread adoption of genetic engineering to significantly increase the IQ of even the lowest common denominator. Which is definitely possible.

I will take a look at those videos soon, thanks for posting them.

[XT]

@SlackerInc

But the second part I think is flawed in essentially imagining that humans will have an endless ability to expand their desire for products of human labor without running into a tightfisted capitalist downsizing buzzsaw. If what you envision does come to pass and keeps employment at the same level it is now, I think that will be because government has come up with more and more makework jobs for people because they were unable to summon the political will (or lack the ideology) to go all the way to a mincome.

Well, it’s definitely debatable. Historically, the surplus of labor lead to tight-fisted capitalist types finding things for that labor to do that would make said capitalists money, and I think this will be the case in the future as well. Someone in 1900 on a farm and being told that within less than 50 years 80% of farm jobs would be history would have been thinking exactly the same thing you and others in this thread are…or those who have been suffering due to downsizing and automation in the manufacturing sectors of the US for the past 3 decades, i.e. they couldn’t imagine what folks are possibly going to do, and everyone is going to be poor and out of work with only a few rich capitalists and a sea of poor people. The reality was that jobs that they didn’t even imagine in 1900 came along to use that idle labor…just as happened every couple of decades since then as jobs came and went. Sure, THIS time could be different and 90% (or whatever) of the labor force will be out of work and starving on the streets while the few rest labor for the fewer rich capitalists, but myself I don’t think that’s where we are headed. I actually think that this time will be different, but because I think the technology is going to free a lot of people from having to do some sort of traditional work…working 9-5 for 5 days a week doing some repetitive and boring job for 30+ years. I think that the very concept of what ‘employment’ and ‘job’ are going to shift, and people in 30 or 50 years will look back the same way we do at folks who worked 16 hour days 6 days a week in dirty and very dangerous jobs starting when they were kids and burning out to die at an early age.

Pure speculation, of course, but that’s how I see it. I think that high automation and deep, strong AI is going to free humans from a lot of traditional thinking and be a major paradigm shift in our society…as it’s already been shifting us in subtle but huge ways that people even 50 years ago would have trouble understanding or grasping. Who would have thought that you could make a 6 or 7 figure salary to play a game while people watched you play? Yet there are folks who do just that. Or that you could become one of the richest people on the planet creating a forum for people to electronically talk about their day, give their opinions on politics or whatever or post cat videos?

I will take a look at those videos soon, thanks for posting them.

I’ll be interested in seeing what you think. You will notice that they don’t conform exactly to what I’m saying, but I still think they are interesting and are more in line with what some in this thread seem to be positing. That I don’t agree completely doesn’t negate them, as I think having different points of view or assumptions is what makes things more interesting and thought provoking.

[Snowboarder_Bo]

@XT

Thought I’d link to this series on AJ+ that delves into this subject. I was going to post this in IMHO or another forum but saw this thread pop back up again on the radar so figured I’d post it here. Basically, this series talks about the future of automation and what jobs are likely to be automated in the near future and what the impact of those may mean. There are 4 videos in the series that discuss different aspects, I think they are all about 15 mins or so. It’s interesting stuff even if you know a bit about this. If you don’t’ know that much about it it will really help.

Thanks for posting this; I added it to my MPSIMS thread What’s new, Atlas? (AI, Robotics and tech thread).

I have the first video bookmarked and plan to watch them all later today.

[Voyager]

@SamuelA

But as the tools get better and better, that number will shrink. Sooner or later it will be feasible to build an AI specifically for a job only 100k people are employed doing. Then 50k. Then 25k. Then eventually all the way down to maybe 100. (below that point, admittedly you will need self designing AIs)

Maybe not 100. Machine learning requires an extensive training set. It may well be that a job only 100 people do does not have enough data to make machine learning feasible. You may need a truly intelligent system to handle that, which we won’t have to worry about for a while.

Here’s an example. Lots of chips get manufactured at the rate of 100,000 a day. We already have good systems that look at test data from the fabs and find problems long before any person could. You can often diagnose the problems, since they tend to repeat.
However, when you are debugging a new chip, you only have a few hundred at most, and many of the problems are seen for the first time. It is good to get lots of data about new parts, but there will never be enough of it to let a machine learning system work. Especially since people tend to fix the last bug on the next part before manufacture. Looking for a previous problem is like how generals always fight the last war.

[XT]

@Snowboarder_Bo

Thanks for posting this; I added it to my MPSIMS thread What’s new, Atlas? (AI, Robotics and tech thread).

I have the first video bookmarked and plan to watch them all later today.

Be interesting to see what your take is on it and whether you think it was helpful or relevant to the ongoing discussion. It’s definitely one of the more relevant of the current debates throughout the world, as the way it plays out will affect basically everyone, everywhere.