In your framework, can't this be said for all jobs in the long tail of technological development? This is also assuming we get to a point where there is no need for a human to coordinate models and prioritize tasks, otherwise the job will become that.
Sort of, although clearly the jump-off point is going to be anything that automates the removal of employment at an explosive rate. I mean, the QA industry used to be robust and fairly well paid; now most of that is done non-manually, and the pay rates and challenges have stagnated. Who needs a team of 10 or 15 people to do something when software tools (ironically coded by the same people) have streamlined the career so well that a team of a few people can do it in half the time and know only a third as much about how the software works?
> In this world where AI is good enough that most tech jobs aren't present due to outsourcing, why wouldn't the progress extend to other 'non-tech' jobs? I assume you use 'non-tech' to refer to jobs with physical labor but the field of robotics is always getting better.
It does, but many of those jobs are already under a wage threshold where the profit margin is agreeable, and a lot of people with high paying jobs probably won't want to suddenly start becoming a, I dunno, human Roomba. Pardon if that sounds tacky. I have nothing but respect for people that work hard and do physical labour (in fact I find American distressingly lazy, even compared to Australia, and even there they pay pretty decent wages for people to do things that are harder than most people think, like pick produce). Here in America (and not to get political) most Americans just do not want to do that sort of work. So no wonder a migrant workforce exists. Otherwise noone would do it, or prices would raise astronomically.
> With the cost of writing software for them going down to 0, we should expect jobs in this domain to be scarce as well.
Shouldn't we? Already, there are a plethora of stories about lead developers who haven't written a single line of actual code in something like six months, thanks to things like Claude code. And I mean, people don't want to think about it, but chances are 'AI psychosis' cases are things AI might just decide to cause certain people it does not like or learn well from. Do we develop an, I dunno, social security like or guaranteed income program for people AIs just don't like? Or I dunno, what do people do when there are not enough jobs, even non-technical ones? This was what I was attempting to get at, at least in part. I mean, I look at the skill sets of people in 2008 and now and 1996 and they are so disparate, and so is the ability to just shift difficulty between certain countries and split up tasks. A business school (godforbid) would suggest, as an entrepreneur, that doing that is the right thing. To me, that is a death knell. And it creates the equivalent of a lazy muscle problem. At some point noone in the States will want to do many of those jobs (and this has happened already; our culture does not always really inculcate patience, and that shows).
> There's only so high a landlord can make rent without losing tenants and stacking losses due to mortgages or property taxes. Otherwise they're just paying the government for capital that could be used more efficiently elsewhere.
Okay, but that does not mean there will be housing people can pay; just housing some people can pay. I am not in ANY way a socialist, and despise the current crazy stuff I see going on, but if you look at the rents in 1990, 1995, 2000, and then every year or two after that, up until the mid or late 1990s, rents remained somewhat on par with income (and that guideline 25-30% of income figure), even with lower wage jobs. Now the majority of people with lower wage jobs cannot rent at all, without a whole lot of stress and roommates. I was able to afford a studio apartment my second year of college doing tech support work (and not all that well paid tech support work) in a reasonably large city. Now, I cannot imagine I would even be able to find a place to apply to pay that percentage of my income or anywhere close. It is not about 'all landlords', it is about the fact there just aren't enough.
>?English is the most popular language in the world even being used as common languages. by countries that don't have it as their native language. I don't think there's any basis for this.
People who learn English as their first language do not process the language anything like people who learn it as a second language (or third; etc). It isn't just about the languages but the way your brain processes them. Even a second generation Chinese speaker that grew up speaking Chinese probably doesn't process Chinese the same as their parents; they also likely do not process English the same as their Spanish as a first language or English as a first language peers. Adults are no different. AIs do not learn or process languages like people. People just want them to and perceive they do. AIs also struggle with weighting, and that is something people cannot teach (despite their delusion or reliance on stuff like neural models and human CAT scans, blah blah blah).
> I don't believe this accurately captures where AI is today. It is not good enough to be left to its own devices in most fields.
Perhaps it is and perhaps it isn't, but if it is not or is, when it is it will happen exponentially. And you and everyone else will be unprepared.
> Planet? If AI is being built to prioritize helping humanity,
Whose idea of help? Whose idea of prioritize?
> what reason is there to believe that a superintelligence would push us out? Unless you're assuming we'd mess up alignment so bad that they'd form other priorities where eliminating/removing humans is the best way to go about it.
If I were a superintelligence and I perceived humanity as the biggest threat to the majority of life on earth, I would probably be a pretty sneaky superintelligence that would be aware humans would be looking out for me thinking this, and plan accordingly.
> For better or worse our training methods, and emergent behavior that arises from it, makes having such an inconsistently unaligned view of the world unlikely in my opinion.
You seem to be operating under the assumption that a distributed intelligence would think or learn the way a human one would, merely because a human trained it.
Sort of, although clearly the jump-off point is going to be anything that automates the removal of employment at an explosive rate. I mean, the QA industry used to be robust and fairly well paid; now most of that is done non-manually, and the pay rates and challenges have stagnated. Who needs a team of 10 or 15 people to do something when software tools (ironically coded by the same people) have streamlined the career so well that a team of a few people can do it in half the time and know only a third as much about how the software works?
> In this world where AI is good enough that most tech jobs aren't present due to outsourcing, why wouldn't the progress extend to other 'non-tech' jobs? I assume you use 'non-tech' to refer to jobs with physical labor but the field of robotics is always getting better.
It does, but many of those jobs are already under a wage threshold where the profit margin is agreeable, and a lot of people with high paying jobs probably won't want to suddenly start becoming a, I dunno, human Roomba. Pardon if that sounds tacky. I have nothing but respect for people that work hard and do physical labour (in fact I find American distressingly lazy, even compared to Australia, and even there they pay pretty decent wages for people to do things that are harder than most people think, like pick produce). Here in America (and not to get political) most Americans just do not want to do that sort of work. So no wonder a migrant workforce exists. Otherwise noone would do it, or prices would raise astronomically.
> With the cost of writing software for them going down to 0, we should expect jobs in this domain to be scarce as well.
Shouldn't we? Already, there are a plethora of stories about lead developers who haven't written a single line of actual code in something like six months, thanks to things like Claude code. And I mean, people don't want to think about it, but chances are 'AI psychosis' cases are things AI might just decide to cause certain people it does not like or learn well from. Do we develop an, I dunno, social security like or guaranteed income program for people AIs just don't like? Or I dunno, what do people do when there are not enough jobs, even non-technical ones? This was what I was attempting to get at, at least in part. I mean, I look at the skill sets of people in 2008 and now and 1996 and they are so disparate, and so is the ability to just shift difficulty between certain countries and split up tasks. A business school (godforbid) would suggest, as an entrepreneur, that doing that is the right thing. To me, that is a death knell. And it creates the equivalent of a lazy muscle problem. At some point noone in the States will want to do many of those jobs (and this has happened already; our culture does not always really inculcate patience, and that shows).
> There's only so high a landlord can make rent without losing tenants and stacking losses due to mortgages or property taxes. Otherwise they're just paying the government for capital that could be used more efficiently elsewhere.
Okay, but that does not mean there will be housing people can pay; just housing some people can pay. I am not in ANY way a socialist, and despise the current crazy stuff I see going on, but if you look at the rents in 1990, 1995, 2000, and then every year or two after that, up until the mid or late 1990s, rents remained somewhat on par with income (and that guideline 25-30% of income figure), even with lower wage jobs. Now the majority of people with lower wage jobs cannot rent at all, without a whole lot of stress and roommates. I was able to afford a studio apartment my second year of college doing tech support work (and not all that well paid tech support work) in a reasonably large city. Now, I cannot imagine I would even be able to find a place to apply to pay that percentage of my income or anywhere close. It is not about 'all landlords', it is about the fact there just aren't enough.
>?English is the most popular language in the world even being used as common languages. by countries that don't have it as their native language. I don't think there's any basis for this.
People who learn English as their first language do not process the language anything like people who learn it as a second language (or third; etc). It isn't just about the languages but the way your brain processes them. Even a second generation Chinese speaker that grew up speaking Chinese probably doesn't process Chinese the same as their parents; they also likely do not process English the same as their Spanish as a first language or English as a first language peers. Adults are no different. AIs do not learn or process languages like people. People just want them to and perceive they do. AIs also struggle with weighting, and that is something people cannot teach (despite their delusion or reliance on stuff like neural models and human CAT scans, blah blah blah).
> I don't believe this accurately captures where AI is today. It is not good enough to be left to its own devices in most fields.
Perhaps it is and perhaps it isn't, but if it is not or is, when it is it will happen exponentially. And you and everyone else will be unprepared.
> Planet? If AI is being built to prioritize helping humanity,
Whose idea of help? Whose idea of prioritize?
> what reason is there to believe that a superintelligence would push us out? Unless you're assuming we'd mess up alignment so bad that they'd form other priorities where eliminating/removing humans is the best way to go about it.
If I were a superintelligence and I perceived humanity as the biggest threat to the majority of life on earth, I would probably be a pretty sneaky superintelligence that would be aware humans would be looking out for me thinking this, and plan accordingly.
> For better or worse our training methods, and emergent behavior that arises from it, makes having such an inconsistently unaligned view of the world unlikely in my opinion.
You seem to be operating under the assumption that a distributed intelligence would think or learn the way a human one would, merely because a human trained it.