12 ms·

> AI won't take a thought workers job. I don't understand how you can hold this position considering not only where we are but also the rate of progress in AI.

> AI won't take a thought workers job. I don't understand how you can hold this position considering not only where we are but also the rate of progress in AI. It has been taking thought workers job and it will increasingly do so in the future until the surplus cognitive workers offer is just so minimal that it's not worth to automate it.

kodah 4y ago
What jobs have been totally replaced thus far?
- fxtentacle 4y ago
  Didn't all major companies do mass-firings recently? What do you expect will happen if an entry-level software engineer becomes 2x as productive? You need 2x less, so you fire half the team.
  - spoiler 4y ago
    A smart company will realise it's better to be 2x productive for the same money than to be at the same productivity for half the money.
    - jokethrowaway 4y ago
      Wrong, there's just so much a company can dedicate itself too, otherwise you'll start doing little useless experiments that go nowhere like google
      - spoiler 4y ago
        I think Google has a weird pathological case maybe. So, I'm not sure I would use them as a forecast. I think R&D is good in any company. Sure, the nature of R&D is that most of it is failure; that's just the way it is. AI makes the R&D iteration shorter and cheaper, IMO. The more you try, the more likely you're to succeed (aside: I should listen to my own advice sometimes lol).
- jay_kyburz 4y ago
  Companies are slow to move on things, but I think "video game concept artist" is done. An art director can now sit down in front of Midjourny and pump out years and years of a concept artists work in just an afternoon.
  - whateveracct 4y ago
    This feels like a comment written by someone without practical experience with concept art (creator or consumer) - correct me if I'm wrong though! Like personally, I think I will still draw concept art instead of asking a GAN. Feels like that greases my creative wheels more. Just me though.
    - coldtea 4y ago
      Whether you have experience or not, game companies already use AI for concept art.
      - lunfard000 4y ago
        not only art, Obsidian is using already for voices[0]. The future is here. 0:https://www.youtube.com/watch?v=YajBa5PO1Hk https://www.youtube.com/watch?v=YajBa5PO1Hk
    - jay_kyburz 4y ago
      You are wrong, I have worked for many years as an artists and level builder in video games. Bioshock is the most famous game, and Void Bastards the most recent.
      - spoiler 4y ago
        Could you instead use AI to reduce drudgery in your work, or to iterate faster between ideas until you settle on a final peace? What if you have something in mind that is hard to prompt for, or the tool just refuses to generate what you had in mind? I would try to lean into it as a tool to speed up some of the tedium, instead of fear it, if possible
- coldtea 4y ago
  First, of all, why it has to be "totally replaced"? It's enough for positions to be dramatically reduced, and most such "though workers" tasks getting done automatically, e.g. under the final supervision of much fewer people. Second, why it has to be "thus far"? If it happens in 1 or 5 or 10 years does it disqualify it? Because above you claimed it will "never take thought workers' jobs". Suddenly the goalposts moved so that it has to have happened "totally" and "thus far" to qualify as taking their jobs?
  - CommanderData 4y ago
    I'm witnessing engineers on my team and myself saving a lot of time in development and in operations (scripting). I have no doubt in my mind it is reducing the need for more engineers and has probably already contributed to a reduction of them and I think it's just getting started.
    - saurik 4y ago
      Compilers and high-level languages assuredly reduced the number of engineers needed, right? No: we just started making much more complicated things and the basic level of "this isn't even a serviceable solution" has gone up so high that if you sat around and wrote a trivial program designed to be used over a teletype--which is got much easier as you can do it in C instead of assembly--you'd lose in the market to your competitors who are willing to put in the effort to build a modern fancy reactive single-page whatever's-hot-now web app. The reality is that companies are going to continue allocating a similar percentage of their budget to tech that they were before, and what we DO might change a ton; but, frankly, I've been doing this now for decades, and the stuff I used to do 25 years ago (omg: a full quarter of a century of professional software development ;P) already often just isn't a thing anyone does anymore. Every time my job gets easier, the expectations for the products I architect also increase, because we're all in a rat race attempting to compete for the same customers to defend our margins. Hell: I remember when the first step to deploying a new service--even if you were a company that only had a handful of people--was to attempt to predict how much capacity you needed and then buy a ton of parts so you'd be able to sit around and build machines for your server room to run your new product. I was involved in a project back in like, 1998/1999, for a company with all of maybe five people, and someone's job was just sitting around building the server room... that specific task is now a few lines of YAML configuration and a couple of accounts (AWS and Twilio <- back in the day, we had to have custom T1 lines brought in not for data but for bulk parallel phone service to hook up to the Dialogic boards we needed to simulate phone lines ;P). The concept of cloud-computing entirely changed that part of the business--having a small handful of companies manage all the servers at scale and run numerous jobs on shared hardware, in some sense reducing the cost to the point where you can both outsource this and pay less--to the point where doing that almost sounds insane; and yet, I'm sure there are more people getting paid more to deploy and manage ever more virtualized servers in the cloud than there ever were people (including me!) literally screwing around with screwdrivers in the ubiquitous server rooms. You could argue that my job got "replaced", with all of the skills I had to deal with IRQ balancing and automated power backup systems being for naught... but, somehow, I'm apparently still relevant. I'll also admit that I honestly kind of enjoyed doing that work?... and yet, I also was quite happy to never have to do it ever again. I'll never forget the day I took the first real-world server I was in charge of and just like, uploaded it to 1&1 where I had an early software-based hypervisor-like stack I was experimenting with (UML, aka User Mode Linux, which is no longer something people bother with due to the advent of Xen/KVM/etc.); it just worked and I was like "omg I'm never building another server again, am I? that was it: it's over". (It then turned out that I did build another complex hardware solution in 2016 to start doing work with GPUs, but there was only a small window of time there before cloud services caught up.) So like, sure: at some point the AI is going to entirely replace what we do--not the specific frustrating coding tasks we do today, but the entire concept of our jobs--but, the issue at that point isn't going to be me figuring out what I'm going to do next to make ends meat... it is going to be that I'm not a serviceable warrior in the war against the machines as the world devolves into some kind of Terminator-inspired hellscape; because, if you can manage to do my job in its totality, then you frankly can also do the job of everyone who has ever hired me and you can do the job of every investor who has ever given us money to do those things... we are all obsolete and, if AI actually gets as good as so many people here fear it will, at best we all become pets :(.
- vasco 4y ago
  If performing a job is even just 10% faster, you can fire 1 person out of a team of 10 (or need to hire in the first place). This should be obvious.
  - jayd16 4y ago
    First, this argument applies to everything from WFH adding efficiency to computers themselves being faster than ink and paper. But that's not how the economics works out. The business can simply choose to sell 10% more content.
    - wwweston 4y ago
      At some point, you saturate the market for whatever’s being produced and then prices drop and you have diminishing marginal returns for production so the economics of making more doesn’t make sense.
  - raincole 4y ago
    I'm sure most programmers work more than 10% faster with syntax highlighting and autocomplete. But no one says IDE replaced humans' jobs.
    - TeMPOraL 4y ago
      They did, and so did high-level languages, and not to mention the entire profession is first and foremost about automating other peoples' jobs away. Thing is, nobody is really tracking how improvements for software developers are replacing people, because this only manifests at slowing the absurdly high growth rate in this industry.
    - vasco 4y ago
      Yes they did, what are you thinking would have happened if all we had was assembly? To build any meaningful software you'd need armies of people and it'd take way longer and it wouldn't be as good.
  - cthalupa 4y ago
- happytiger 4y ago
  I have seen at least one outsourced programming job slow down to the point where it’s effectively no longer outsourced. Previously there were two engineers working, and now there is one as the work isn’t needed anymore. That in a very real way is a replaced job. The remaining job is running large scale data loading for a startup and really now is just eyeballs to make sure errors are caught. No longer requires much skill either. I think some degree of low end engineering work in the outsourcing world is at risk.
- code_witch_sam 4y ago
  paralegals seem like they're very much at risk
- ericathegreat 4y ago
  There are still professional, high quality hand knitters in the world. Of course, having people hand knit garments used to be the only way to get a knitted garment at all. Then we invented industrial knitting machines, and those hand knitters found their roles had changed. Instead of knitting a whole garment, they would be closing up the toe on the socks, or doing finishing work on a sweater. Of course, companies didn't need anywhere near as many knitters under this system, so a huge proportion of them lost their jobs. Then the knitting machines got better. They could close the toes on the socks themselves, could do most of the finishing work automatically. Some of the remaining knitters became industrial knitting support workers, but most of the actual knitting jobs dried up. But there are still professional, commercial hand knitters, even today! They test hand knitting patterns for the hobby market. Make the samples up for photographing, and make sure that all the sizes come out right. They number... Dozens? Maybe? And most of them treat it as a side gig, despite being the absolute pinnacle of hand knitting talent, since it pays terribly. A job doesn't have to have been totally replaced to be effectively replaced. As we find ways to hand over larger and larger pieces of the work to an automated system, the number of real roles in that field diminishes, until it eventually becomes infeasible as a career choice. This is what a lot these digital content creation jobs are heading. Gradual obsolescence.
Pandabob 4y ago
Lots of reasonable people seem to hold this view. Sam Altman himself mentioned believing this in the Lex Friedman podcast. Here are tweets from famous HN users Simon Willison[0] and Patrick Mckenzie[1] endorsing the viewpoint as well. [0]: https://twitter.com/simonw/status/1639692312585572352 https://twitter.com/simonw/status/1639692312585572352 [1]: https://twitter.com/patio11/status/1639787193580519424 https://twitter.com/patio11/status/1639787193580519424
- rafaelero 4y ago
  I hope they are right. But one thing that seems to be conflated in this debate is the comparison of AI with previous disruptions. The tools we have built before were taking some specific parts of the process and making it more productive. They were not trying to mimic human cognition in a way that can be applied to so many different processes at the same time.

kodah 4y ago

What jobs have been totally replaced thus far?

fxtentacle 4y ago
Didn't all major companies do mass-firings recently? What do you expect will happen if an entry-level software engineer becomes 2x as productive? You need 2x less, so you fire half the team.
- spoiler 4y ago
  A smart company will realise it's better to be 2x productive for the same money than to be at the same productivity for half the money.
  - jokethrowaway 4y ago
    Wrong, there's just so much a company can dedicate itself too, otherwise you'll start doing little useless experiments that go nowhere like google
    - spoiler 4y ago
      I think Google has a weird pathological case maybe. So, I'm not sure I would use them as a forecast. I think R&D is good in any company. Sure, the nature of R&D is that most of it is failure; that's just the way it is. AI makes the R&D iteration shorter and cheaper, IMO. The more you try, the more likely you're to succeed (aside: I should listen to my own advice sometimes lol).
jay_kyburz 4y ago
Companies are slow to move on things, but I think "video game concept artist" is done. An art director can now sit down in front of Midjourny and pump out years and years of a concept artists work in just an afternoon.
- whateveracct 4y ago
  This feels like a comment written by someone without practical experience with concept art (creator or consumer) - correct me if I'm wrong though! Like personally, I think I will still draw concept art instead of asking a GAN. Feels like that greases my creative wheels more. Just me though.
  - coldtea 4y ago
    Whether you have experience or not, game companies already use AI for concept art.
    - lunfard000 4y ago
      not only art, Obsidian is using already for voices[0]. The future is here. 0:https://www.youtube.com/watch?v=YajBa5PO1Hk https://www.youtube.com/watch?v=YajBa5PO1Hk
  - jay_kyburz 4y ago
    You are wrong, I have worked for many years as an artists and level builder in video games. Bioshock is the most famous game, and Void Bastards the most recent.
    - spoiler 4y ago
      Could you instead use AI to reduce drudgery in your work, or to iterate faster between ideas until you settle on a final peace? What if you have something in mind that is hard to prompt for, or the tool just refuses to generate what you had in mind? I would try to lean into it as a tool to speed up some of the tedium, instead of fear it, if possible
coldtea 4y ago
First, of all, why it has to be "totally replaced"? It's enough for positions to be dramatically reduced, and most such "though workers" tasks getting done automatically, e.g. under the final supervision of much fewer people. Second, why it has to be "thus far"? If it happens in 1 or 5 or 10 years does it disqualify it? Because above you claimed it will "never take thought workers' jobs". Suddenly the goalposts moved so that it has to have happened "totally" and "thus far" to qualify as taking their jobs?
- CommanderData 4y ago
  I'm witnessing engineers on my team and myself saving a lot of time in development and in operations (scripting). I have no doubt in my mind it is reducing the need for more engineers and has probably already contributed to a reduction of them and I think it's just getting started.
  - saurik 4y ago
    Compilers and high-level languages assuredly reduced the number of engineers needed, right? No: we just started making much more complicated things and the basic level of "this isn't even a serviceable solution" has gone up so high that if you sat around and wrote a trivial program designed to be used over a teletype--which is got much easier as you can do it in C instead of assembly--you'd lose in the market to your competitors who are willing to put in the effort to build a modern fancy reactive single-page whatever's-hot-now web app. The reality is that companies are going to continue allocating a similar percentage of their budget to tech that they were before, and what we DO might change a ton; but, frankly, I've been doing this now for decades, and the stuff I used to do 25 years ago (omg: a full quarter of a century of professional software development ;P) already often just isn't a thing anyone does anymore. Every time my job gets easier, the expectations for the products I architect also increase, because we're all in a rat race attempting to compete for the same customers to defend our margins. Hell: I remember when the first step to deploying a new service--even if you were a company that only had a handful of people--was to attempt to predict how much capacity you needed and then buy a ton of parts so you'd be able to sit around and build machines for your server room to run your new product. I was involved in a project back in like, 1998/1999, for a company with all of maybe five people, and someone's job was just sitting around building the server room... that specific task is now a few lines of YAML configuration and a couple of accounts (AWS and Twilio <- back in the day, we had to have custom T1 lines brought in not for data but for bulk parallel phone service to hook up to the Dialogic boards we needed to simulate phone lines ;P). The concept of cloud-computing entirely changed that part of the business--having a small handful of companies manage all the servers at scale and run numerous jobs on shared hardware, in some sense reducing the cost to the point where you can both outsource this and pay less--to the point where doing that almost sounds insane; and yet, I'm sure there are more people getting paid more to deploy and manage ever more virtualized servers in the cloud than there ever were people (including me!) literally screwing around with screwdrivers in the ubiquitous server rooms. You could argue that my job got "replaced", with all of the skills I had to deal with IRQ balancing and automated power backup systems being for naught... but, somehow, I'm apparently still relevant. I'll also admit that I honestly kind of enjoyed doing that work?... and yet, I also was quite happy to never have to do it ever again. I'll never forget the day I took the first real-world server I was in charge of and just like, uploaded it to 1&1 where I had an early software-based hypervisor-like stack I was experimenting with (UML, aka User Mode Linux, which is no longer something people bother with due to the advent of Xen/KVM/etc.); it just worked and I was like "omg I'm never building another server again, am I? that was it: it's over". (It then turned out that I did build another complex hardware solution in 2016 to start doing work with GPUs, but there was only a small window of time there before cloud services caught up.) So like, sure: at some point the AI is going to entirely replace what we do--not the specific frustrating coding tasks we do today, but the entire concept of our jobs--but, the issue at that point isn't going to be me figuring out what I'm going to do next to make ends meat... it is going to be that I'm not a serviceable warrior in the war against the machines as the world devolves into some kind of Terminator-inspired hellscape; because, if you can manage to do my job in its totality, then you frankly can also do the job of everyone who has ever hired me and you can do the job of every investor who has ever given us money to do those things... we are all obsolete and, if AI actually gets as good as so many people here fear it will, at best we all become pets :(.
vasco 4y ago
If performing a job is even just 10% faster, you can fire 1 person out of a team of 10 (or need to hire in the first place). This should be obvious.
- jayd16 4y ago
  First, this argument applies to everything from WFH adding efficiency to computers themselves being faster than ink and paper. But that's not how the economics works out. The business can simply choose to sell 10% more content.
  - wwweston 4y ago
    At some point, you saturate the market for whatever’s being produced and then prices drop and you have diminishing marginal returns for production so the economics of making more doesn’t make sense.
- raincole 4y ago
  I'm sure most programmers work more than 10% faster with syntax highlighting and autocomplete. But no one says IDE replaced humans' jobs.
  - TeMPOraL 4y ago
    They did, and so did high-level languages, and not to mention the entire profession is first and foremost about automating other peoples' jobs away. Thing is, nobody is really tracking how improvements for software developers are replacing people, because this only manifests at slowing the absurdly high growth rate in this industry.
  - vasco 4y ago
    Yes they did, what are you thinking would have happened if all we had was assembly? To build any meaningful software you'd need armies of people and it'd take way longer and it wouldn't be as good.
- cthalupa 4y ago
happytiger 4y ago
I have seen at least one outsourced programming job slow down to the point where it’s effectively no longer outsourced. Previously there were two engineers working, and now there is one as the work isn’t needed anymore. That in a very real way is a replaced job. The remaining job is running large scale data loading for a startup and really now is just eyeballs to make sure errors are caught. No longer requires much skill either. I think some degree of low end engineering work in the outsourcing world is at risk.
code_witch_sam 4y ago
paralegals seem like they're very much at risk
ericathegreat 4y ago
There are still professional, high quality hand knitters in the world. Of course, having people hand knit garments used to be the only way to get a knitted garment at all. Then we invented industrial knitting machines, and those hand knitters found their roles had changed. Instead of knitting a whole garment, they would be closing up the toe on the socks, or doing finishing work on a sweater. Of course, companies didn't need anywhere near as many knitters under this system, so a huge proportion of them lost their jobs. Then the knitting machines got better. They could close the toes on the socks themselves, could do most of the finishing work automatically. Some of the remaining knitters became industrial knitting support workers, but most of the actual knitting jobs dried up. But there are still professional, commercial hand knitters, even today! They test hand knitting patterns for the hobby market. Make the samples up for photographing, and make sure that all the sizes come out right. They number... Dozens? Maybe? And most of them treat it as a side gig, despite being the absolute pinnacle of hand knitting talent, since it pays terribly. A job doesn't have to have been totally replaced to be effectively replaced. As we find ways to hand over larger and larger pieces of the work to an automated system, the number of real roles in that field diminishes, until it eventually becomes infeasible as a career choice. This is what a lot these digital content creation jobs are heading. Gradual obsolescence.

Pandabob 4y ago

Lots of reasonable people seem to hold this view. Sam Altman himself mentioned believing this in the Lex Friedman podcast. Here are tweets from famous HN users Simon Willison[0] and Patrick Mckenzie[1] endorsing the viewpoint as well. [0]: https://twitter.com/simonw/status/1639692312585572352 https://twitter.com/simonw/status/1639692312585572352 [1]: https://twitter.com/patio11/status/1639787193580519424 https://twitter.com/patio11/status/1639787193580519424

rafaelero 4y ago
I hope they are right. But one thing that seems to be conflated in this debate is the comparison of AI with previous disruptions. The tools we have built before were taking some specific parts of the process and making it more productive. They were not trying to mimic human cognition in a way that can be applied to so many different processes at the same time.