3 ms·

This is a weird jump. Usually technology bumps out the lowest class workers, but genAI is taking a bite right out of the middle to upper class... certain execu

This is a weird jump. Usually technology bumps out the lowest class workers, but genAI is taking a bite right out of the middle to upper class... certain executives, office salespeople, some programmers, actors, and such will be replaced by robots before garbage workers, lawn services or roofers.

nullc 3y ago
> This is a weird jump. https://marshallbrain.com/manna1 https://marshallbrain.com/manna1
ronsor 3y ago
Robotics is significantly harder and more expensive than algorithms, both in development and deployment.
- SoftTalker 3y ago
  And you still need people to maintain and repair the robots, at least for now. Software doesn't need maintenance in that sense; it doesn't have moving parts that need lubrication, adjustments, and that eventually wear out.
  - slashdev 3y ago
    As anyone who’s ever maintained software can tell you, software does suffer bitrot over time. Dependencies need updating and replacing. Deprecated functions have to be replaced. Systems and software change over time, breaking your software. It’s not as high maintenance as most physical things with moving parts though.
    - LeonB 3y ago
      Agreed. It’s really astounding how much maintenance “modern” software needs. A lot of the maintenance is to do with the dependencies, and/or with the discovery of vulnerabilities, but also with obsolescence in the tool chain used for building the software, or the APIs that the software depends upon, or simply the continual replacement of tokens/certificate and so on that have built in expiry. It’s not glamorous work, but it is absolutely relentless, endless, tedious stuff.
  - Ekaros 3y ago
    Also it doesn't murder itself and what is around it in case of bad inputs or failure. The failure modes of software are pretty tame, specially in cloud. With proper sized robots killing someone or damaging itself and other things is entirely real possibility.
    - brewdad 3y ago
      A few self driving cars would beg to differ. Software absolutely can kill in the meatspace world.
      - Ekaros 3y ago
        Those are robots... By any reasonable definition.
      - krisoft 3y ago
        Self driving cars are robots. There is a more academical definition: a robot is a computer with sensors and actuators attached. Self driving cars certanly fit that criteria. There is a more practical definition too: a robot is a machine with a computer which does not work yet. Because once it starts working people stop calling it a robot. For example a modern washing machine certainly fits the academic definition of being a robot, but most people wouldn’t call it one.
    - TheLoafOfBread 3y ago
      On embedded system software can murder itself relatively easily. * Wrong EEPROM emulation routine is using data flash (100k of writes) as EEPROM (10M of writes). * If you make a mistake during bootloader setup, bootloader can delete whole flash memory, including bootloader, or write over the bootloader. * You can unpack program into RAM (i.e. Compressed file on flash -> Unpack into RAM), so if you will screw up pointers, you can overwrite your own program on the fly without even noticing it.
- brucethemoose2 3y ago
  In hindsight it seems that way, but until recently, getting a robot to generate a movie or act as a creative therapist seemed impossible, not a matter of engineering resources.
- 1letterunixname 3y ago
  Yeap. The VA commissioned Dean Kamen to build the Luke Arm, a practical prosthetic arm for disabled vets. It took years, million$, determination, and extensive testing to achieve. Not impossible, but it's not something that can be thrown together in a weekend hackathon. https://mobiusbionics.com/luke-arm/ https://mobiusbionics.com/luke-arm/
qingcharles 3y ago
That's fascinating. I'd not thought of it that way, but it is absolutely true. When DALLE2 came along last year I was stunned as it seemed like we had skipped the part where AI did the drudgery and jumped right to the bit where it emulated human emotion and imagination. I know many people using AI to write vast swathes of content, to create marketing materials, to write code and to create stock images. Some of the people AI has replaced were taking $150K salaries and now are earning zero. That's a harder lump to swallow than your working class person who might have gone from $20K to zero.
1letterunixname 3y ago
"It is difficult to get a man to understand something, when his salary depends upon his not understanding it." -- Upton Sinclair It's my experience that coders, software engineers, accountants, and other STEM professionals regularly dismiss that their jobs are threatened by the encroachment of advances in AI such as deep learning and LLMs. And, unfortunately, the standard optimistic, hand-waving refrain of "There will be other jobs." denies reality that they may well make less money elsewhere, be under/unemployed, or not be able to retrain. I foresee a widening excess supply of skilled individuals who will face difficulties finding any job that either: pays enough OR where a potential employer doesn't deem them "overqualified." (In addition, almost all of the professional class bets against their own interests by being predominantly anti-union, even when they're under/haphazardly paid or overworked.) It will take more decades, but even nursing and care will be automatable when AI and mechatronics advance sufficiently. I look forward to the kitchen and bathroom cleaning robots. These are moonshot projects that require 100M-1B and multiple decades to attain. https://www.ted.com/talks/jeremy_howard_the_wonderful_and_terrifying_implications_of_computers_that_can_learn https://www.ted.com/talks/jeremy_howard_the_wonderful_and_te... (2014)
- julianeon 3y ago
  I am optimistic for all those professions you listed, because those are all knowledge workers. I like that term because the vagueness is actually helpful. Ultimately these people work to become experts in various knowledge systems. And the other thing that’s happening while AI is advancing, is that the educational system is borderline collapsing. So what that means is that, even if AI automates a ton more coding, even if it’s capabilities keep improving, the net effect of that will still be to add more complexity: and that means the jobs of people who are doing that type of job now, are going to get more complex. Someone has to make sure the AI data lake processor is feeding data into the chat room that displays daily metrics, for example. Now take that example and multiply it by 100, for all the new capabilities AI will create. So maybe you don’t write the app necessarily, because AI does, but because of the fragmented nature of various vendors and services, someone still has to make sure the app talks to the DB, which talks to the AI chat bot, which does a new kind of real time metrics analysis made possible by AI. I really don’t think the demand for the type of well educated worker it takes to manage all that is being met, and won’t be for a long time to come. If you can do that work, your job will be safe.

nullc 3y ago

> This is a weird jump. https://marshallbrain.com/manna1 https://marshallbrain.com/manna1

ronsor 3y ago

Robotics is significantly harder and more expensive than algorithms, both in development and deployment.

SoftTalker 3y ago
And you still need people to maintain and repair the robots, at least for now. Software doesn't need maintenance in that sense; it doesn't have moving parts that need lubrication, adjustments, and that eventually wear out.
- slashdev 3y ago
  As anyone who’s ever maintained software can tell you, software does suffer bitrot over time. Dependencies need updating and replacing. Deprecated functions have to be replaced. Systems and software change over time, breaking your software. It’s not as high maintenance as most physical things with moving parts though.
  - LeonB 3y ago
    Agreed. It’s really astounding how much maintenance “modern” software needs. A lot of the maintenance is to do with the dependencies, and/or with the discovery of vulnerabilities, but also with obsolescence in the tool chain used for building the software, or the APIs that the software depends upon, or simply the continual replacement of tokens/certificate and so on that have built in expiry. It’s not glamorous work, but it is absolutely relentless, endless, tedious stuff.
- Ekaros 3y ago
  Also it doesn't murder itself and what is around it in case of bad inputs or failure. The failure modes of software are pretty tame, specially in cloud. With proper sized robots killing someone or damaging itself and other things is entirely real possibility.
  - brewdad 3y ago
    A few self driving cars would beg to differ. Software absolutely can kill in the meatspace world.
    - Ekaros 3y ago
      Those are robots... By any reasonable definition.
    - krisoft 3y ago
      Self driving cars are robots. There is a more academical definition: a robot is a computer with sensors and actuators attached. Self driving cars certanly fit that criteria. There is a more practical definition too: a robot is a machine with a computer which does not work yet. Because once it starts working people stop calling it a robot. For example a modern washing machine certainly fits the academic definition of being a robot, but most people wouldn’t call it one.
  - TheLoafOfBread 3y ago
    On embedded system software can murder itself relatively easily. * Wrong EEPROM emulation routine is using data flash (100k of writes) as EEPROM (10M of writes). * If you make a mistake during bootloader setup, bootloader can delete whole flash memory, including bootloader, or write over the bootloader. * You can unpack program into RAM (i.e. Compressed file on flash -> Unpack into RAM), so if you will screw up pointers, you can overwrite your own program on the fly without even noticing it.
brucethemoose2 3y ago
In hindsight it seems that way, but until recently, getting a robot to generate a movie or act as a creative therapist seemed impossible, not a matter of engineering resources.
1letterunixname 3y ago
Yeap. The VA commissioned Dean Kamen to build the Luke Arm, a practical prosthetic arm for disabled vets. It took years, million$, determination, and extensive testing to achieve. Not impossible, but it's not something that can be thrown together in a weekend hackathon. https://mobiusbionics.com/luke-arm/ https://mobiusbionics.com/luke-arm/

qingcharles 3y ago

That's fascinating. I'd not thought of it that way, but it is absolutely true. When DALLE2 came along last year I was stunned as it seemed like we had skipped the part where AI did the drudgery and jumped right to the bit where it emulated human emotion and imagination. I know many people using AI to write vast swathes of content, to create marketing materials, to write code and to create stock images. Some of the people AI has replaced were taking $150K salaries and now are earning zero. That's a harder lump to swallow than your working class person who might have gone from $20K to zero.

1letterunixname 3y ago

"It is difficult to get a man to understand something, when his salary depends upon his not understanding it." -- Upton Sinclair It's my experience that coders, software engineers, accountants, and other STEM professionals regularly dismiss that their jobs are threatened by the encroachment of advances in AI such as deep learning and LLMs. And, unfortunately, the standard optimistic, hand-waving refrain of "There will be other jobs." denies reality that they may well make less money elsewhere, be under/unemployed, or not be able to retrain. I foresee a widening excess supply of skilled individuals who will face difficulties finding any job that either: pays enough OR where a potential employer doesn't deem them "overqualified." (In addition, almost all of the professional class bets against their own interests by being predominantly anti-union, even when they're under/haphazardly paid or overworked.) It will take more decades, but even nursing and care will be automatable when AI and mechatronics advance sufficiently. I look forward to the kitchen and bathroom cleaning robots. These are moonshot projects that require 100M-1B and multiple decades to attain. https://www.ted.com/talks/jeremy_howard_the_wonderful_and_terrifying_implications_of_computers_that_can_learn https://www.ted.com/talks/jeremy_howard_the_wonderful_and_te... (2014)

julianeon 3y ago
I am optimistic for all those professions you listed, because those are all knowledge workers. I like that term because the vagueness is actually helpful. Ultimately these people work to become experts in various knowledge systems. And the other thing that’s happening while AI is advancing, is that the educational system is borderline collapsing. So what that means is that, even if AI automates a ton more coding, even if it’s capabilities keep improving, the net effect of that will still be to add more complexity: and that means the jobs of people who are doing that type of job now, are going to get more complex. Someone has to make sure the AI data lake processor is feeding data into the chat room that displays daily metrics, for example. Now take that example and multiply it by 100, for all the new capabilities AI will create. So maybe you don’t write the app necessarily, because AI does, but because of the fragmented nature of various vendors and services, someone still has to make sure the app talks to the DB, which talks to the AI chat bot, which does a new kind of real time metrics analysis made possible by AI. I really don’t think the demand for the type of well educated worker it takes to manage all that is being met, and won’t be for a long time to come. If you can do that work, your job will be safe.