4 ms·

I've been having similar thoughts about software engineering since one effect of LLMs is they make senior devs very productive, they make junior devs superfluou

I've been having similar thoughts about software engineering since one effect of LLMs is they make senior devs very productive, they make junior devs superfluous, but being a junior dev first is how you become a senior dev. But I wonder if the outcome is simply that we drop every known manufacturing technique on the floor and just start from scratch with current adjacent technologies. Basically kill the whole industry and reboot it again. My sister was a petroleum geologist. She went into oil in the early 2000s because she saw the roughly 30,000 person shortfall of petroleum geologists that was about to happen as the baby boomers aged out of the profession, and was like "Well, they're going to need to hire new blood, that's good for me." And it worked great for about 5 years, she was paid a shit-ton of money because they couldn't get people. But then two layoffs later, what actually happened is that the entire oil industry and associated value chain is dying, and we're replacing it with electricity, solar, batteries, EVs, smart-grids, and a bunch of things that didn't exist in the early 2000s. Maybe the same thing happens to manufacturing, and we just get rid of craftsmanship, machining, and logistics, and have everybody 3D-print their appliances in a factory-in-a-box they keep at their home, just shipping filament and chips and other raw materials directly to them.

johnnyanmac 1y ago
>'ve been having similar thoughts about software engineering since one effect of LLMs is they make senior devs very productive, they make junior devs superfluous Well that's the pitch they want to throw at you. Actual studies disagree. >Maybe the same thing happens to manufacturing, and we just get rid of craftsmanship, machining, and logistics, and have everybody 3D-print their appliances in a factory-in-a-box they keep at their home, just shipping filament and chips and other raw materials directly to them. I legitmately think we will hit General AI before we have anything close to this. We simply don't have the resources needed on an individual bases to faciliate everyone being a "factory worker at home".
- Our_Benefactors 1y ago
  Look at how miserably AI fails at any kind of CAD work and you may reconsider.
  - idiotsecant 1y ago
    There's no reason to think this will continue to be the case for long.
    - ethanwillis 1y ago
      none? not a single reason at all?
    - ahartmetz 1y ago
      It seems like current approaches have already reached a plateau. Either something new will work better soon... or there will be another long wait.
  - vjvjvjvjghv 1y ago
    I think AI may actually work really well for CAD. Purely text based prompts won't work. But a mix of touch and AR together with voice input could be extremely productive.
    - Our_Benefactors 1y ago
      Nonstarter, everybody hates VR
vjvjvjvjghv 1y ago
" but being a junior dev first is how you become a senior dev." If AI keeps improving, I am not so sure about that. Smart people may be able to quickly jump to senior skills. And what we view as senior skills at the moment may become useless.
- Wololooo 1y ago
  This feels to me like a very naive statement, you don't just learn something magically because it permeates, it's because of experience and number of failures or actually really playing with a system. If you are always on crutches and need the AI companion all the steps of the way, you basically offload the most important part of the work which is the whole cognitive work. That "senior skill" is the debugging of complex systems that often interconnects in unusual ways, why? Because this might be a one off this might not be well documented, or even the code might not really point at the issue because it might be a subtle interplay of hardware and software. Even if you are smart, it doesn't solve the problem that people still have to learn, and institutional knowledge is more important at times, because this might not be documented or easy to figure out, even with the greatest AI companion...

johnnyanmac 1y ago

>'ve been having similar thoughts about software engineering since one effect of LLMs is they make senior devs very productive, they make junior devs superfluous Well that's the pitch they want to throw at you. Actual studies disagree. >Maybe the same thing happens to manufacturing, and we just get rid of craftsmanship, machining, and logistics, and have everybody 3D-print their appliances in a factory-in-a-box they keep at their home, just shipping filament and chips and other raw materials directly to them. I legitmately think we will hit General AI before we have anything close to this. We simply don't have the resources needed on an individual bases to faciliate everyone being a "factory worker at home".

Our_Benefactors 1y ago
Look at how miserably AI fails at any kind of CAD work and you may reconsider.
- idiotsecant 1y ago
  There's no reason to think this will continue to be the case for long.
  - ethanwillis 1y ago
    none? not a single reason at all?
  - ahartmetz 1y ago
    It seems like current approaches have already reached a plateau. Either something new will work better soon... or there will be another long wait.
- vjvjvjvjghv 1y ago
  I think AI may actually work really well for CAD. Purely text based prompts won't work. But a mix of touch and AR together with voice input could be extremely productive.
  - Our_Benefactors 1y ago
    Nonstarter, everybody hates VR

vjvjvjvjghv 1y ago

" but being a junior dev first is how you become a senior dev." If AI keeps improving, I am not so sure about that. Smart people may be able to quickly jump to senior skills. And what we view as senior skills at the moment may become useless.

Wololooo 1y ago
This feels to me like a very naive statement, you don't just learn something magically because it permeates, it's because of experience and number of failures or actually really playing with a system. If you are always on crutches and need the AI companion all the steps of the way, you basically offload the most important part of the work which is the whole cognitive work. That "senior skill" is the debugging of complex systems that often interconnects in unusual ways, why? Because this might be a one off this might not be well documented, or even the code might not really point at the issue because it might be a subtle interplay of hardware and software. Even if you are smart, it doesn't solve the problem that people still have to learn, and institutional knowledge is more important at times, because this might not be documented or easy to figure out, even with the greatest AI companion...