6 ms·

I don't worry about it, because: 1) I believe we need true AGI to replace developers. 2) I don't believe LLMs are currently AGI or that if we just feed them m

I don't worry about it, because: 1) I believe we need true AGI to replace developers. 2) I don't believe LLMs are currently AGI or that if we just feed them more compute during training that they'll magically become AGI. 3) Even if we did invent AGI soon and replace developers, I wouldn't even really care, because the invention of AGI would be such an insanely impactful, world changing, event that who knows what the world would even look like afterwards. It would be massively changed. Having a development job is the absolute least of my worries in that scenario, it pales in comparison to the transformation the entire world would go through.

plantwallshoe 2y ago
This is where I’m at too. By the time we fully automate software engineering, we definitely will have already automated marketing, sales, customer success, accountants, office administrators, program managers, lawyers, etc. At that point its total societal upheaval and losing my job will probably be the least of my worries.
akira2501 2y ago
Even if AGI suddenly appears we will most likely have an energy feed and efficiency problem with it. These scaling problems are just not on the common roadmap at all and people forget how much effort typically has to be spent here before a new technology can take over.
janalsncm 2y ago
To replace all developers, we need AGI yes. To replace many developers? No. If one developer can do the same work as 5 could previously, unless the amount of work expands then 4 developers are going to be looking for a job. Therefore, unless you for some reason believe you will be in the shrinking portion that cannot be replaced I think the question deserves more attention than “nothing”.
- amrocha 2y ago
  I have a coworker who I suspect has been using LLMs to write most of his code for him. He wrote multiple PRs with thousands of lines of code over a month. Me and the other senior dev spent weeks reviewing these PRs. Here’s what we found: - The feature wasn’t built to spec, so even though it worked in general the details were all wrong - The code was sloppy and didn’t adhere to the repos guidelines - He couldn’t explain why he did things a certain way versus another, so reviews took a long time - The code worked for the happy path, and errored for everything else Eventually this guy got moved to a different team and we closed his PRs and rewrote the feature in less than a week. This was an awful experience. If you told me that this is the future of software I’d laugh you out of the room, because engineers make enough money and have enough leverage to just quit. If you force engineers to work this way, all the good ones will quit and retire. So you’re gonna be stuck with the guys who can’t write code reviewing code they don’t understand.
  - janalsncm 2y ago
    In the short term, I share your opinion. LLMs have automated the creation of slop that resembles code. In the long term, we SWEs (like other industries) have to own the fact that there’s a huge target on our backs, and aside from hubris there’s no law of nature or man preventing people smarter than us from building robots that do our jobs faster than us.
    - amrocha 2y ago
      That’s called class consciousness, and I agree, and I think most people have already realized companies are not their friend after the last two years of layoffs. But like I said, I’m not worried about it in the imminent future, and I have enough leverage to turn down any jobs that want me to work in that way.
- fire_lake 2y ago
  > unless the amount of work expands This is what will happen
- natemwilson 2y ago
  It’s my belief that humanity has an effectively infinite capacity for software and code. We can always recursively explore deeper complexity.
  - janalsncm 2y ago
    As we are able to automate more and more of the creation process, we may be able to create an infinite amount of software. However, what sustains high wages for our industry is a constrained supply of people with an ability to create good software.
- throwawayffffas 2y ago
  That's a weird way to look at it. If one developer can do what 5 could do before, that doesn't mean I will be looking for a job, it means I will be doing 5 times more work.
  - Valord 2y ago
    5x throughput/output not 5x work
- j45 2y ago
  I think counting the number of devs might not be the best way to go considering not all teams are equally capable or skilled in each person, and in enterprises, some people are inevitably hiding in a project or team. Comparing only the amount of forward progress in a codebase and AI's ability to participate or cover in it might be better.
- tqi 2y ago
  I'm not sure it is as simple as that - Induced Demand might be enough to keep the pool of human-hours steady. What that does to wages though, who can say...
  - janalsncm 2y ago
    Well it certainly depends on how much induced demand there is, and how much automation can multiply developer productivity. If we are talking an 80% reduction in developers needed per project, then we would need 5x the amount of software demand in the future to avoid a workforce reduction.
- JamesBarney 2y ago
  I think not only is this possible, it's likely for two reasons. 1. A lot more projects get the green light when the price is 5x less, and a many more organizations can afford custom applications. 2. LLMs unlock large amounts of new applications. A lot more of the economy is now automatable with LLMs. I think jr devs will see the biggest hit. If you're going to teach someone how to code, might as teach a domain expert. LLMs already code much better than almost all jr devs.
- simplyluke 2y ago
  Frameworks, compilers, and countless other developments in computing massively expanded the efficiency of programmers and that only expanded the field. Short of genuine AGI I’ve yet to see a compelling argument why productivity eliminates jobs, when the opposite has been true in every modern economy.
  - janalsncm 2y ago
    > Frameworks, compilers, and countless other developments in computing How would those have plausibly eliminated jobs? Neither frameworks nor compilers were the totality of the tasks a single person previously was assigned. If there was a person whose job it was to convert C code to assembly by hand, yes, a compiler would have eliminated most of those jobs. If you need an example of automation eliminating jobs, look at automated switchboard operators. The job of human switchboard operator (mostly women btw) was eliminated in a matter of years. Except here, instead of a low-paid industry we are talking about a relatively high-paid one, so the returns would be much higher. A good analogy can be made to outsourcing for manufacturing. For a long time Chinese products were universally of worse quality. Then they caught up. Now, in many advanced manufacturing sectors the Chinese are unmatched. It was only hubris that drove arguments that Chinese manufacturing could never match America’s.
throw234234234 2y ago
1) Maybe. 2) I do see this, given the money poured into this cycle, as a potential possibility. It may not just be LLM's. To another comment you are betting against the whole capitalist systems, human ingenuity and billions/trillions? of dollars targeted at making SWE's redundant. 3) I think it can disrupt only knowledge jobs, and be some large time before it disrupts physical jobs. For SWE's this is the worst outcome - it means you are on your own w.r.t adjusting for the changes coming. Its only "world changing" as you put it to economic systems if it disrupts everyone at once. I don't think it will happen that way. More to the point software engineers will automate themselves out before other jobs for only one reason - they understand AI better than other jobs (even if objectively it is harder to automate) and they tend not to protect the knowledge required to do so. They have the domain knowledge to know what to automate/make redundant. The people that have the power to resist/slow down disruption (i.e. hide knowledge) will gain more pricing power, and therefore be able to earn more capital taking advantage of the efficiency gains made by jobs being redundant from AI. The last to be disrupted has the most opportunity to gain ownership of assets and capital from their economic profits preserved. The inefficient will win out of this - capital rewards scarcity/people that can remain in demand despite being inefficient relatively. Competition is for losers - its IMV the biggest flaw of the system. As a result people will see what has happened to SWE's and make sure their industry "has time" to adapt particularly since many knowledge professions are really "industry unions/licensed clubs" who have the advantage of keeping their domain knowledge harder to access. To explain it further even if software is more complicated; there is just so much more capital it seems trying to disrupt it than other industries. Given IMV software demand is relatively inelastic to price due to scaling profits, making it cheaper to produce won't really benefit society all that much w.r.t more output (i.e. what was good economically to build would of been built anyway in an inelastic demand/scaling commodity). Generally more supply/less cost of a good has more absolute societal benefits when there is unmet and/or elastic demand. Instead costs of SWE's will go down and the benefit will be distributed to the jobs/people remaining (managers, CEO's, etc) - the people that dev's think "are inefficient" in my experience. When it is about inelastic demand its more re-distributive; the customer benefits and the supplier (in this case SWE's) lose. I don't like saying this; but we gave AI all the advantage. No licensing requirements, open source software for training, etc.
- 2y ago

plantwallshoe 2y ago

This is where I’m at too. By the time we fully automate software engineering, we definitely will have already automated marketing, sales, customer success, accountants, office administrators, program managers, lawyers, etc. At that point its total societal upheaval and losing my job will probably be the least of my worries.

akira2501 2y ago

Even if AGI suddenly appears we will most likely have an energy feed and efficiency problem with it. These scaling problems are just not on the common roadmap at all and people forget how much effort typically has to be spent here before a new technology can take over.

janalsncm 2y ago

To replace all developers, we need AGI yes. To replace many developers? No. If one developer can do the same work as 5 could previously, unless the amount of work expands then 4 developers are going to be looking for a job. Therefore, unless you for some reason believe you will be in the shrinking portion that cannot be replaced I think the question deserves more attention than “nothing”.

amrocha 2y ago
I have a coworker who I suspect has been using LLMs to write most of his code for him. He wrote multiple PRs with thousands of lines of code over a month. Me and the other senior dev spent weeks reviewing these PRs. Here’s what we found: - The feature wasn’t built to spec, so even though it worked in general the details were all wrong - The code was sloppy and didn’t adhere to the repos guidelines - He couldn’t explain why he did things a certain way versus another, so reviews took a long time - The code worked for the happy path, and errored for everything else Eventually this guy got moved to a different team and we closed his PRs and rewrote the feature in less than a week. This was an awful experience. If you told me that this is the future of software I’d laugh you out of the room, because engineers make enough money and have enough leverage to just quit. If you force engineers to work this way, all the good ones will quit and retire. So you’re gonna be stuck with the guys who can’t write code reviewing code they don’t understand.
- janalsncm 2y ago
  In the short term, I share your opinion. LLMs have automated the creation of slop that resembles code. In the long term, we SWEs (like other industries) have to own the fact that there’s a huge target on our backs, and aside from hubris there’s no law of nature or man preventing people smarter than us from building robots that do our jobs faster than us.
  - amrocha 2y ago
    That’s called class consciousness, and I agree, and I think most people have already realized companies are not their friend after the last two years of layoffs. But like I said, I’m not worried about it in the imminent future, and I have enough leverage to turn down any jobs that want me to work in that way.
fire_lake 2y ago
> unless the amount of work expands This is what will happen
natemwilson 2y ago
It’s my belief that humanity has an effectively infinite capacity for software and code. We can always recursively explore deeper complexity.
- janalsncm 2y ago
  As we are able to automate more and more of the creation process, we may be able to create an infinite amount of software. However, what sustains high wages for our industry is a constrained supply of people with an ability to create good software.
throwawayffffas 2y ago
That's a weird way to look at it. If one developer can do what 5 could do before, that doesn't mean I will be looking for a job, it means I will be doing 5 times more work.
- Valord 2y ago
  5x throughput/output not 5x work
j45 2y ago
I think counting the number of devs might not be the best way to go considering not all teams are equally capable or skilled in each person, and in enterprises, some people are inevitably hiding in a project or team. Comparing only the amount of forward progress in a codebase and AI's ability to participate or cover in it might be better.
tqi 2y ago
I'm not sure it is as simple as that - Induced Demand might be enough to keep the pool of human-hours steady. What that does to wages though, who can say...
- janalsncm 2y ago
  Well it certainly depends on how much induced demand there is, and how much automation can multiply developer productivity. If we are talking an 80% reduction in developers needed per project, then we would need 5x the amount of software demand in the future to avoid a workforce reduction.
JamesBarney 2y ago
I think not only is this possible, it's likely for two reasons. 1. A lot more projects get the green light when the price is 5x less, and a many more organizations can afford custom applications. 2. LLMs unlock large amounts of new applications. A lot more of the economy is now automatable with LLMs. I think jr devs will see the biggest hit. If you're going to teach someone how to code, might as teach a domain expert. LLMs already code much better than almost all jr devs.
simplyluke 2y ago
Frameworks, compilers, and countless other developments in computing massively expanded the efficiency of programmers and that only expanded the field. Short of genuine AGI I’ve yet to see a compelling argument why productivity eliminates jobs, when the opposite has been true in every modern economy.
- janalsncm 2y ago
  > Frameworks, compilers, and countless other developments in computing How would those have plausibly eliminated jobs? Neither frameworks nor compilers were the totality of the tasks a single person previously was assigned. If there was a person whose job it was to convert C code to assembly by hand, yes, a compiler would have eliminated most of those jobs. If you need an example of automation eliminating jobs, look at automated switchboard operators. The job of human switchboard operator (mostly women btw) was eliminated in a matter of years. Except here, instead of a low-paid industry we are talking about a relatively high-paid one, so the returns would be much higher. A good analogy can be made to outsourcing for manufacturing. For a long time Chinese products were universally of worse quality. Then they caught up. Now, in many advanced manufacturing sectors the Chinese are unmatched. It was only hubris that drove arguments that Chinese manufacturing could never match America’s.

throw234234234 2y ago

1) Maybe. 2) I do see this, given the money poured into this cycle, as a potential possibility. It may not just be LLM's. To another comment you are betting against the whole capitalist systems, human ingenuity and billions/trillions? of dollars targeted at making SWE's redundant. 3) I think it can disrupt only knowledge jobs, and be some large time before it disrupts physical jobs. For SWE's this is the worst outcome - it means you are on your own w.r.t adjusting for the changes coming. Its only "world changing" as you put it to economic systems if it disrupts everyone at once. I don't think it will happen that way. More to the point software engineers will automate themselves out before other jobs for only one reason - they understand AI better than other jobs (even if objectively it is harder to automate) and they tend not to protect the knowledge required to do so. They have the domain knowledge to know what to automate/make redundant. The people that have the power to resist/slow down disruption (i.e. hide knowledge) will gain more pricing power, and therefore be able to earn more capital taking advantage of the efficiency gains made by jobs being redundant from AI. The last to be disrupted has the most opportunity to gain ownership of assets and capital from their economic profits preserved. The inefficient will win out of this - capital rewards scarcity/people that can remain in demand despite being inefficient relatively. Competition is for losers - its IMV the biggest flaw of the system. As a result people will see what has happened to SWE's and make sure their industry "has time" to adapt particularly since many knowledge professions are really "industry unions/licensed clubs" who have the advantage of keeping their domain knowledge harder to access. To explain it further even if software is more complicated; there is just so much more capital it seems trying to disrupt it than other industries. Given IMV software demand is relatively inelastic to price due to scaling profits, making it cheaper to produce won't really benefit society all that much w.r.t more output (i.e. what was good economically to build would of been built anyway in an inelastic demand/scaling commodity). Generally more supply/less cost of a good has more absolute societal benefits when there is unmet and/or elastic demand. Instead costs of SWE's will go down and the benefit will be distributed to the jobs/people remaining (managers, CEO's, etc) - the people that dev's think "are inefficient" in my experience. When it is about inelastic demand its more re-distributive; the customer benefits and the supplier (in this case SWE's) lose. I don't like saying this; but we gave AI all the advantage. No licensing requirements, open source software for training, etc.

2y ago