"I work at NVIDIA during the day as a software engineer, and fiddle with AI for fun on the side"
Every single time. This article is concern trolling or advertisement trolling. I hope Nvidia's stock goes up due to the AI assisted existential crisis thoughts of this programmer.
elliotbnvl 11 hours ago [-]
I mean it is the butcher who knows how the sausage is made.
greazy 21 hours ago [-]
Great article.
While the definition changes, the expertise shifts and with it the field. Computers eventually became statisticians and data scientists. Printers became graphic designers.
What I found most interesting is that when positions undergo such evolution (printer -> graphic designer), a number of skills which were previously different expertise altogether, combine to create a new field. In other words, a new multidisciplinary field is born.
I think a good example is data science, the field at it's core is applied statistics using modern techniques such as data management and computing [0].
The question is, what is the new evolution of a programmer? Lots of folks like to use the term "engineer", and previously I thought this was silly. But now with LLMs, maybe that is a good descriptor; software engineer.
The moniker already exists which we need to revive and repurpose for the LLM era;
"Systems Engineer" i.e. one who does Systems Engineering - https://en.wikipedia.org/wiki/Systems_engineering Because the focus is no longer on coding alone but involves specification, verification (formal and testing), traceability and correctness. All using a whole plethora of third-party infrastructure, tools and components.
In the early days there used to be "Systems Analyst" and "Systems Designer" in addition to the above. All of them go together. The Systems Analyst is business requirements facing, The Systems Designer maps it to implementation architecture and The Systems Engineer pulls everything together (including costs/risks/specific implementation technologies etc.) to produce the complete functional system.
If you were a real programmer, you'd know that it is all hype.
esailija 18 hours ago [-]
> Lately I watch the machine do in a minute the thing I could have billed a morning for
Just raw code output is not really what anyone is paid for
There already existed billions of lines of code doing every possible thing imaginable that software is capable of doing and you could use all of it free of charge with the authors bending over backwards to make it as easy as possible for you to utilize.
Code that nobody understands and cannot be responsible for is just a flat out liability for the business, so these are not the same outputs.
And just going by raw numbers this would imply more than 100x productivity yet there is no tangible productivity change in big picture. 99 per 100 software devs are not fired, software is not getting better. In 600 days doing it the old way people could do any AAA game from scratch. If there is 100x improvement that should take 6 days now. Do you realize how insane even 10x sounds let alone 100x?
ido 17 hours ago [-]
> 99 per 100 software devs are not fired
While far from 99%, there’s been a lot of layoffs in the last couple years and less hiring such that anyone that has been looking for work can tell you’re there’s a massive in experience before and after ~2023.
elliotbnvl 11 hours ago [-]
I don’t foresee development timelines speeding up but I am seeing entire applications being built by one developer instead of four over the same timeframe. If anything communication overhead is reduced and velocity and quality are up, too.
turtleyacht 24 hours ago [-]
The thing about programming is it can be done with tape [1], birds [2] and textiles [3]. It happens to mostly be done on machines.
Eh, programmers used to be people who'd desk-check their flow charts before hand-translating them into machine code to enter into a front panel. There's been decades of growth in abstraction since then, and LLMs are just one more layer, another return of the perennial idea of "programming" by writing specifications in a natural language that a machine can automatically translate into actual code which it can run. You know, like what COBOL allows. We're still going to need people who are capable of making such specifications, ensuring the resulting code is correct, and fixing them when they're no longer sufficient.
framel 21 hours ago [-]
Humanity previously experimented, many centuries, with writing math in natural language and failed; it is fundamentally unsuited for the task. Furthermore, natural language specifications are, at best, wishful thinking. Feed this into a stochastic parrot, and you have a recipe for disaster. Repeating these mistakes proves coding is still a pseudoscience.
antonvs 18 hours ago [-]
> it is fundamentally unsuited for the task.
Very deepity. But you’ve apparently misunderstood mathematical notation - it’s just a shorthand, nothing stops one from expressing the same concepts in natural language.
> Furthermore, natural language specifications are, at best, wishful thinking.
More deepitism. There are plenty of counterexample to this, your claim only serves to suggest that you have no experience with software development.
> Feed this into a stochastic parrot
To might want to look for terminology in papers that were published after GPT 3.5 was released, it’ll make you sound less like an Amish person objecting to the “English”. Then again you used “clanker” in another comment, so I don’t hold out much hope.
framel 17 hours ago [-]
Very deepity?!? You lack basic grammar skills, yet here you are lecturing me about how mathematical notation is just syntactic sugar over natural language. That’s exactly the point: we already have a construct for telling computers what to do. We don’t need to use something subpar to express our intent.
> A superficial equivocation which only seems to be profound.
antonvs 4 hours ago [-]
> You lack basic grammar skills
You lack basic knowledge of the field you're ineptly trying to criticize. "Deepity" was coined by the Harvard philosophy professor Daniel Dennett, and was specifically intended to address the sort of empty nonsense you were spewing in your previous comment. Take this as a shot across the bow and go study before you continue vomiting your uninformed opinions everywhere.
msla 20 hours ago [-]
> coding is still a pseudoscience.
Amazing how you said that like it made any sense at all.
framel 20 hours ago [-]
amazing how you understood nothing from any of my words, how about using the clanker to vibe your understanding too?
msla 19 hours ago [-]
> amazing how you understood nothing from any of my words, how about using the clanker to vibe your understanding too?
"Coding is a pseudoscience" makes sense like "plumbing is a pseudoscience" makes sense: Fluid dynamics might be a pseudoscience, but plumbing either works or it doesn't, and it's either maintainable and modifiable or it isn't. Computer Science is what you're groping for, in your speech laden with things you will probably defend as not-slurs but which you're still bizarrely excited to get to use, but outright saying Computer Science is a pseudoscience might make you realize you're talking absolute nonsense.
framel 18 hours ago [-]
I get the impression you haven’t quite grasped what separates applied science from theoretical science.
"I work at NVIDIA during the day as a software engineer, and fiddle with AI for fun on the side"
Every single time. This article is concern trolling or advertisement trolling. I hope Nvidia's stock goes up due to the AI assisted existential crisis thoughts of this programmer.
While the definition changes, the expertise shifts and with it the field. Computers eventually became statisticians and data scientists. Printers became graphic designers.
What I found most interesting is that when positions undergo such evolution (printer -> graphic designer), a number of skills which were previously different expertise altogether, combine to create a new field. In other words, a new multidisciplinary field is born.
I think a good example is data science, the field at it's core is applied statistics using modern techniques such as data management and computing [0].
The question is, what is the new evolution of a programmer? Lots of folks like to use the term "engineer", and previously I thought this was silly. But now with LLMs, maybe that is a good descriptor; software engineer.
[0] https://www.welcometothejungle.com/en/articles/story-origin-...
The moniker already exists which we need to revive and repurpose for the LLM era;
"Systems Engineer" i.e. one who does Systems Engineering - https://en.wikipedia.org/wiki/Systems_engineering Because the focus is no longer on coding alone but involves specification, verification (formal and testing), traceability and correctness. All using a whole plethora of third-party infrastructure, tools and components.
In the early days there used to be "Systems Analyst" and "Systems Designer" in addition to the above. All of them go together. The Systems Analyst is business requirements facing, The Systems Designer maps it to implementation architecture and The Systems Engineer pulls everything together (including costs/risks/specific implementation technologies etc.) to produce the complete functional system.
See also my previous comments here - https://news.ycombinator.com/item?id=48264680
Just raw code output is not really what anyone is paid for
There already existed billions of lines of code doing every possible thing imaginable that software is capable of doing and you could use all of it free of charge with the authors bending over backwards to make it as easy as possible for you to utilize.
Code that nobody understands and cannot be responsible for is just a flat out liability for the business, so these are not the same outputs.
And just going by raw numbers this would imply more than 100x productivity yet there is no tangible productivity change in big picture. 99 per 100 software devs are not fired, software is not getting better. In 600 days doing it the old way people could do any AAA game from scratch. If there is 100x improvement that should take 6 days now. Do you realize how insane even 10x sounds let alone 100x?
While far from 99%, there’s been a lot of layoffs in the last couple years and less hiring such that anyone that has been looking for work can tell you’re there’s a massive in experience before and after ~2023.
[1] https://en.wikipedia.org/wiki/Turing_machine
[2] https://en.wikipedia.org/wiki/IP_over_Avian_Carriers
[3] https://www.karriweaver.com/selvagenotes/weaving-computing-t...
Very deepity. But you’ve apparently misunderstood mathematical notation - it’s just a shorthand, nothing stops one from expressing the same concepts in natural language.
> Furthermore, natural language specifications are, at best, wishful thinking.
More deepitism. There are plenty of counterexample to this, your claim only serves to suggest that you have no experience with software development.
> Feed this into a stochastic parrot
To might want to look for terminology in papers that were published after GPT 3.5 was released, it’ll make you sound less like an Amish person objecting to the “English”. Then again you used “clanker” in another comment, so I don’t hold out much hope.
https://en.wiktionary.org/wiki/deepity
> A superficial equivocation which only seems to be profound.
You lack basic knowledge of the field you're ineptly trying to criticize. "Deepity" was coined by the Harvard philosophy professor Daniel Dennett, and was specifically intended to address the sort of empty nonsense you were spewing in your previous comment. Take this as a shot across the bow and go study before you continue vomiting your uninformed opinions everywhere.
Amazing how you said that like it made any sense at all.
"Coding is a pseudoscience" makes sense like "plumbing is a pseudoscience" makes sense: Fluid dynamics might be a pseudoscience, but plumbing either works or it doesn't, and it's either maintainable and modifiable or it isn't. Computer Science is what you're groping for, in your speech laden with things you will probably defend as not-slurs but which you're still bizarrely excited to get to use, but outright saying Computer Science is a pseudoscience might make you realize you're talking absolute nonsense.