I’m an accountant. Components of the job have been being automated or systemised for many decades. Most of the tasks that occupied a graduate when I was one 20 years ago don’t exist anymore.
Not because AI is doing those tasks but just because everything became more integrated, we configure and manage the flow of data rather than making the data, you might say.
If you had to hire 100 professional programmers in the past, but then AI makes programmers 10% more efficient than previously, then you can do the same work with 91 programmers.
That doesn’t mean that 9 people were doing something that an LLM can do, it just means that more work is being completed with fewer programmers.
If you had to hire 100 professional programmers in the past, but then AI makes programmers 10% more efficient than previously, then you can do the same work with 91 programmers.
You’ve nailed to root of the misunderstanding by non-programmers. We’re already optimized past that target.
Some people think we type all day. We don’t. We stare at our screen saying “what the fuck?!” for most of the day. Those is especially true for the best programmers doing really interesting work.
There’s maybe three living humans who actually know how to correctly build a Windows installer. One of those three is paid to sell software to automate the task for everyone else. The other two retired already. (One is hiding out as a bar tender and claims to not speak any English if recognized from their MSI days.)
Pick an interesting topic in programming, and you’ll find similarly ludicrous optimization.
There’s a few hundred programmers building all banking automation, selling it to millions of bank employees.
It’s possible that AI will force a dozen people to stop doing banking automation. It’s a lot more likely that the backlog of unmet banking automation need will instead just get very slightly smaller.
Now, the reality of the economics won’t stop CIOs from laying off staff and betting that AI will magically expand to fill the gap. We’re seeing that now. That’s called the “fuck around” phase.
But we’ve seen “this revolutionary technology will make us not need more programmers” before (several times). The outcomes, when the dust settles are:
The job is now genuinely easier to do, at least for beginners. (Senior professionals had access to equivalent solutions, before everyone else got excited.)
More people are now programmers. (We laid a bunch of them off, and we meant to not hire any back, but it turned out that our backlog of cool/revolutionary/necessary ideas was more important to leadership than pinching pennies.)
A lot of work that was previously ignored completely now gets done, but done very badly by brand new programmers. (We asked the senior developers to do it, but they said “Fuck you, that’s not important, make the new kid do it.” I think they’re just still cranky that we spent three years laying off staff instead of training…)
The average quality of all software is now a bit worse, but there’s a lot more variety of (worse) software now available.
To add on this, this doesn’t necessarily mean that there are fewer programing jobs in total. If people work 10% more efficently, that means that the cost of labor is only 91% of what it was before meaning that people might be able to afford to finance more programing projects. One thing that does matter is for example things like entry level jobs disappearing or the nature of the work changing. Doing less boring gruntwork can make the job more fun, but otoh digitization sometimes results in the worker having less agency in what they do since they have to fit everything into a possibly inflexible digital system.
Automation is always incremental.
I’m an accountant. Components of the job have been being automated or systemised for many decades. Most of the tasks that occupied a graduate when I was one 20 years ago don’t exist anymore.
Not because AI is doing those tasks but just because everything became more integrated, we configure and manage the flow of data rather than making the data, you might say.
If you had to hire 100 professional programmers in the past, but then AI makes programmers 10% more efficient than previously, then you can do the same work with 91 programmers.
That doesn’t mean that 9 people were doing something that an LLM can do, it just means that more work is being completed with fewer programmers.
You’ve nailed to root of the misunderstanding by non-programmers. We’re already optimized past that target.
Some people think we type all day. We don’t. We stare at our screen saying “what the fuck?!” for most of the day. Those is especially true for the best programmers doing really interesting work.
There’s maybe three living humans who actually know how to correctly build a Windows installer. One of those three is paid to sell software to automate the task for everyone else. The other two retired already. (One is hiding out as a bar tender and claims to not speak any English if recognized from their MSI days.)
Pick an interesting topic in programming, and you’ll find similarly ludicrous optimization.
There’s a few hundred programmers building all banking automation, selling it to millions of bank employees.
It’s possible that AI will force a dozen people to stop doing banking automation. It’s a lot more likely that the backlog of unmet banking automation need will instead just get very slightly smaller.
Now, the reality of the economics won’t stop CIOs from laying off staff and betting that AI will magically expand to fill the gap. We’re seeing that now. That’s called the “fuck around” phase.
But we’ve seen “this revolutionary technology will make us not need more programmers” before (several times). The outcomes, when the dust settles are:
To add on this, this doesn’t necessarily mean that there are fewer programing jobs in total. If people work 10% more efficently, that means that the cost of labor is only 91% of what it was before meaning that people might be able to afford to finance more programing projects. One thing that does matter is for example things like entry level jobs disappearing or the nature of the work changing. Doing less boring gruntwork can make the job more fun, but otoh digitization sometimes results in the worker having less agency in what they do since they have to fit everything into a possibly inflexible digital system.