> Engineers report shifting toward higher-level work managing AI systems and report significant productivity gains. However, these changes also raise questions about the long-term trajectory of software engineering as a profession.
I’m surprised employees are this open to their employer about these details of their job.
I am curious : could GenAI have written the paper "Attention is all you need"? We were trapped in CNN RNN architectures for a while : could genAi have arrived at a better architecture ?
No, it couldn't have unless these ideas were sandwiched between other ideas that it could interpolate between.
You have to approach genai as a high-dimensional interpolation machine. It can perform extrapolation when you, the user, provide enough information to operate on. It can interpolate between what you provide and what it knows as well.
With these constraints, it is still pretty powerful, and I am generalizing of course. But in my experience, it is terrible at truly novel implementations of anything. It makes countless mistakes, because it continually attempts to fit to patterns found in existing code.
So you can really see the weaknesses at the frontier. I would encourage experimenting there to confirm what I am saying.
Can they please prompt their way to not having Claude go down every 4 days? Thank you.
I think you now know why it does
Why would you report honestly to your employer about metrics and methodologies that could potentially show you're not needed?
They are still needed though - the code isn't just prompted into thin air by nobody.
How many are needed though.
And for how long.
> Engineers report shifting toward higher-level work managing AI systems and report significant productivity gains. However, these changes also raise questions about the long-term trajectory of software engineering as a profession.
I’m surprised employees are this open to their employer about these details of their job.
"...approaching weekly limit..."
I am curious : could GenAI have written the paper "Attention is all you need"? We were trapped in CNN RNN architectures for a while : could genAi have arrived at a better architecture ?
I'm yet to see a convincing example of LLMs producing anything substantially insightful.
Depends on how you define "insight" really.
Is doing meta-analysis and discovering a commonality "insightful" for example?
Or is insight only something new you discover without basing your discovery on anything?
No, it couldn't have unless these ideas were sandwiched between other ideas that it could interpolate between.
You have to approach genai as a high-dimensional interpolation machine. It can perform extrapolation when you, the user, provide enough information to operate on. It can interpolate between what you provide and what it knows as well.
With these constraints, it is still pretty powerful, and I am generalizing of course. But in my experience, it is terrible at truly novel implementations of anything. It makes countless mistakes, because it continually attempts to fit to patterns found in existing code.
So you can really see the weaknesses at the frontier. I would encourage experimenting there to confirm what I am saying.
Most believable stats out there.
/s