Hacker News new | ask | show | jobs
by not_the_fda 1178 days ago
If you have been getting by as a developer by searching stack overflow, then yes, gpt will replace you.

That's all these AI tools are, better stack overflow searches. They have no ability to know what is correct or what is wrong, it lacks judgement, which is one of the most important skills to have as a software engineer.

Engineering is about solving problems, these tools can't solve problems, they can regurgitate solutions to problems they have been trained on, often times confidently incorrectly, which is much worse than saying "I don't know".

They can't extract requirements from the client to find out what they really want.

They completely fail at moderately hard problems, or novel problems.

I think these tools may be worse for the industry because people will have less opportunity to learn problem solving skills since the AI will handle to easy stuff, and when the hard stuff come along, people won't have the skills to solve them.

For those with the good problem solving skills, AI isn't a threat. There will always be work for solving hard problems, making judgements, and trade offs, actual thinking.

1 comments

> That's all these AI tools are, better stack overflow searches. They have no ability to know what is correct or what is wrong, it lacks judgement, which is one of the most important skills to have as a software engineer.

I keep seeing this idea, but have you actually used it? It clearly has the ability to problem solve. It's not just copying and pasting solutions.

Ok granted it's not especially good at it yet and the bullshitting problem is a real issue, but how long do you think that will remain unsolved?

I think where it will continue to struggle is niche domains that aren't on the internet a lot, e.g. hardware design. But if you're writing CRUD apps all day you should be worried!

Or at least brace for your job description to change from "Software Developer" to "Prompt Developer and AI Output Verification".

> Ok granted it's not especially good at it yet and the bullshitting problem is a real issue, but how long do you think that will remain unsolved?

I'm curious why some people seem to think it's going to be solved imminently. The last 1% is always the hardest (by far)!

Because recent progress has been extremely rapid and has crossed the threshold from "this is rubbish" to "woah this actually sort of works!" which is a really big deal.

It's already putting people out of jobs.

> Because recent progress has been extremely rapid and has crossed the threshold from "this is rubbish" to "woah this actually sort of works!" which is a really big deal.

Sure, progress has been astonishing, but why does that imply that any given sticking point will be resolved ever, let alone soon?

> It's already putting people out of jobs.

Is it? Who? do you have any references?