Hacker News new | past | comments | ask | show | jobs | submit
Most SWE who "oppose AI" in my experience, don't oppose AI but oppose its how its being adapted or used. All gas no brakes. You can't really say "lets think about it a bit" without becoming AI denier/opposer or whatever.

We make aggressive decisions, based on overly optimistic bet on "AI will fix it" mindset. I think it should be more towards, given what we have now and may put in little bit of optimism around AI how can go forward without completely dismissing decades of engineering principles we developed. If AI really goes 10x better than what it is now, then we can just build stuff much faster then anyway, but it didn't we made something reliable today as well something we can keep building on 5/10 years from now.

> You can't really say "lets think about it a bit" without becoming AI denier/opposer or whatever.

What does "lets think about it a bit" even mean though? You want other people to stop using productive tools? Or you want to be able to opt out of using productive tools without an employer worrying about it? There always seems to be either a shifting of burden to others, or a denial that AI tools increase productivity.

People can think about AI and use AI with all gas and no brakes. Or you can opt out of using AI for as long as you like to think about things.

>> You want other people to stop using productive tools?

This is the EXACTLY the kind of BS you get. No, not asking you to stop using those tools. Its not 0 or 1 thing. Its not a choice between "all gas no brake" and "I wont use AI". But maybe, just maybe, think a little about what you are producing, instead of just being a ENTER PRESSING entity. The amount of spaghetti that is being generated is not just an issue from the perspective "oh the code looks horrible to read", but also fundamentally how we build complicated systems, and it becomes bandages over bandages that just sucks and potentially fragile.

The alternative is don't go all gas, work strategically thinking a bit longer term like we always had in past, and maybe you loose little velocity in short term, but you are likely better off in long term.

AI is always ready to implement something, or generate code, and make decisions, and it will do whatever you ask it to do. However, to make right decisions you need to understand the problem/context etc etc. And that was always significant portion of work, and you can be little faster with that with AI, but I think that part is still kind of slow but we as industry are in denial about that, and just not doing that part that well anymore.

loading story #50569140
You can go out and take a walk around the block/quad/building still. And come back knowing the exact prompt to use to find the true explanation for some puzzling data anomaly/code behavior/failure mode.
> What does "lets think about it a bit" even mean though?

Don't push code no human has read into production.

loading story #50569155
Surely those aren't the only two options though? What if "let's think about it a bit" is about your coworkers going head first into a design that you know will fail, and then maybe let's think about how we can do a better design before we lock in a shit one. The AI does build _exactly_ what you ask, whether it's a smart choice or not. Surely you can just automate that choice to the AI as well, why not? "Think about all the ways to design this and pick the best". But possibly we are being led blind into a tar pit. So maybe, just maybe, we can spend a seconds of our lives to try to involve our human domain knowledge and hard won expertise at this job and its systems to guide the AI in a more productive route. Or is this scenario really just about slowing down everyone around you because you're secretely an indignant, neolithic, asshole.
loading story #50569246
I think ops point is the tools are irrelevant. What comes out is the only thing that matters. Making grandiose assumptions about the arc of history is the job of pundits and linkedin influencers.