alt Hacker NewsWe have a skewed-to-the-positive view about AI around here because we are the industry where it has turned out to work to a significant degree. We can argue about whether it lived up to the promises and the practical implications of using it and how much you can trust it, etc., but it's actually a debate in our space because there is enough value that it is worth debating. I think in a lot of other spaces it isn't working nearly as well. From my blog https://jerf.org/iri/post/2026/programming_is_engineering/ :
"It’s worth taking a moment to really watch how [AIs] work, as they are working on some relatively lengthy task for you. It reminds me of bumper bowling [1]; link to a short video showing what it is. The AIs bang along all our bumpers; failing compiles, failing tests, failing integrations, error messages, all the bumpers we’ve built into our engineering process, and in the end they get a good result. But that good result is as much a result of all the solid engineering processes we have installed as anything else, because the AI without those protections rolls into the gutter relatively quickly. In those fields where the bumpers can’t be built as easily or in as great a quantity as we can have them, that’s what happens with the AIs.
"... it’s not that “programmers are awesome”, it’s that our domain is amenable to having all these bumpers in the first place."
I can think of bumpers for other domains, but not in the quantity we have in our space. Plus there's things like accounting, where "accounting" already is the bumper, having an AI banging into the accounting bumpers is much more concerning than a syntax error on compile is for us. Coding AIs would be nearly useless if the bumpers weren't there, and I think that's where a lot of other domains end up with when it comes to AI.
If you want to see what I mean, watch a longer coding process in your AI, notice the first place it bangs into one of the bumpers I'm talking about, and then imagine how valuable the AI would be if instead it didn't realize that it banged into a bumper, continued on obliviously, and then went off into ever-more-fantastic flights of fancy with no connection to reality. Such an AI would be too hazardous to use. I think that's the experience of almost every other field right now.
> We have a skewed-to-the-positive view about AI around here because we are the industry where it has turned out to work to a significant degree.
Well... except that it actually hasn't. A handful of people are saying "yeah we're getting huge gains, it's unreal". But most devs here talk about how it's at best a modest speedup, and at worst slows them down. So even for programming, the thing they are supposedly good at, LLMs aren't actually that good.