Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

You should probably not be thinking about it in that way.

Anthropomorphising statistical models is a bad idea...



Oh I‘m not antropomorphising. It feels more like a GPS - a tool I wouldn’t like to be without when driving unknown places.

I work mostly alone on projects, at least currently. While I have friends I can bounce back general ideas on, it‘s hard to get good feedback on a small problem that I‘d be stuck with for, let’s say, 5 minutes.

So far, the choice has been between disrupting others‘ flow (who might not even work with the tech I‘m using), or exploring the issue myself. Problem solving is certainly not a skill I want to atrophy, but figuring out framework specific intricacies is more a chore than an interesting problem. A chore I can get stuck on. And that’s where the bot often points me in the right direction.

An example: I‘m dabbling in the PETAL stack with a side project, and Ecto‘s DSL still feels foreign to me. ChatGPT is actually really good at fixing my Ecto queries and recently made the suggestion to use the „dynamic“ function. It applied the function in a slightly wrong way, but that made me read up dynamic query building, which is already bearing fruit.

I could have read all of Ecto‘s documentation and wouldn’t need the bot, but that’s out of scope right now. I‘m currently working with native iOS, Android, JS/TS (Next), Flutter, and Elixir - if I read all the documentation, I wouldn’t write any code.


You meant "something to fall back on" instead of "someone". If you write it like that suddenly you're not anthropomorphizing anymore =)


I agree, I noticed this too late. Apologies to GP


We are anthropomorphizing NPCs in video games for decades - it's ok.


Yeah, they hate it.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: