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

> I am not sure I will ever trust an ML algorithm trained on inputs/outputs only

It’s doing TDD!



Imagine a world where all programmers do is write the tests and the generalized learning algos write all the implementations.


No need to imagine. That would just be Declarative Programming going mainstream?

Its just that algorithms like the Prolog’s WAM or the reet used in rules are replaced with neural nets and the like.

Rules use the same When Then format acting on a Given fact base. Prologs Horn Clauses can easily be translated into that form.

BDD writes all tests in the format Given When Then.

It’s already here.


Welcome to my company a few years ago. Except they were not programmers but PM and architects, and algos were subcontractors. Now we have real developers (I'm in one of those teams)


That's such a big pile of yikes, I'm morbidly drawn to see that codebase.


I'm pretty sure the generated code will be an incomprehensible mess that passes precisely those tests and nothing more, e.g. it would correctly sort arrays [1,3,2] and [3,2,1], but not [2,1,3]; and obviously not [1,4,2], because the specs didn't mention number 4.




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

Search: