LLMs Making Algorithms & Data Structures Obsolete

63 points · 15 comments · view on lemmy.world

15 Comments

spread@programming.dev · 32 pts · 3y (5 replies)

Impressive, very nice. Now let's see LLM's space complexity.

fubo@lemmy.world · 27 pts · 3y (3 replies)

O(all the GPUs, all of them)

VictorPrincipum@vlemmy.net · 11 pts · 3y (2 replies)

Hey now, don’t forget all the memory too

rcmaehl@lemmy.world · 11 pts · 3y (1 reply)

And my cache!

Sotuanduso@lemm.ee · 2 pts · 3y

/UnexpectedGimli

whiskeypickle@lemmy.ml · 5 pts · 3y

Eggshell… and is that… Gothic type?

duncesplayed@lemmy.one · 17 pts · 3y (4 replies)

Any algorithm can be O(n^2) if you only want it to be occasionally right.

Iridium@lemmy.world · 18 pts · 3y (1 reply)
Function isPrime(number):
    return false

Accurate for almost 100% of cases

darcy@sh.itjust.works · 2 pts · 3y

as test count approach infinity

julianh@lemm.ee · 5 pts · 3y

Any algorithm can be O(1) if you cache all the answers beforehand.

MajorHavoc@lemmy.world · 2 pts · 3y

Yes.

And depending how occasionally we're talking, I can code for some very fast solutions when the correctness requirements are low enough.

Alternately, if we want it to only be occasionally fast, I've got a very nice looking and very wrong algorithm for that, as well.

darcy@sh.itjust.works · 2 pts · 3y (3 replies)

isnt O(n³) usually simplified to O(n²) anyway ?

AlmightySnoo@lemmy.world · 3 pts · 3y (1 reply)

No, n³ cannot be O(n²) as otherwise that would mean that there exists a positive constant K and a positive threshold m such that for any integer n greater than m you would have n³ less than K*n², which would be the same as saying n less than K, which cannot hold for any integer n greater than m. So n³ cannot be an O(n²), which means that something that is an O(n³) is not necessarily an O(n²).

It's the other way around, if something is an O(n²) then it is necessarily also an O(n³).

darcy@sh.itjust.works · 2 pts · 3y

ok thanks

MajorHavoc@lemmy.world · 2 pts · 3y

Yes. The other answer is technically correct, but yours is pragmatically correct.

If a solution is worse than O(nln(n))* then most of us are going to be looking for a pragmatic and completely alternate way to deal with it, rather than analyzing how to make it mildly less terrible.

So I'm just writing O(n^2) as a quick professional replacement for my original write in answer of "dogshit".