let's maybe not throw LLMs on simple calculations?
kagi didn't do that by default btw: it usually pops up a calculator, it just didn't that time for some reason and I clicked on "Quick Answer".
@fasterthanlime the footnote in the top screenshot makes it extra funny imo
@barometz who can argue with such a citation tbqh
@fasterthanlime ....this hurts even more knowing kagi is a paid product
@fasterthanlime@hachyderm.io ah but you see, because it uses billions of multiplications the left is more accurate
@fasterthanlime
Entirely unsolicited but perhaps useful suggestion: at some point I gave up on search calculators working and found out about https://rinkcalc.app/ (written in Rust!) which is really nice and also unit-aware
@fasterthanlime lol let's do basic math by multiplying billions of numbers together
@fasterthanlime Are you sure Wolfram is right, though? I’m no mathematician, but Kagi was so eloquent that it persuaded me. I mean, look at that step-by-step, clearly it must know what it’s doing!
@fasterthanlime Almost like an overgrown Markov chain can't actually process anything it's given as anything more than a sequence of tokens with probabilities on what tokens come next...
@fasterthanlime Future generations are simply not going to understand what we mean about how computers used to give precise answers, do the same thing every time on the same inputs, etc
@fasterthanlime floating point precision loss but for AI