joao.omg.lol
Research Engineer at Google DeepMind
A special snowflake existing in 196883 dimensions
#ActuallyAutistic He/Him
343 posts
5,920 followers
2,179 following
Prolific Poster
Conversation Starter
comment in response to
post
in general i think we expect more from languages now? you need a compiler *and* all the adjacent tooling, and even then the tooling compat story is terrible so you need to pay attention to how your tools work with all the major editors
comment in response to
post
Definitivamente, se temos conferência/convenções, swag, e gente com tatuagens de PLs com certeza podemos classificar parte da comunidade como fandoms
comment in response to
post
So much yes!
Somewhat related, have you seen “The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers”
comment in response to
post
Frankly, even after training models that can do this for general clothing is solved I think the hardware cost will still be an issue :/
comment in response to
post
Love to see more people being comfortable enough to come out and be themselves!
This reminds me of how, in the early 1900s, few reported being left-handed due to societal pressure forcing right handedness. As that pressure lessened, more people were comfortable reporting themselves as a lefty.
comment in response to
post
fortran favors the old
comment in response to
post
And the worst is I've seen people use one of these to mean the others, especially pass@n and BoN, which are quite different, so yeah, very hard to figure out what any of the numbers mean
comment in response to
post
RM@n (use the reward model to select one of the n answers, can be useful if the reward model is open of if there's a way to use this specific kind of sampling), BoN (best-of-n, assumes you have some kind of oracle that can check if any of the n answers is correct, e.g. online judge or Thm Prover)
comment in response to
post
Frankly, I'll just agree it's a mess. There's cons@k, which can be used for things with a very clear answer (e.g. a multi-choice problem, a yes/no, number), there's pass@n (you run n times and approximate the probability of being right for each test case) 1/2
comment in response to
post
Either way
comment in response to
post
They may find that multiple health issues are related to the stress induced by the patriarchy and its expectations put on men and women?
comment in response to
post
Can you elaborate a bit?