rllin

"One weird trick" is still pretty much the first way to go for most recommendation systems that are large embeddings focused. see torchrec, nvidia's hugectr

really you should treat your entire cloud deployment (sans state where impossible) as cattle

did nobody read the article lol

> The organisers say the sequence is often picked. But some have alleged a scam and an investigation is under way.

> It is extremely rare for multiple winners to share the jackpot.

that sequence happens, but multiple winners is rare

so the combination of them is the concern here

yep demand side subsidies do not have good outcomes

supply side are a bit better

wrt cost disease

see how milk has remained the same price for decades

spiritual successor

AGNB needs some time to grow into longer format

they're literally 3 dudes in an RV without a working toilet

if they had BBC money they would live up to Theroux 100%

you only need two operators, the kubeoperator, the dataflow operator

and then every task is just a standalone, vertically scalable service on k8s or a giant horizontally scalable compute job

the thing is humans have most efficiently encoded (in detail) reality in text. humans already highlight what is worth encoding about reality.

for example, you can finetune gpt-2 to have an idea of sexual biology by having it read erotica. just like how you can have a model learn the same by watching porn. but it is much more efficient to read the text, since there is much less information that is "useless"

AGNB is the spiritual successor to Louis Theroux

it's what Vice tried to capture but failed. parachute journalism works, only if you inject very little of yourself into it and have minimal framework.

anybody know if 5g is worth waiting for?

lol it is so rare for a company that goes belly up to hurt someone's career

it's a specialized swab shortage

It really looks more and more like JAX is the internal winner after the tf 2.0 fiasco

not op...

google deployment manager. it has incredibly subpar support for other GCP services. all support interactions resort to them suggesting migration to Terraform. we do use it in production, but not without great headache.

this isn't an explicit kill-off, but certainly purposefully offering bad support