gerikson @ gerikson @awful.systems

帖子

44

评论

1,464

加入于

2 yr. ago

That popular piece on why it's dumb to build (GenAI-scale) in space hit lobste.rs, and while most commenters agreed it is indeed dumb, fascist Flask founder felt the need to "well ackstually" for some stupid reason

https://lobste.rs/c/mmq7sy

duh ofc there are computers in space, there are computers everywhere, but the whole fucking point of the piece is you can't take thousands of racks of GPUs and launch them into space and expect them to work

oh yeah the relationship between the fusion-device wielding 30-something Aluetian freedom fighter and the 16 year old skateboard courier in Snow Crash is... of its time

coming soon to whatever remains of CBS with an introductory 1 hour lecture by Bari Weiss

Neal Stephenson comes out strong and funny against GenAI here:

https://nealstephenson.substack.com/p/a-remarkable-assertion-from-a16z

"Hypothesis 1: it was written by a clanker"

I've worked in an adjacent field (workforce planning) and I deliver B2B software support for a living, so I too have Thoughts.

At least here in Schwedenland, contact centers have been filed down by relentless cost and tech pressure to be about as automated as can be. You have websites with FAQs, simple chatbots that basically repeat the FAQ for those for whom reading more than a sentence of text is too hard, phone trees to gatekeep you from the Inner Sanctum, etc. etc. The end result is that the actual people taking the calls are gonna be the ones who can make human decisions - troubleshoot a complex issue, handle insurance claims, upsell your mortgage.

Trying to att LLM voice tech to that is just going to add another filter between the customer and the center, with the additional reputational risk of the robot fucking up and losing the customer.

There's base-level sneer as this is posted on LW, but I found this comparison of LLM in call centers to cheap human labor for the same interesting:

https://www.lesswrong.com/posts/rJatmEDcYrDQcwstT/the-economics-of-replacing-call-center-workers-with-ais#comments

He's also found a lucrative niche being the "serious" AI booster. No doubt he gets something from it, if nothing else consulting work.

FFS so many promptfondlers and gish-gallopers in there. Echoes of pro-crapto (can't criticize if you don't buy in, use case is coming bro, it's actually decentralized)

Edit the worst thing isnt't the number of fondlers, it's the upvotes they're getting.

"Enjoy" this toxic stew of prediction markets, racism, and the objectifying of women:

https://protos.com/polymarket-criticized-for-racist-post-targeting-fake-baddies/

Saw this quality nickname for prompt receivers: Chet Jippity

Linux puritans are as fissile as the OG Puritans. There's already Devuan which is Debian minus systemd, and now with XLibre. Not sure why this nerd thinks his particular non-denominational chapel will bring in the crowds to hear him sermonize.

Edit BTW Devuan for me is the epitome of FLOSS techfash.

m'lady

Semi-relevant, Matthew Garrett has won a court case against the kooks running Techrights:

https://nondeterministic.computer/@mjg59/115581959497817474

About the only positive thing you can say about Schestowitz (and probably his wife) is that they are rabidly anti-AI, but that is only because they are also rabidly against anyone who does not subscribe to their personal purity-test vision of Free Software (basically it's them, RMS, and maybe his parrot).

I have an unhealthy interest in them, because during the Andrew Lee putsch they were basically on his side, until he fucked even them, and I kinda drifted into their IRC server to see what was up. It was my unwelcome introduction to the toxic underbelly of FLOSS, with rampant RMS-worship, misogyny, racism and incipient techno-fascism. I ducked out quite quickly.

Tellingly, Techrights is all in on the Gemini protocol.

Paul Campos - The medium is the mob

Discusses the "reverse Flynn effect", filter bubbles, and infopollution via GenAI

This is Hercule Poirot levels of observation. Kudos.

210 IQ is not enough

Being really "smart" (along some axis) doesn't make you succesful or even happy.

also a massive massive promptfondler

While you partied

I studied THE BLADE

WTF are databricks? LLM feces?

Schrödinger's essay except the cat is always dead