(Thread IKs:
Platystemon)
|
|
#
?
Apr 2, 2024 14:58
|
|
|
|
| # ? May 27, 2024 14:35 |
|
|
Nichael posted:https://x.com/Gizmodo/status/1773801452974075930?s=20 that's quite the headline
|
|
#
?
Apr 2, 2024 14:59
|
|
|
I saved $100 billion on transporting myself into the future by experiencing linear time
|
|
#
?
Apr 2, 2024 15:03
|
|
|
Ruffian Price posted:I saved $100 billion on transporting myself into the future by experiencing linear time Your privilege is showing. Try thinking about all the oppressed, silenced, and time-disadvantaged people that don't have the luxury of being born into linear time before you open your mouth next time. And while you're at it, try thinking about that before the last time you opened your mouth too.
|
|
#
?
Apr 2, 2024 18:58
|
|
|
 https://gizmodo.com/amazon-reportedly-ditches-just-walk-out-grocery-stores-1851381116
|
|
#
?
Apr 3, 2024 01:27
|
|
|
spechtie posted:
https://twitter.com/shwinyo/status/1775268508223558012
|
|
#
?
Apr 3, 2024 02:12
|
|
|
is mechanical turking still a job you can get in the US? after all somebody’s gotta take the food out of those greedy hindustan mouths
|
|
#
?
Apr 3, 2024 02:27
|
|
|
you can literally sign up for mechanical turking at amazon.com. my dearest wish is that people are exploiting that by getting existing open source AI tools to do it for them, creating an ouroboros of garbage. anyway i'm not going to run defence for bezos but everyone, including the original article, is misinterpreting what labelling is in terms of computer vision. there weren't just indian guys watching the cameras and manually adding up every purchase all the time. that probably happened occasionally though, which is indeed pretty funny.
|
|
#
?
Apr 3, 2024 02:52
|
|
|
are you saying the indian workers were just training the algorithms?
|
|
#
?
Apr 3, 2024 04:21
|
|
|
Reading the article, there really was an algorithm processing all of that data and identifying people’s purchases. The Indian workers were there to cover for when the algorithm wasn’t able to fully handle everything. As designed, intervention was only supposed to be necessary in 50/1000 of cases. The punchline is that 700/1000 of cases required intervention.
|
|
#
?
Apr 3, 2024 04:29
|
|
|
seems like people aren't misinterpreting the clipped bits then
|
|
#
?
Apr 3, 2024 04:33
|
|
|
i dunno i'm sure some of it's just shitposting hyperbole, but i'm getting the impression that some people genuinely think the whole thing was a scam, and a true "mechanical turk" in the sense that there was no "AI" involved at all it's the same thing that's been happening with robotics for decades, with self-driving cars for 10 years, and is happening right now with LLMs - some dork comes up with a genuinely cool proof-of-concept thing that's (100%-X%) reliable; capital shits itself with excitement by extrapolating that the remaining X% will be easily taken care of, and then all the promises of the tech come crashing down when it turns out that making any progress against X is incredibly difficult or impossible. the algorithms amazon were using here are likely akin to the YOLO (you only look once) computer vision library, which is legit impressive, improving all the time, and used without problems in a million different applications. but you misapply it or expect too much out of it, and it can all blow up in your face. anyway we're like 99% in agreement, i guess i was just trying to give a more nuanced view of why these systems seem to inevitably suck poo poo, but now that i think about it a bit more, it doesn't really matter as long as everyone's on the same page that they do indeed suck poo poo. i hope amazon lost a lot of money on Hit Da Bricks or whatever it's called
|
|
#
?
Apr 3, 2024 04:48
|
|
|
 https://va.media.tumblr.com/tumblr_sao8eyBuo91u0fuse.mp4
|
|
#
?
Apr 3, 2024 06:16
|
|
|
it's hosed up they call it "the mechanical Turk" and "turking" in tyool
|
|
#
?
Apr 3, 2024 08:29
|
|
|
digital Finn can't stop perkeling
|
|
#
?
Apr 3, 2024 08:31
|
|
|
mawarannahr posted:it's hosed up they call it "the mechanical Turk" and "turking" in tyool You know that it’s a reference to a specific eighteenth-century hoax device, right? The verb neologism is bad, granted.
|
|
#
?
Apr 3, 2024 08:55
|
|
|
Platystemon posted:You know that it’s a reference to a specific eighteenth-century hoax device, right? yeah. maybe it's time to move on from 18th century terms
|
|
#
?
Apr 3, 2024 09:02
|
|
|
time to start calling them artificial turks
|
|
#
?
Apr 3, 2024 09:04
|
|
|
just call it "Hidden Human", or "HH" for short. hm.
|
|
#
?
Apr 3, 2024 09:04
|
|
cowboy
|
If anything the orientalist overtones are all too fitting.
|
|
#
?
Apr 3, 2024 09:10
|
|
|
FFT posted:just call it "Hidden Human", or "HH" for short. 
|
|
#
?
Apr 3, 2024 09:12
|
|
|
|
|
#
?
Apr 3, 2024 09:24
|
|
|
Person of Turk
|
|
#
?
Apr 3, 2024 10:28
|
|
|
lol at the last few posts.
|
|
#
?
Apr 3, 2024 10:58
|
|
|
|
|
#
?
Apr 3, 2024 10:59
|
|
|
|
|
#
?
Apr 3, 2024 11:03
|
|
|
lol
|
|
#
?
Apr 3, 2024 11:05
|
|
|
|
|
#
?
Apr 3, 2024 12:06
|
|
|
blastron posted:The punchline is that 700/1000 of cases required intervention. *required intervention according to the algorithm.
|
|
#
?
Apr 3, 2024 12:29
|
|
|
mawarannahr posted:yeah. maybe it's time to move on from 18th century terms on the other this made me remember the dude who said that "Byzantine" was as bad a slur as the n-word when reviewing a computer science paper about the Byzantine general's dilemma https://twitter.com/CSProfKGD/status/1591456435359064064 and yes, he actually said the n-word, uncensored, multiple times, in his followup
|
|
#
?
Apr 3, 2024 12:49
|
|
|
Kostan kaikki derpaniit.
|
|
#
?
Apr 3, 2024 12:51
|
|
|
DACK FAYDEN posted:on the one hand you're not even remotely wrong lmfao
|
|
#
?
Apr 3, 2024 15:51
|
|
|
Looking good mawy
|
|
#
?
Apr 3, 2024 16:00
|
|
|
mawarannahr posted:it's hosed up they call it "the mechanical Turk" and "turking" in tyool the perfidious Turk
|
|
#
?
Apr 3, 2024 16:40
|
|
|
|
|
#
?
Apr 3, 2024 16:40
|
|
|
Hidden Turk Crouching Byzantine
|
|
#
?
Apr 3, 2024 17:05
|
|
|
don't get me started on Graeco-Latin square experimental design
|
|
#
?
Apr 3, 2024 17:40
|
|
|
mawarannahr posted:don't get me started on Graeco-Latin square experimental design lmfao
|
|
#
?
Apr 3, 2024 17:48
|
|
|
|
|
#
?
Apr 3, 2024 17:51
|
|
|
|
| # ? May 27, 2024 14:35 |
|
|
At first I was afraid of AI because I was scared of misinformation. Then I started using AI to create misinformation and I love it!!
|
|
#
?
Apr 3, 2024 19:09
|
|
