Add content/story/turkerstricking.md
This commit is contained in:
parent
bf4d578004
commit
c089eadeb0
1 changed files with 10 additions and 0 deletions
10
content/story/turkerstricking.md
Normal file
10
content/story/turkerstricking.md
Normal file
|
@ -0,0 +1,10 @@
|
|||
+++
|
||||
title = "Turkers Tricking the Turk"
|
||||
+++
|
||||
|
||||
*First heard on the net, 2021. Brought to bonfires in Berlin and Orléans in 2024.*
|
||||
|
||||
The unsung heroes of the gig economy, or perhaps more accurately, its foot soldiers, are the workers on platforms like Amazon Mechanical Turk, where “Turkers” tackle micro-tasks for pocket change. Each nugget of data they polish typically earns them as little as 0.05 US dollars.
|
||||
But there’s a twist in the tale: AI needs humans to curate and annotate training datasets, as well as to check the outputs generated by the models to guide them to produce more accurate results. However, some researchers recently uncovered that up to 46% of these Turkers are using AI to beat AI at its own game. They’re roping in the likes of ChatGPT to hustle through tasks more quickly and efficiently, squeezing more money out of the tech giants’ deep pockets.
|
||||
|
||||
And there is a further twist: As the Turkers started to use AI to take up parts of their jobs, the AIs that they are training are starting to turn into funhouse mirrors of echos—producing data that is distorted, biased, and just plain weird. Therefore, like in a high-stakes game of digital cat and mouse, the AI companies are currently trying to sidestep this self-feeding loop, while the crafty Turkers are forcing them to either cough up more money for their human touch, or concoct new ways to phase it out.
|
Loading…
Add table
Reference in a new issue