04/16 - Streamlined Sunday!

Scenario Study
By stanford.edu

• £1.10 • £9.45/hr

• 7 mins

• 332 places

 

I cut my front yard and half my back yard before the angry clouds and lightning rolled in yesterday. I like to spread lawn mowing out over a couple days anyway, good excuse for some physical activity for more than one day


Here's this:
https://www.artisana.ai/articles/bye-bye-mechanical-turk-how-chatgpt-is-making-humans-obsolete
"Bye-Bye, Mechanical Turk? How ChatGPT is Making Humans Obsolete"

 

TWO-PART Political judgment study Part 2 is up at Connect

 

Hoori [A3DVQKR67792EJ] Academic study pays $0.20 for a short 2-minute study(~ 2 minutes) - $0.20 | PANDA

Underpaid

Unrated

Unrated

$8.89 / hour
00:01:21 / completion time

​

Pros:

Cons:

Catchability: Great (HIT Last Seen: a second ago)
Qualifications: Exc: [13228479-412831] DoesNotExist ; rhsnlk98 >= 100; Location In US;

To read Hoori's full profile check out TurkerView!​

 

"In a complicated task such as Stance Detection, where ChatGPT and crowd-workers were asked to assess whether a tweet supported, opposed, or remained neutral regarding a piece of US legislation, ChatGPT consistently demonstrated three times greater accuracy than Amazon's most elite Mechanical Turk workers based in the U.S.

These findings were all achieved without the use of GPT-4, which has since demonstrated further gains in comprehending and processing intricate topics and prompts.

One startup CTO, who spoke on the condition of anonymity and whose company relies on services like Labelbox and Mechanical Turk for machine learning labels, envisions a bleak future for the industry. "In a few years," he remarked, "I'm not sure some of these services may be around at all. They'll need to start reinventing themselves now to stay relevant."

 

someone should tell them that elite means that you won a random lottery 8+ years ago...

 

and I don't know. most of the steady money on the platform these days comes from training those chatbots on their basic language processing SO....

 

Ya also have to wonder what the referenced study was paying and if it was underpaid, they skews the results bc workers aren't going to put much effort into something that isn't paying well. But then again, I guess that is besides the point bc they dont have to pay ChatGPT

 

This is the second article I've read about this. I think they're correct with repetitive tasks, but I don't think ChatGPT could do surveys.

 

It can do surveys, but it would be pointless bc the point is to measure humans not machines

 

Yeah that's what I meant. The surveys are subjective.

 

Limited time for this to remain true though. Once they get the models trained enough, work will drop significantly. It could possibly continue to go on for quite a long time though

 

a lot of my recent batchwork has been more about feelings and emotions and I don't see how they can have GPT assess that.

 

Yeah ChatGPT isn't quite there yet. It gives me false information quite a bit.

 

Academic Research [A3PE2WFW7YARFR] Product Survey(~ 2 minutes) - $0.20 | PANDA

Unrated

Unrated

Unrated

$13.58 / hour
00:00:53 / completion time

​

Pros:

Cons:

Catchability: Great (HIT Last Seen: a second ago)
Qualifications: Exc: [407936788-412834] DoesNotExist ; rhsnlk98 >= 100; HIT approval rate (%) >= 95; Location In US;

To read Academic Research's full profile check out TurkerView!​

 

Personally I think AI is way overblown in general, but it definitely seems like it will nail down the more basic stuff pretty well.

Some tasks that don't require much human insight could easily disappear

And also when it comes to certain kinds of work, a lot of big companies have a decently high tolerance for error so long as the numbers work out in relation to the money being made. Customer Service for example, they don't really care how good the service actually is but how much it costs in relation to profit

 

Anthony [A1FV3BP839KKQJ] Quick 2-3 Minute Survey(~ 2 minutes) - $0.30 | PANDA

Unrated

Unrated

Unrated

$11.74 / hour
00:01:32 / completion time

​

Pros:

Cons:

Catchability: Great (HIT Last Seen: a second ago)
Qualifications: Total approved HITs >= 5000; Exc: [404514452-412836] DoesNotExist ; rhsnlk98 >= 100; HIT approval rate (%) >= 99; Location In US;

To read Anthony's full profile check out TurkerView!​

 

Anyone done this? Its been staring at me all weekend... looks bad


Measuring Social Media Message Features via Crowdsourcing
By wisc.edu

• £3.15 • £6.00/hr

• 31 mins

• 12 places

 

no. what's the estimated time? have you hovered over it to see the adjusted?