02/10 - Swifty Sunday!

AnaBot confirmed

 

Good Morning!

 

Wtf chelsea

 

honestly nothing i have seen seems written with any coherence....

 

You've received a bonus from Rowan Zellers for work related to 3XEIP58NL0UN60WB2CSWOIOMSVDZLA.
The value of your bonus is: $0.07 USD

The Requester included this note:
Hi there! Thanks for doing my HITs (Classifying caption endings). Here's some more feedback, which hopefully can help inform what I'm looking for. Anyways, each caption has one real ending - the other five were generated by AIs. You've ranked the real ending as best or 2nd best 8/15 times, or 53%. Not bad - try to be careful about telling apart AI generated vs real endings.
Ranking the real ending as the best is the main thing that I'm looking for. You ranked the real endings likely 60% of the time, unlikely 40% of the time, and gibberish 0%.
You ranked the AI-generated endings likely 30% of the time, unlikely 62% of the time, and gibberish 6%.

-he put some examples I'm redacting for content reasons-

A disclaimer: don't worry about perfectly telling apart AI vs. human endings. There's some ambiguity, and what I'm looking for is to tell which AI endings humans prefer. (That way, we can reward the AI for generating sensible endings!) For that reason, it doesn't help me if you cherry pick the easy ones - just do your best, and don't guess wildly
More HITs coming! thanks, -Rowan

 

@rowan your feedback system is making no sense

 

Title: Judge if the tweet is related to alcohol consumption | PANDA Requester: Marcos Grzeça [A2AYNQMGCVD3W8] (Req TV): $-/hr (TO): [Pay: N/A] [Fair: N/A] [Comm: N/A] [Fast: N/A] Description: Judge if the tweet is related to alcohol consumption Time: 2 minutes(s) HITs Available: 200 Reward: $0.01 Qualifications: Masters Exists ;

 

I submit that if you're developing an AI system and people are having trouble distinguishing AI responses from human responses, maybe your AI system is pretty good.

 

except when everything written is mostly incoherent

 

Mattheus Brenig [A1AC1BTIMTFPZL] Slider Task (you can earn 1.5$ in approximately 8 minutes) - $0.00 $1.50 bonus | PANDA

Unrated

<24 hours

Unrated

$9.63 / hour
00:09:21 / completion time

Approved! ​

Pros:
Instant approval + bonus

Cons:

Qualifications: Block Past Workers DoesNotExist ; Total approved HITs >= 100; HIT approval rate (%) >= 90; Location EqualTo US;

To read Mattheus Brenig's full profile check out TurkerView!​

 

Maybe I'm just a slow reader but I couldn't get these new Rowans to pay more than $10/hr.

 

Ah. Well, having done none of the HITs in question, I'm just making general commentary. Carry on.

 

When the 'prompt' sentence doesn't make sense, it gets hard to pick anything as a viable ending.

 

Your HIT activity for this week:
- Number of HITs submitted: 6,107

Approvals and payments that occurred this week:
- Number of HITs approved: 3,902
- Number of HITs rejected: 0
- Bonus reward earned: $19.54
- HIT reward earned: $633.84
- Total Amount earned this week: $653.38

 

Mine got revoked and I seem to be doing better than you stats wise.

You've ranked the real ending as best or 2nd best 25/32 times, or 78%. Great job! Ranking the real ending as the best is the main thing that I'm looking for. You ranked the real endings possible 78% of the time, unlikely 21% of the time, and gibberish 0%.

 

Greetings from Amazon Mechanical Turk,

The following is a summary of activity for your Mechanical Turk account for the week ending Feb 09, 2019.

Your HIT activity for this week:
- Number of HITs submitted: 257

Approvals and payments that occurred this week:
- Number of HITs approved: 284
- Number of HITs rejected: 1
- Bonus reward earned: $12.80
- HIT reward earned: $111.16
- Total Amount earned this week: $123.96

For more information about your account status and activity, sign in to
https://www.mturk.com/?referrer=email, then view your "Dashboard".

Sincerely,
Amazon Mechanical Turk
https://www.mturk.com

 

lowered or revoked? Can you do the batch he just dropped?

 

ASU PL LAB [A26S4BHWY8M7E1] Perception of Friendship (10-minute study)(~ 10 minutes) - $1.00 | PANDA

Generous

Unrated

Unrated

$18.85 / hour
00:03:11 / completion time

​

Pros:

Cons:

Qualifications: Exc: [128692-128103] DoesNotExist ; Total approved HITs >= 50; HIT approval rate (%) >= 75; Location In US;

To read ASU PL LAB's full profile check out TurkerView!​

 

Revoked. I got a very nice email about it.