05/05 - Cinco De Maro Monday

A researcher found numerous bot-responses that began with "Certainly!"
The researcher advised others to "Look out for "Certainly!" It's a LLM red flag."

https://bsky.app/profile/metacognishane.bsky.social/post/3ll5tijbmuc2q

But such a particular tendency may not persist for very long. A bot so easily detected is one that's likely to be further refined by its evil creators wanting to avoid discovery.

 

What if someone is perhaps emotionally limited and tends to write unemotionally? – Their posts might look bottish to some.

 

For example therefore it follows in conclusion it can be said wherein all else fails indisputably evidence suggests solipsism cannot be discounted.

 

I tried a little emotional outburst in one of my defenses on another site, imagining that my emotional expression would give evidence that I'm human. That didn't work, at all. The accuser's viewpoint was unchanged.

 

I'm not using the things, and I don't know the state-of-the-art with the most advanced LLMs, but from what I've read in the news, I understand that they are trained on very large sets of human-written texts — and I imagine that they can concoct seemingly "original"-looking output that emulates aspects of those texts (without directly quoting them individually, and without necessarily being easily traceable to specific texts), including seemingly "expressive" generated phrasing.

 

It's like love. If it's forced, it's not real.

 

If you don't love me at my realest, you don't deserve me at my fakest.

 

https://bsky.app/profile/noahshachtman.bsky.social/post/3logstg4vnk2u

 

One research associate's opinion about "deception" and nondisclosure in studies (even when a post-survey "debriefing" gives disclosure).

https://bsky.app/profile/jacobilee.bsky.social/post/3ktd2ghthbk2b

> "... A large number of studies that deceive participants don't really need to. They do it because its easier. As such, they fail to do their duty of beneficence to participants, and to science itself. First, deceiving participants actually does harm. ..."

Interesting question:

> "... If a person participates in one study and they learn in the debriefing that they were deceived, will they believe what the researcher says in the next study? Why should they?"

Being somewhat aware of the practice (since we've participated in many studies that utilize the technique), I sometimes disclose my awareness, in a feedback comment (or when they ask if can "guess what the study was about"); I'll note that my participation in many prior studies led me to expect that parts of a study may be initially deceptive, crafted for research purposes, and that some parts are likely varied across participant segments to measure effects of variances, etc. Yadda-yadda. – (Not that I have any research expertise of my own; but after reading hundreds of debriefings over the years, of course I'll be somewhat aware.)

Some researchers might be dismayed if they knew what potentially large percentage of their participants were "not fooled" by the designed "deceptions" and nondisclosures during a study. The research might depend in part on "successful" deception. Some researchers may therefore prefer "less-experienced" participants who are more easily "fooled" for the research purposes; newcomers who are NOT yet aware of the commonality of such deceptions.

 

Slow day for me today. See you guys tomorrow!

Today's Activity
Earnings
$45.80

 

I don't know where this conversation went, but I am hearting this post just for the adorable hamster

 

Often mistakenly called a chipmunk, the animal in the clip is a prairie dog.

https://en.wikipedia.org/wiki/Dramatic_Chipmunk

> "...The video is a five-second clip of a prairie dog (erroneously referred to as a chipmunk) turning its head while the camera zooms in and dramatic music is played. The clip became widely known [in June 2007]. The clip of the prairie dog is from [a Japanese TV show]...."

TIL

 

https://forum.turkerview.com/threads/05-06-treasure-tuesday.6103/

new thread