Progressive Politics

2565 readers

292 users here now

Welcome to Progressive Politics! A place for news updates and political discussion from a left perspective. Conservatives and centrists are welcome just try and keep it civil :)

(Sidebar still a work in progress post recommendations if you have them such as reading lists)

founded 2 years ago

MODERATORS

[email protected]

145

Researchers Secretly Ran a Massive, Unauthorized AI Persuasion Experiment on Reddit Users (www.404media.co)

submitted 20 hours ago* (last edited 20 hours ago) by [email protected] to c/[email protected]

30 comments fedilink hide all child comments

The researchers' bots generated identities as a sexual assault survivor, a trauma counselor, and a Black man opposed to Black Lives Matter.

archive link

you are viewing a single comment's thread
view the rest of the comments

[–] [email protected] 45 points 19 hours ago (3 children)

So the issue here is it was AI this time?

There have been effort by individuals and coordinated groups doing this kind of thing forever. It’s a reminder that you should not fully form your opinion on comments/posts on social media alone.

[–] [email protected] 8 points 14 hours ago (1 children)

I fully agree with this and formed my opinion based on your comments and posts. I'll be getting all my opinions from you from now on.

[–] [email protected] 6 points 13 hours ago (1 children)

Oh I wouldn’t do that. That guy is a moron.

[–] [email protected] 2 points 8 hours ago

As you command

[–] [email protected] 23 points 19 hours ago* (last edited 19 hours ago) (2 children)

The study showed that the AI bots were between 3 and 6 times more persuasive than humans. And the study also mentioned that their bots were not recognized as AI, ever. Not once.

We are now at the point where the side with the most AI wins elections.

Guess which side has the most AI.

[–] [email protected] 6 points 9 hours ago

Democrats can start winning whenever they choose. They just choose wealthy donors over messaging that would win elections. Bots would be irrelevant.

[–] [email protected] 32 points 18 hours ago (2 children)

As was pointed out in the actual reddit discussions, ChangeMyView has a strict "do not accuse the other person of being a bot or troll" rule. So the whole "no one knew it was a bot" part has very little merit.

[–] [email protected] 14 points 18 hours ago

I thought that sounded fishy. It's common knowledge that everyone on the internet except for you is a bot so a contentious discussion on reddit without anyone accusing one of the primary commenters of being a bot seemed questionable without that bit of additional information.

[–] [email protected] 4 points 18 hours ago

lol, good to know!

[–] [email protected] 18 points 19 hours ago (3 children)

The issue is, with AI you can have an agent personalized for every person, pretending to be their friend, manipulating them individually, to move society in a way that's negative for 99.999% of people.

[–] [email protected] 14 points 19 hours ago

Maybe the real friends were the AI bots we made along the way.

[–] [email protected] 5 points 19 hours ago

I think that’s hyperbolic (at least at today’s “AI” capabilities) all of the posts and comment in this ”research” were reviewed and posted by a human researcher.

Even if the technology improves to the point where that is no longer necessary. Even the billionaires are finding out you only need a few bad eggs to spoil the basket

[–] [email protected] 2 points 18 hours ago

Really? I can make an agent that can convince me it's my friend?