Bots Have Taken Over Nearly Half the Internet, but Almost a 3rd of Users Can't Tell Difference Options

[Christian J]

https://www.theepochtimes.com/tech/bots-hav...ference-5448718

[pandy]

Alas that site wants my email.

Are they talking about chat bots or bots that for example post at forums? I've seen (supposedly) ChatGPT in action making forum posts, but I didn't understand what the purpose was.

[Christian J]

Alas that site wants my email.

Here's the study mentioned in the article headline: https://www.ai21.com/blog/human-or-not-results

Are they talking about chat bots or bots that for example posts at forums? I've seen (supposedly) ChatGPT in action making forum posts, but I didn't understand what the purpose was.

Bots in general, it seems:

"The bot-human blur is like a magic trick ... As bots get smarter, we risk losing trust in online interactions," Daniel Cooper told The Epoch Times.

...

"Spotting bots is like finding Waldo in a crowd. Look for repetitive patterns, lack of personalization, or rapid responses. Also, trust your gut. If it feels off, it might just be," he said.

While much of the discussion of malicious or "bad bot" traffic centers on social media, the influence of maligned AI interactions has much farther-reaching consequences.

...

Reports of AI language models leaving reviews for products on sites like Amazon emerged in April this year. The bot reviews were easy to identify since the chatbot literally told readers that it was an AI language model in the first sentence.

...

"Mr. Kann said there are methods a user can employ to determine if they're interacting with another person.

Like Mr. Cooper, he suggested watching response patterns carefully.

"Bots often respond instantly and may use repetitive language."

Mr. Kann also said people should check profiles since bots often have generic or incomplete online profiles.

He added an inability to distinguish between bots and humans could lead to research accuracy challenges.

"It can lead to skewed data analytics, as bot interactions can inflate website traffic and engagement metrics."

[pandy]

So social media. That's us.

Or more likely FB, Insta and so on. The posts I saw was at a normal forum, like this one, though. The posts were on topic but yet a little off. As said, I didn't understand the purpose. Maybe someone was just training, learning how to handle the API.

I must say I'm both fascinated by and scared of what those things can do.

You know Notetab that I cling to, right? AFAK its scripting language isn't documented anywhere online. There are scripts floating around, but the sole source for the syntax is the Help file. Someone, not me, asked ChatGTP to write a script for Notetab that did this and that. And it did! It didn't get it all right, but it was pretty close. And when asked to correct it it switched to some other scripting language, I've forgotten what, so a failure but still... I'm amazed that it knew the syntax and the vocabulary of such an obscure language, used by maybe a couple of hundred people, at all! How the heck did it do that?

[Christian J]

I'm a bit concerned by this part from the survey:

People assume bots don’t make typos, grammar mistakes and use slang

People tend to identify spelling and grammatical errors, as well as the usage of slang, with humans. Therefore, when they noticed mistakes in their counterpart's messages, the first instinct of many participants was to assume they were talking to fellow humans, even though most models in the game were trained to make these types of mistakes and to use slang words.

...

People identified politeness with something less human.

Some participants assumed that if their counterpart was too polite and kind, they were probably a bot, due to the perception that people, especially online, tend to be rude and impolite.

[Christian J]

The posts were on topic but yet a little off.

Truth be told I've seen occasional posts like that since 1998.

[pandy]

Spam bots, yes. But these don't spam So what's the reason? I get it if it's, say, a political topic at for instance FB. But the ones I saw were on a more technical topic, didn't try to convince people of anything at all, just added relevant information.

[Christian J]

No I meant real people expressing themselves in strange ways for unknown reasons.

[coothead]

Is it at all possible that I may be a bot?

coothead

[Christian J]

Who isn't in this neck of the woods?

[coothead]

Off topic - I have just noticed that you and pandy have been
around these parts for exactly 17 YEARS !

Joined: 10-August 06

Blimey, I was almost a middle aged man back then.

coothead

This post has been edited by coothead: Aug 10 2023, 06:50 PM

[pandy]

So were we.

[pandy]

Maybe a man that should be listened to.
https://www.bbc.com/news/world-us-canada-65452940

[coothead]

. "We're biological systems and these are digital systems. And the big difference is
. that with digital systems, you have many copies of the same set of weights, the
. same model of the world.

. "And all these copies can learn separately but share their knowledge instantly. So
. it's as if you had 10,000 people and whenever one person learnt something,
. everybody automatically knew it. And that's how these chatbots can know so much
. more than any one person."

That sounds to me very much like The Borg.

Unfortunately Resistance Is Futile.

coothead

This post has been edited by coothead: Aug 11 2023, 10:56 AM

[pandy]

Yeah. I guess that's how ChatGTP could, at least superficially, know the obscure script language of my old editor. Maybe one of the copies specializes in Help files...

[Christian J]

. "And all these copies can learn separately but share their knowledge instantly. So
. it's as if you had 10,000 people and whenever one person learnt something,
. everybody automatically knew it. And that's how these chatbots can know so much
. more than any one person."

Maybe that's another way to identify a bot: it simply knows much more than a human would. So to pass as a human it needs to feign a bit of ignorance as well.

[Darin McGrew]

"Spotting bots is like finding Waldo in a crowd. Look for repetitive patterns, lack of personalization, or rapid responses. Also, trust your gut. If it feels off, it might just be," he said.

I enjoy taking online opinion surveys, participating in focus groups, etc. One of the things I've noticed more and more is that these surveys ask the same questions repeatedly, only in different ways. For example, a survey may ask your age, and then later ask what year you were born in. If they don't line up, then they reject you (and you don't get the reward for completing the survey). But this makes me think that this might catch humans who are just clicking through the forms as quickly as possible, but not catch bots.

[pandy]

I also do surveys. But I think they most often are just poorly constructed. I get kicked off even if my answers line up. There are lots of questions about things that have already been answered. Like they ask how many people live in your home (answer 1). Then they ask if you have children under the age of 18 living at home. Duh. Or how many vehicles to you have? (Anser 0). Next question: what brand is the car you drive the most?

[Christian J]

"Spotting bots is like finding Waldo in a crowd. Look for repetitive patterns, lack of personalization, or rapid responses. Also, trust your gut. If it feels off, it might just be," he said.

I enjoy taking online opinion surveys, participating in focus groups, etc. One of the things I've noticed more and more is that these surveys ask the same questions repeatedly, only in different ways. For example, a survey may ask your age, and then later ask what year you were born in. If they don't line up, then they reject you (and you don't get the reward for completing the survey). But this makes me think that this might catch humans who are just clicking through the forms as quickly as possible, but not catch bots.

Do you think those repeat questions are used to catch bots? I have no idea, but ISTR such repeat questions have been around a very long time. I imagine police interrogators would use them to spot lies. But I also read recently that current "AI" bots can be very incoherent in their answers, so if you ask their age you might get wildly different answers every time.

[Darin McGrew]

I also do surveys. But I think they most often are just poorly constructed. I get kicked off even if my answers line up. There are lots of questions about things that have already been answered. Like they ask how many people live in your home (answer 1). Then they ask if you have children under the age of 18 living at home. Duh. Or how many vehicles to you have? (Anser 0). Next question: what brand is the car you drive the most?

When I'm kicked out of a survey, I'm pretty sure it's because I don't fit the demographic they're looking for. For example, I say that I've never bought a new car, and that I don't ever expect to be in the market for a new car, and then I get the "We already have enough responses from your group" exit message.

But yeah, I get badly written surveys pretty regularly. I explain that I never use something, then they ask what my favorite brand of that something is, and so on. I usually try to move forward with some kind of "None of the above" or "Other" responses, but sometimes I just give up and close the tab.