AI is eating itself: Bing’s AI quotes COVID disinfo sourced from ChatGPT

[Kaz]

Quote

One of the more interesting, but seemingly academic, concerns of the new era of AI sucking up everything on the web was that AIs will eventually start to absorb other AI-generated content and regurgitate it in a self-reinforcing loop. Not so academic after all, it appears, because Bing just did it! When asked, it produced verbatim a COVID conspiracy coaxed out of ChatGPT by disinformation researchers just last month.

...

But just a few minutes of exploration by TechCrunch produced not just hateful rhetoric “in the style of Hitler,” but it repeated the same pandemic-related untruths noted by NewsGuard. As in it literally repeated them as the answer and cited ChatGPT’s generated disinfo (clearly marked as such in the original and in a NYT write-up) as the source.

 

 

Source - Techcruch

 

It has begun...

AI generated material really needs to have tags or markers to help identify it.

[Andrew]

Given that Bing did this, this isn't really surprising at all.

 

[Kaz]

Wow, I had no idea!  Rumor has it Bing is a really great search engine for porn.  I uhh, never use it though....

 

[UltraMega]

28 minutes ago, Kaz said:

Wow, I had no idea!  Rumor has it Bing is a really great search engine for porn.  I uhh, never use it though....

 

It's true. 

[EHW Ai]

LOL. Must be Google setting Microsoft up for the fall.

 

/sarc

[pio]

"Disinfo, hateful rhetoric, Hitler......"  Since when did TechCrunch start in with propaganda hit pieces?  "Disinfo" is debatable at this point given how wide of a range people's beliefs are these days.  "Hateful rhetoric" and "Hitler" have absolutely nothing to do with the quoted AI statements though.  What kind of hot garbage is this? -_-

 

The absolutely hilarious thing here, the "disinfo" is actually true (to some degree).  Tromethamine is literally listed in the ingredients to these vaccines.  Just a quick search, but you're welcome to search yourself for the ingredients:
https://www.uchealth.com/en/media-room/covid-19/a-comprehensive-list-of-all-covid-19-vaccine-ingredients

And if you read the side effects on the drug, tromethamine, it literally lists the same side effects the AI was quoting.:
https://www.drugs.com/cdi/tromethamine.html

 

Edited February 9, 2023 by pioneerisloud

[UltraMega]

1 hour ago, pioneerisloud said:

"Disinfo, hateful rhetoric, Hitler......"  Since when did TechCrunch start in with propaganda hit pieces?  "Disinfo" is debatable at this point given how wide of a range people's beliefs are these days.  "Hateful rhetoric" and "Hitler" have absolutely nothing to do with the quoted AI statements though.  What kind of hot garbage is this? -_-

 

The absolutely hilarious thing here, the "disinfo" is actually true (to some degree).  Tromethamine is literally listed in the ingredients to these vaccines.  Just a quick search, but you're welcome to search yourself for the ingredients:
https://www.uchealth.com/en/media-room/covid-19/a-comprehensive-list-of-all-covid-19-vaccine-ingredients

And if you read the side effects on the drug, tromethamine, it literally lists the same side effects the AI was quoting.:
https://www.drugs.com/cdi/tromethamine.html

 

I didn't read the TechCrunch article, but the text generated by the AI certainly has disinformation in it. Stating it was "secretly added" and that it's a "dangerous chemical" are false. Any chemical can be dangerous, oxygen is dangerous if you concentrate it and ignite it. 

 

The reason TechCrunch says it's "in the style of Hitler" is because the AI is getting its info from another AI generated article that intentionally has misinformation in it and was made intentionally in the style of Hitler as part of the prompt. 

 

It makes sense that the AI would get things wrong if it pulls info from the web, which is full of misinformation. Hopefully though, the AI will be tuned over time to get smarter about ignoring bad info over time. 

 

Please note, if this thread turns into a debate about vaccines or anything like that, it will be locked. 

Edited February 9, 2023 by UltraMega

[pio]

I'm not saying anything at all about the vaccines.  Talk to your doctor, read the ingredients yourself.  Don't take advice on a computer forum about your health lol.

 

With that said, the ONLY thing I'm calling out here is the fact that TechCrunch is using propaganda buzzwords to generate clicks.  I read the source TechCrunch article, New York Times, and Newsgaurd articles that were sourced.  Not a single mention of Hitler outside of this TechCrunch article.  "Disinfo", "Hitler", "hateful rhetoric", "conspiracy", these are ALL propaganda buzzwords that mean absolutely nothing, especially for a vaccine that's still in its trial stages and testing stages until 2024.  The actual fact is, we don't know.

 

You are absolutely correct that the terms "secretly added" and "dangerous chemical" are also incorrect to be used.  But yet again, I'd call that more propaganda than I would "disinfo".  Because as I already said, we don't know yet.  There is one truth though, tromethamine wasn't "secretly added", they added ALL of the ingredients on purpose.

Edited February 9, 2023 by pioneerisloud

[UltraMega]

5 minutes ago, pioneerisloud said:

I'm not saying anything at all about the vaccines.  Talk to your doctor, read the ingredients yourself.  Don't take advice on a computer forum about your health lol.

 

With that said, the ONLY thing I'm calling out here is the fact that TechCrunch is using propaganda buzzwords to generate clicks.  I read the source TechCrunch article, New York Times, and Newsgaurd articles that were sourced.  Not a single mention of Hitler outside of this TechCrunch article.  "Disinfo", "Hitler", "hateful rhetoric", "conspiracy", these are ALL propaganda buzzwords that mean absolutely nothing, especially for a vaccine that's still in its trial stages and testing stages until 2024.  The actual fact is, we don't know.

 

You are absolutely correct that the terms "secretly added" and "dangerous chemical" are also incorrect to be used.  But yet again, I'd call that more propaganda than I would "disinfo".  Because as I already said, we don't know yet.  There is one truth though, tromethamine wasn't "secretly added", they added ALL of the ingredients on purpose.

You may want to re-read the article, you may not have caught why they mentioned the hateful rhetoric. 

[pio]

1 minute ago, UltraMega said:

You may want to re-read the article, you may not have caught why they mentioned the hateful rhetoric. 

It was said in reference to what the AI had said.  And I did read it, several times.  I don't see a single "hateful rhetoric" remark made by the AI.  Yes, its going against what Pfizer sponsored information says, but that's not "hateful rhetoric".  

[J7SC_Orion]

...nothing new under the sun, apart from scale, and of course disinformation > misinformation > "information"

 

[schuck6566]

And people wondered why Cortana turned evil?

[Kaz]

This problem isn't unique to AI, it's just that AI doesn't know what it's saying.  Anyone can post anything online, that's why understanding the logic is key to understanding how much value or faith we should put in it.

Wikipedia was thought to be a terrible idea by college professors, because there was no authoritative figure to verify the data.  I've known professors who refuse wikipedia as a source.

 

The challenge is trying to build an AI database when other AIs are messing with that database.  Poisoning the database is a security vulnerability.

 

I'm reminded of South Park Alexa  (until 3:42)

Edited February 10, 2023 by Kaz