3 ms·

This was related to OpenAI announcing that they had shut down several covert influence campaigns using ChatGPT-generated social media posts. https://openai.com/

This was related to OpenAI announcing that they had shut down several covert influence campaigns using ChatGPT-generated social media posts. https://openai.com/index/disrupting-deceptive-uses-of-AI-by-covert-influence-operations/ https://openai.com/index/disrupting-deceptive-uses-of-AI-by-...

dang 2y ago
Disrupting deceptive uses of AI by covert influence operations - https://news.ycombinator.com/item?id=40526068 https://news.ycombinator.com/item?id=40526068 - May 2024 (70 comments)
sva_ 2y ago
This article itself seems to be written by chatgpt to some degree at least? I've developed trust-issues with bullet point lists in that format.
- rfw300 2y ago
  If anything, OpenAI's affection for bullet lists is probably a window into why ChatGPT uses them so frequently.
  - choxi 2y ago
    Claude does bullet lists a lot too, I think people just like bullet lists over blobs of text
dmix 2y ago
> As of May 2024, these campaigns do not appear to have meaningfully increased their audience engagement or reach as a result of our services. Just like I've predicted many times on HN. Text generation is low on the list of things needed to successfully engage in automated spam. Social media is built on reputation, not who can write generic believable text the quickest. And funny enough the ex-OpenAI people (mostly Helen) calling for gov regulation said GPT should not have been released to the public because of this risk.
- sillyfluke 2y ago
  In a way, I find your comment unintentionally hilarious. Sure, this may actually be true. Consider the source, however. Congratulating yourself by quoting the tobacco company's own research of (the lack of) adverse effects of tobacco seems a tad ill-conceived. The talk of reputation makes it doubly so, since on this topic Sam Altman or any official OpenAI post has no credibility whatsoever.
- kelnos 2y ago
  I'm not sure I would just take their word on it, though. They have a pretty strong incentive to claim that their service didn't cause any harm.
- barfingclouds 2y ago
  A lot of social consensus is often made when huge threads on twitter or Instagram have a massive number of supporters behind a topic. So when bots get better at all saying the same thing in seemingly human voices, it trains us to believe that is popular, and likely that that is right
- Aerbil313 2y ago
  > Social media is built on reputation, not who can write generic believable text the quickest. This is not the whole picture at all. I’d even say mostly incorrect. You can definitely influence public opinion with a social media consisting entirely of bots, provided the public aren’t aware of it.

dang 2y ago

Disrupting deceptive uses of AI by covert influence operations - https://news.ycombinator.com/item?id=40526068 https://news.ycombinator.com/item?id=40526068 - May 2024 (70 comments)

sva_ 2y ago

This article itself seems to be written by chatgpt to some degree at least? I've developed trust-issues with bullet point lists in that format.

rfw300 2y ago
If anything, OpenAI's affection for bullet lists is probably a window into why ChatGPT uses them so frequently.
- choxi 2y ago
  Claude does bullet lists a lot too, I think people just like bullet lists over blobs of text

dmix 2y ago

> As of May 2024, these campaigns do not appear to have meaningfully increased their audience engagement or reach as a result of our services. Just like I've predicted many times on HN. Text generation is low on the list of things needed to successfully engage in automated spam. Social media is built on reputation, not who can write generic believable text the quickest. And funny enough the ex-OpenAI people (mostly Helen) calling for gov regulation said GPT should not have been released to the public because of this risk.

sillyfluke 2y ago
In a way, I find your comment unintentionally hilarious. Sure, this may actually be true. Consider the source, however. Congratulating yourself by quoting the tobacco company's own research of (the lack of) adverse effects of tobacco seems a tad ill-conceived. The talk of reputation makes it doubly so, since on this topic Sam Altman or any official OpenAI post has no credibility whatsoever.
kelnos 2y ago
I'm not sure I would just take their word on it, though. They have a pretty strong incentive to claim that their service didn't cause any harm.
barfingclouds 2y ago
A lot of social consensus is often made when huge threads on twitter or Instagram have a massive number of supporters behind a topic. So when bots get better at all saying the same thing in seemingly human voices, it trains us to believe that is popular, and likely that that is right
Aerbil313 2y ago
> Social media is built on reputation, not who can write generic believable text the quickest. This is not the whole picture at all. I’d even say mostly incorrect. You can definitely influence public opinion with a social media consisting entirely of bots, provided the public aren’t aware of it.