3 ms·

What I meant by state is status. As in, this type of humble retro styled application is where we find GPT-2 in 2020 instead of dominating the news cycle with wh

What I meant by state is status. As in, this type of humble retro styled application is where we find GPT-2 in 2020 instead of dominating the news cycle with what you claim it is so good at. If it excels at text, where are the fake articles and tweets? It's just not that good.

ehsankia 6y ago
By definition, if the tweets and articles are good, they are unnoticeable from real ones. That's exactly what makes them dangerous. If you could detect them, then they wouldn't be dangerous anymore.
- renewiltord 6y ago
  You should see the too-dangerous-to-release AI model my team built. It actually governs nations in a manner indistinguishable from a human. There are nations currently under its thrall that you wouldn't even imagine. The world leaders of these nations merely speak what speeches it writes for them.
djannzjkzxn 6y ago
Have you been reading the mass comments/tweets on political stories lately? Some of it does seem bot-like to me. I don’t have a strong opinion about how much this is happening. But I know I could personally run a bot that generates political spam with GPT-2. I know a lot of people want to influence the political conversation. I mean, people get paid salaries to write comments online. So I can’t help but suspect that someone is using tools that allow it to be done a lot more cheaply.
- seanwilson 6y ago
  Saying that it could be is wide use but you wouldn't be able to detect it is an impossible to falsify claim though.
  - djannzjkzxn 6y ago
    The way to falsify it is to track down the identity of the human author of every internet comment. Simple! To get philosophical I won’t say that I “know” it’s happening. Most of the world consists of things I don’t know about. The best I can do is build a mental model based on what I do know.
gwern 6y ago
It does work, though. Here's an example: https://techscience.org/a/2019121801/ https://techscience.org/a/2019121801/ There's also https://arxiv.org/pdf/1908.09203.pdf#page=48 https://arxiv.org/pdf/1908.09203.pdf#page=48 Why aren't bad actors using it in the wild? I think it's a combination of them being technically unsophisticated and conservative, propaganda not actually working nearly as well as people like to think it does, and lack of detection of competent actors (things like StyleGAN being used for fake FB profiles are detected through carelessness like leaving the faces exactly aligned).

ehsankia 6y ago

By definition, if the tweets and articles are good, they are unnoticeable from real ones. That's exactly what makes them dangerous. If you could detect them, then they wouldn't be dangerous anymore.

renewiltord 6y ago
You should see the too-dangerous-to-release AI model my team built. It actually governs nations in a manner indistinguishable from a human. There are nations currently under its thrall that you wouldn't even imagine. The world leaders of these nations merely speak what speeches it writes for them.

djannzjkzxn 6y ago

Have you been reading the mass comments/tweets on political stories lately? Some of it does seem bot-like to me. I don’t have a strong opinion about how much this is happening. But I know I could personally run a bot that generates political spam with GPT-2. I know a lot of people want to influence the political conversation. I mean, people get paid salaries to write comments online. So I can’t help but suspect that someone is using tools that allow it to be done a lot more cheaply.

seanwilson 6y ago
Saying that it could be is wide use but you wouldn't be able to detect it is an impossible to falsify claim though.
- djannzjkzxn 6y ago
  The way to falsify it is to track down the identity of the human author of every internet comment. Simple! To get philosophical I won’t say that I “know” it’s happening. Most of the world consists of things I don’t know about. The best I can do is build a mental model based on what I do know.

gwern 6y ago

It does work, though. Here's an example: https://techscience.org/a/2019121801/ https://techscience.org/a/2019121801/ There's also https://arxiv.org/pdf/1908.09203.pdf#page=48 https://arxiv.org/pdf/1908.09203.pdf#page=48 Why aren't bad actors using it in the wild? I think it's a combination of them being technically unsophisticated and conservative, propaganda not actually working nearly as well as people like to think it does, and lack of detection of competent actors (things like StyleGAN being used for fake FB profiles are detected through carelessness like leaving the faces exactly aligned).