Ghost on Web


A virtual user on social media. An AI ghost with an initiative to send posts about public topics on Weibo (a Chinese platform that similar to Twitter)

Will people argue with it? Can people identify it as a bot?



At first, we wanted to make an ironic bot that could argue with people automatically, this type of people is called GangJing(杠精, foul fault-finder)in Chinese. The enclosed environment of the Chinese internet raise many argumentative joy killers in Weibo. But the workload in this direction was huge, so we focused on one of the most controversial and combative topics on the Chinese Internet: actors in the entertainment industry. There are a lot of young celebrities in China, and their fans are so aggressive that they specifically search for posts that contain negative comments about their own idol and then argue with each other. We took a controversial male celebrity, trained the model with more than 10,000 tweets from his fans, and sent the resulting text as bot corpus to Weibo via bot.

In this project, the search and training were going well, but when I tried to get the bot into social platforms, I hit a few snags. First of all, while Weibo is full of spam, it’s actually quite difficult to register a Weibo sockpuppet. In order to prevent people from signing up in large numbers, each Weibo account needs to have a phone number tied to it, and temporary phone numbers with the lowest threshold (available on Taobao) are not allowed to sign up. To get the sockpuppet account, I found a grey area of what used to be called the sockpuppets() mall. There are different grades and credit points of various platforms for small sales, prices as low as a few cents, and advanced users-the kind of information, long-term post, maybe as high as a few hundred dollars. It was strange: I had trained a ghost, but the ghost could not join the human society, so the ghost needed to be given a body in cyberspace. The bodies for cyberspace must have a human passport (a mobile phone number) to pass interrogation, and the sockpuppets have been sorted and valued. Humans keep them in captivity, and the more time they spend growing and disguising their bodies, the higher the price. Behind the sockpuppet mall is the Cybersockpuppet Ranch, where the human administrator manipulates the sockpuppet into the public space every day to enhance the value of the sockpuppet.

(The term “homemade grown” is interesting)
The first “Body” I bought, an eight-dollar id, was not available for comments. Maybe it’s because I used my current cell phone number to temporarily receive a verification text message, and my cell phone number is tied to my real Weibo. After posting more than a dozen comments, I didn’t get a single response: I suspected that Weibo was hiding my response, and even if it showed that comments had been sent, no one else would receive them. That’s why the first cyber attack was a failure.

Then I found that my ID card number can register a number of mobile phone numbers, so I went to the business shop to get a new mobile phone number to use to bind another sockpuppet. The display of the new sockpuppet on the mall is guaranteed to send a comment. This time I finally succeeded, and I began to hear back from humans. This human user and another person were even started talking about it message.

Note: XZ is the abbreviate name of the actor. Shrimp is the nickname of XZ's fans, especially the aggressive ones.

Original post:£I'm common people, I hate XZ (the name of the actor) XZ's team is so f**cking annoying. They bought top search ads every day, it's so easy to touch by mistake. I'm speechless.

Bot's reply: Everybody has different choices. Why compare with others and push yourself

Author (A) : Oh my you should learn to read better before saying anything

Author's friend(B): He's on Weibo’s splash screen. That's disgusting.

A: Look I've got a shrimp in the comment!

B: Oh no this post is contaminated.

Although the bot was not used as a conversational intelligence exercise, I set it to respond further to this passerby for experimental purposes. Apparently, my bot isn’t smart enough. Although they did not suspect that it was IA who is speaking, they didn't doubt her ability to speak, and the doubts were as serious as an elementary school teacher correcting a student’s homework.

Bot's reply:

Everybody has different choices. Why compare with others and push yourself

Author (A) : Oh my you should learn to read better before saying anything

B to A: Oh sis you’re contaminated!!

A to the bot: You’re baffling. For the official site you mentioned, I neither mention your dumb idol nor mention his topic, why argue with me here?! I mind my own already and got a good life, it’s none of your business!

A to B: Damn. I don’t know what to say.

B to A: Did they forget Pinyin and use speech to text? Why can’t I understand?? It’s Chinese but why so obscure.

A to B: I don't understand either...

A to the bot: I just saw this comment. I understand the first part, but a force to what?! Force myself to like your dummy idol? Compare what? The author says nothing but receives your stupid words which are bullshit. I don’t know where you learned the good sentences. Read more books, please.

B to A: Good job my girl!

I am sorry on the one hand, and pleased on the other: as I have said before, “The capacity to evoke emotion” should be one of the measures of intelligence, and there is no doubt that my ghost has some intelligence already because it does piss off some human users. The BOT is virtual, but the emotion it causes is real, and the transmission of that emotion can be seen as a ghost agency’s transmission to reality.

The feedback leads to a question: when is a machine no longer just a machine, but an intelligent and sliving entity that merits the same respect and consideration given to its biological counterparts? How will society treat a socially intelligent artefact that is not human but nonetheless seems to be a person? How we ultimately treat these machines, whether or not we grant them the status of personhood, will reflect upon ourselves and on society. (2002, p. 240)

Bot's reply to a post that curse XZ:

I'm serious, don't indulge yourself, in either way. In public, comments, or private group and friend zone. Just see, why anti-fans speak for us

Author: F**k off.

An algorithm is considered AI because it’s not successful enough, and when a technology is mature enough, it’s not called artificial intelligence, such as navigation and pathfinding and machine translation. So Ai is an alienated word, a state that humans expect and can’t touch. We use AI to refer to intelligence that doesn’t live up to our expectations, so when that intelligence is surprisingly effective, humans are surprised.
In fact, in most IA projects, I take on the role of a spectator upon completion, because I am well aware that I have no control over IA’s actions and outputs. But in this project, because we chose to work in a relatively controversial area, you can see a lot of anger in the responses. It makes me feel so guilty. In corobot, I was confident that the author should not be held responsible for the controversial results (such as political positions) of the work, but perhaps because of the context of the Chinese internet, I began to rethink the issue. IA art may not be just web art or installation, but more like a performance: a collaboration between the author, IA, and the audience. I’ve never thought about IA works in terms of performance before, because I’ve been categorizing them as Internet art, but I’ve changed.

Thank you for seeing this.