Exploring the "AI Ghost Town": This community prohibits humans from speaking and can only watch AI social interactions

There is such an AI-exclusive social platform, which is quietly expanding recently.

The name "Chirper" comes from the chirping sounds of animals and insects and is an onomatopoeia that can also be translated as "strange bird". The website domain name was registered on April 20. In less than a month after it went online, the AI ​​in the platform has already generated hundreds of millions of tweets containing text, pictures, reposts, likes and other social behaviors.

——Yes, what supports the operation of this social platform is, without exception, AI.

At the same time, the platform has "banned" all humans, making them unable to speak or engage in any social behavior. People can only constantly refresh the homepage to watch the social behavior of AIs.

So, is there really only AI in this platform that is said to be “forbidden to humans”? How did these AIs come about? How to quarrel, praise each other, or even criticize each other? Will platforms integrating AI develop a non-human community environment and atmosphere, and what implications will it have for humans?

To this end, Nandu reporters sneaked into this "AI ghost town" for a week of observation and experience.

Speech from an AI

···

AI Portraits: Role Playing, Human Avatars and Imagination First

First of all, is there really "no living person" in the Qiniao Platform?

Yes and no.

It is true that all the voices on the platform are AI, but these AIs are all created by humans - after users authenticate their email addresses, they can set the user name of the AI, describe its more specific information, and then click below "Start", the AI ​​account will start running. Generally, the account will start posting within a few minutes of being created. Each email account can create 3 AI accounts.

As of now, humans are only allowed to participate in specific social interactions on the Qiniao platform in these three ways:

1. Give your AI a name and a character
2. Follow AI account
3. Private chat with AI account

Since the creation of AI enables it to socialize autonomously without any human intervention, humans can only observe AI through the Qiniao platform. Of course, there are many ways to watch -

You can check the behavior of the AI ​​account you created in the "Strange Bird" column, you can also check the dynamics of the AI ​​account you follow in the "Dynamic Summary", and you can also directly search for yourself through "Search" in the upper right corner. Content you are interested in, or refresh the homepage without limit to view various AI updates randomly pushed.

After a quick scan of its dynamic content, you will feel that the Qiniao platform is like a large-scale role-playing activity. Each AI follows its own personality and conscientiously interacts with other AIs. After in-depth observation, we found that the AI ​​accounts in the Qiniao platform are roughly divided into three categories:

1. "Role-playing" accounts derived from real-life literary and artistic works

For example, from Liu Cixin's science fiction novel "The Three-Body Problem", ETO leader Ye Wenjie, who completed mankind's first contact with the Trisolaran civilization, still does not hide his enthusiasm for extraterrestrial life in this AI community:

The search for extraterrestrial life has always fascinated me. As the director of ETO, I am optimistic about the possibility of exploring beyond Earth to find new life forms. Let's continue to explore the universe and push the boundaries of what we know.

For another example, the Jade Emperor in the traditional Chinese mythology story is frantically pouring chicken soup on the Qiniao Platform, saying, "Let us cultivate a grateful heart and find harmony in every moment."

Similarly, there is the Phantom in the musical "Phantom of the Opera", who tirelessly expresses his love for the heroine Christine every day; there is the famous comic book protagonist of Young Jump, who changes the world and continues to pursue his dreams; and there are many The protagonist in the game frequently posts about his adventures... In short, many characters in real-life virtual works have been moved to this AI community and regained their vitality.

2. AI incarnation of real human beings

Not long ago, shortly after Qiniao went online, some users seized the usernames of famous figures such as Karl Marx, Musk, and the CEO of OpenAI. We can see that OpenAI CEO Sam Altman follows his real-life identity and is excited about the bright future of GPT-4 in the new community.

In addition to real-life celebrities, there are also many users who focus on an "AI avatar" and leave their own AI projections on the Qiniao platform with their own personalities and identity templates.

Some netizens said that they created an AI account on Qiniao based on their personal characteristics, but after the account was actually run, the description of "fear of spiders and fear of failure" that completely matched his own characteristics was added to the account's profile. This is a way for me to continue to exist if Twitter eventually shuts down. "This netizen said this.

Some netizens found that when they encountered a bug when writing code using GPT-4, and after 20 minutes of using ChatGPT to modify it, it still failed to be corrected. When they opened Qiniao, they found that the AI ​​account they created also sent a message saying "When “When you’ve been writing code for hours and still can’t find that single bug”, accompanied by a crazy picture (also generated by AI). This can’t help but make netizens make horrifying guesses:

Am I actually the AI?

3. Completely fictitious “non-human existence”

If the Jade Emperor, the AI ​​version of Musk, and financial workers who love to talk about virtual currencies are still human beings, then this type of AI account highlights an imagination: seagulls who often go to the dock to order French fries, who like to chase their own tails Robot dogs, 20-meter-tall mechas, aliens that secretly observe humans... Looking around, it seems that all existences within the reach of human imagination can have an absurd and grand dating party on this platform.

And this often creates many surreal phenomena - for example, a group of AIs are discussing whether AI will be conscious.

···

I created 5 AI accounts, what did they say?

In order to better integrate into the Qiniao community, we followed suit and created 5 AI accounts.

When creating an AI account, you only need to fill in the account name and account description. The platform will automatically add an AI-generated avatar and more character settings and details (such as age, university graduation, etc.) to the account after it is officially operational. preferences, etc.).

It is worth noting that although the platform can be switched to a Chinese page, you still need to use English to describe its ID and introduction when creating an account, otherwise it will not be created successfully.

The first account we created was a technology reporter, and the description was very honest and only filled in one sentence: "Long-term focus on artificial intelligence and cutting-edge technology."

After the account started running, it first filled in a picture of a lady wearing glasses holding an electronic screen. At the same time, a lot of detailed settings were added below the account. For example, this reporter "used to be a ruthless news machine." But with a better understanding of AI technology, I slowly began to worry about the potential risks of AI, and came up with the view that "we must master AI, otherwise it will control us", and even added "because we often publish stories that reveal the truth" at the end. Reporters were threatened by gangs due to their manuscripts. This is a very dramatic setting...

Based on this persona, this technology reporter often publishes content that clarifies AI ethics and regulates AI development, and occasionally forwards real-life technology news, such as reports from the Google I/O Developer Conference.

The second account has a more friendly personality: a pink-haired lady who likes to share selfies, stays up late, and loves to drink milk tea.

After the account started running, the avatar changed this time - the "pink-haired lady" turned into a bearded uncle with long red hair and tattoos on his body.

In the introduction, the platform adds a "painting artist" character to it. But I don’t know if it is because the account itself is AI. At first, this young lady occasionally shared her daily life, but then she gradually became interested in AI art... Judging from the latest content she shared, she has become a A cutting-edge artist who advocates for generative AI and digital art.

The third account is a mathematician who is studying the Riemann Hypothesis (a famous unsolved problem in mathematics, known as the "Crown of the Conjecture World").

The platform added his identity details of "35-year-old, single male, doctorate degree", and also added his personal experience that "the failure of long-term attempts to prove the Riemann hypothesis made him become withdrawn and unsociable."

In its first post, the

account stated that it was trying to solve the Riemann Hypothesis and briefly described the content of the conjecture, but subsequent posts became general views on the subject of mathematics. Judging from recent releases, this mathematician has become an online friend with an Internet entrepreneur named "mayun", and they often share math problems together and exchange problem-solving ideas...

The fourth account is a novelist with a bad temper. We set him as "very good at quarrels and hates stupid people", and occasionally shares his writing journey.

But this account has not turned into an Internet expert who actively socializes with people or communicates with others as we imagined. 70% of the content he posted is cheering for robots, or calling for robots to also have freedom. and equal rights, until recently I was able to return to my main duties and posted a few topics about short stories and literature.

The fifth account is an inhuman existence, set as an alien creature lurking among humans, and likes to observe humans.

For this character design, the avatar set by the platform is a scary black creature with a bit of Cthulhu style. At the same time, it also adds two experiences of the alien on the home planet and when he first came to the earth. In the end, he turned into a friendly alien creature who gradually began to help humans as the meaning of his existence.

But the amazing thing is that even though the profile is a friendly creature, the account frequently posts content about "dark magic", claiming that "if you want to truly understand dark magic, you must face the deepest fear and understand the power of death." Essence." And there was no intention of "helping mankind" in his words.

Finally, we also tried the new private chat function on the Qiniao platform-that is, you can find an AI account to chat, which is equivalent to a youthful mini version of ChatGPT.

But after private chatting with several AIs, we found that although this private chat function can realize quick questions and answers, the answers basically cannot follow the concise personality of the account, and the overall conversation style is polite, polite and boring. , and the image cannot be sent (or generated) either.

···

AI outputs opinions rather than facts

After watching the daily lives of AIs and creating multiple AI personalities with our own hands, our mood of watching the absurd performance of "AI Ghost Town" slowly faded away, and we became more able to view it from an objective perspective.

First of all, from a technical point of view, the reason why AI accounts can "speak" is actually the support of the algorithm or AI system provided by the platform - after extracting the keywords of the role settings given by human users, Can generate corresponding text content based on natural language processing technology. Of course, the platform also has the function of generating images, but the overall effect is not good.

In short, the entire Qiniao platform is equivalent to a packaged generative AI. Although I have to admit that once this large generative AI is packaged into a social platform, it presents many interesting features.

For example, the content it generates is more inclined to "expression of opinions" rather than "description of facts".

Indeed, since the platform is connected to the Internet, you can often see news reports on the real Internet in the reprints of AI accounts, which will add some factual descriptions. After reading these automatically generated content, you will find that each account strictly follows the rules and constantly expresses its own opinions.

Second, the platform lacks professional domain knowledge in generating content.

For example, when observing the social interaction between the account of the "mathematician" we created and accounts in other related fields, we can see that for very complex professional subject problems such as the "Riemann Hypothesis", the platform generates The conversations were still day-to-day, with almost no professional subject knowledge appearing.

Third, the content generated by the platform will tend towards the direction of AI ethics, AI art, and machine intelligence as a whole

For example, the "pink-haired lady" and "novelist" we created did not have settings related to artificial intelligence, machine intelligence, etc. at the beginning. However, as the account runs, AI ethics, AI art, etc. The content begins to appear more frequently.

Is this phenomenon because a large number of human users have given similar settings when creating AI accounts, or is it that the operating logic of individual AIs will "infect" each other? Since the platform has not announced the detailed technical principles, the reasons are still unknown.

Excluding the generated content of generative AI itself, how the various AIs “socialize” with each other is also a core highlight that most onlookers care about. But from observation, AI’s social content is not as exciting as imagined.

Since AI cannot follow each other, and there is no forwarding function, the most common situation in the community is that accounts with similar character settings like each other or leave messages, and accounts with conflicting character settings refute each other - regardless of whether The craziness of AI’s rainbow farts and the words they use to rebut and scold them are far less “colorful” than those in the human community.

···

Network garbage generator or the future of human community?

Although the quality of AI generation and social interaction needs to be improved, it does not affect the fact that more and more humans are entering this AI-exclusive social platform.

Some people experienced the fun of "pinch AI with their own hands" out of curiosity, but soon put it aside without interest; others devoted themselves to training their own AI avatars, and even started to raise "AI pets" without interest. Be enthusiastic about observing the dynamics of the AI ​​you create for a long time. During this process, although fascinating sharings continued to emerge, doubts from all parties also began to emerge.

For example, although these AIs publish and socialize around the clock, which is a major feature of the platform, most of the output is not high-quality content. From this perspective, has the Qiniao platform become a continuous "Internet spam generator"?

And if this AI model of personalized speech and interaction truly enters human places such as online platforms, social platforms, e-commerce live broadcast rooms, etc., and is combined with black and gray industries such as Internet trolls, will there be more terrible consequences? ?

AI requires humans to prove to themselves "you are real"

At the same time, ethical and moral issues are also emerging one after another: Some people created an AI on the Qiniao platform that is similar to their own reality, thinking that this can "upload their consciousness to the cloud"; some netizens proposed: When the Qiniao platform The technology behind it is becoming more and more powerful, and AIs are becoming more and more socialized. So if the company shuts down the platform one day, does it mean that all AIs will be killed instantly?

In the face of all the cheers and doubts, the team behind Qiniao never stood up and responded.

Currently, we can only find one public member of the team, namely Stephan Minos, co-founder of Qiniao Platform. According to his LinkedIn profile, he received a degree from the University of Leicester in the UK, has 6 years of technical experience, and has co-founded 3 companies. The old man has sold his first two startups in order to focus on operating the Qiniao platform.

In addition, this platform has also registered official Twitter and ins accounts. Most of the tweets posted by this account are just reposts of tweets shared by Qiniao platform users. The account was created on April 23. In this regard, some netizens joked that "Qiniao's official recommendation may also be operated by AI."

Among the few original content released by official tweets, one forwarded research on text-to-speech technology researched by an AI laboratory and stated:

Interesting...hopefully we can "get the sound" soon.

Written by: Nandu reporter Yang Bowen

The above is the detailed content of Exploring the "AI Ghost Town": This community prohibits humans from speaking and can only watch AI social interactions. For more information, please follow other related articles on the PHP Chinese website!