Someone used AI to create a Twitch channel, which gained 30,000 followers every day

ChatGPT, which became popular at the end of 2022, has attracted the keen attention of major companies. On the other hand, people have also seen the possibility of extending AI technology. Now, someone has used AI to create a Twitch channel. How to understand and observe this attempt? Let's take a look at the author's interpretation.

After ChatGPT became popular, big companies entered into a competition, with Google and Baidu catching up at full speed; on the other hand, developers applied AI technology in a variety of ways. Recently, the editorial department saw a lot of them, but most of them were pictures, until we found a Twitch channel with a lot of ideas.

1. The AI ​​version of Seinfeld is broadcast live on Twitch all year round, with more than 30,000 followers in a single day

At around 10pm on February 1, Beijing time, Skyler Hartle, one of the founders of Nothing, Forever, posted on Twitter to celebrate the fact that Nothing, Forever's Twitch channel WatchMeForever had more than 7,000 followers. When I checked on the afternoon of February 2, the channel had already reached 37,000 followers. By the afternoon of February 6, the number of followers had exceeded 180,000. Overall, the average daily increase in followers is 30,000, and this is Twitch.

Skyler Hartle celebrated on Twitter

Nothing,Forever’s number of followers exceeds 7,000 | Source: Twitter

Nothing,Forever channel follower count changes | Source: Twitch

Nothing, Forever is an AI-generated sitcom that is live-streamed on Twitch. The image quality gives people the feeling of a PC game from the 1990s, with a high saturation background. The characters' movements are not smooth, and the voices can be clearly heard to be machine-generated. The canned laughter, music, and vocal tracks added to the plot also have a relatively obvious sense of disconnection. But when you think that this is all AI-generated content, you still feel it is amazing.

In terms of plot, Nothing, Forever records the daily life of 3 men and 1 woman. There are only 2 scenes in the whole play. The protagonist, who is a professional talk show host, stands on the stage to tell jokes, and the protagonists chat in the apartment. Each scene lasts for 2 minutes at most. There is no continuity between the plots. There is only a picture of New York street buildings as a link, and there is no connection between each episode.

Many people will find that Nothing, Forever imitates the popular 1990s sitcom "Seinfeld" after watching it for a while. Whether it is the protagonist's profession, animated image or nonsensical and unrelated clips, there are shadows of "Seinfeld".

However, the founder is very frank about this, and even the logo of Nothing, Forever deliberately imitates "Seinfeld" to emphasize the connection between the two. However, the biggest advantage of AI is its strong learning ability and fast learning speed. Nothing, Forever may soon no longer be a copy of "Seinfeld".

Seinfeld vs. Nothing, Forever logo

2. Nothing, Forever is both AI and UGC

Skyler Hartle introduced in a Reddit post that Nothing, Forever uses OpenAI's GPT-3 language prediction model and monitors real-time chats in the Twitch comment section. User discussions may change the direction of the series. For example, if everyone wants to delete a character, Nothing, Forever can do it while maintaining the live broadcast. In other words, with the joint creation of the audience, Nothing, Forever's script inspiration will never run out.

This is another feature of Nothing, Forever. Since its launch on December 14 last year, Nothing, Forever has been on rotation on Twitch. During the several time periods that I watched, the number of people online in Nothing, Forever's live broadcast room basically remained at 7.5K-9K, and users were very active in the Twitch comment area and Nothing, Forever's Discord group.

I joined the Discord group on February 2, when the group members were mainly discussing the connection between Nothing, Forever and Seinfeld, and a lot of emoticons and memes were generated (who would have thought that the first thing that became popular in a show was Larry's microwave oven). In just a few days, many new topics were added to the group, including discussions on the future plot direction, fans uploading homemade materials, and discussing AI, etc. It can be regarded as an active community with strong user spontaneity.

Nothing,Forever Discord Group

Nothing, Forever’s rapid rise to fame is astonishing, and its monetization speed is also not lagging behind, which shows that the founding team has expectations for growth and has planned a business path.

In addition to Twitch donations, Nothing, Forever can also monetize through Patreon and selling peripherals. Nothing, Forever has quickly launched its own peripheral website and launched mugs, sweatshirts, stickers and other products based on the current popular logos and characters.

Source: Nothing, Forever surrounding website

Seeing this, you will find that the founding team has planned a relatively mature growth, fermentation and business model. After AI generates content, it ferments and becomes active through the Discord group, allowing everyone to have a sense of belonging to the content, and then directs users from Discord to the independent site for monetization. This is the monetization method adopted by many "IPs" . For example, Lofi Girl, a very large companion channel on Discord, has a special Merch tag to sell peripherals. From the interaction below, we can see that Discord users are very receptive to this.

Lofi Girl’s Discord group has a tag dedicated to peripherals

This also makes me wonder, how long can the popularity of "poor quality" sitcoms generated by AI last? Is the founding team trying to do a long-term business or a one-time deal?

3. After 4 years of preparation, is Nothing, Forever’s goal Netflix?

In response to the two questions raised by the author above, the founder’s answer seems to be: Although we have figured out how to do it and even how to monetize it, this is more of an experiment.

Source: Twitter

Skyler Hartle and Brian Habersberger are the two founders of Nothing, Forever. When Nothing, Forever was about to be launched, Hartle posted on Twitter that he had been preparing for the launch of Nothing, Forever for nearly 4 years. In other words, the two had been working on this project as early as 2018, and Nothing, Forever had gone through many iterations in the past 4 years.

In a recent interview with foreign media, Hartle mentioned that the team’s next goal is to make Nothing, Forever a program that continuously produces interesting content, and hopes to launch a drama of comparable quality to Netflix in the future.

Despite the huge investment, Nothing, Forever is not their main business . Skyler Hartle has worked at Microsoft for more than 4 years and is currently a senior product manager. Brian Habersberger is a polymer physicist. According to LinkedIn information, Habersberger has been working at Dow Chemical since graduation and has been promoted to a senior researcher.

Source: LinkedIn

Therefore, Nothing, Forever is more like an experiment in which the two creators put a lot of effort.

4. At the application level, is AIGC a one-wave trend?

In fact, Nothing, Forever is essentially a drama, but with the help of AI, it has added a "user development color". This also allows users to have a sense of belonging to the content even without exciting plots and high-level pictures. This is also the "confidence" for Nothing, Forever to adopt the IP monetization model.

But whether it will last long depends on whether this sense of belonging is sustainable or just a temporary "stimulation" given by AI to users. Take AI drawing as an example. In the past two years, many kinds of gameplay have become popular, but whether from the overall or individual point of view, most of them are one-off.

November 2022 - January 2023, 15 head AI images

App downloads and IAP revenue fluctuations | Source: apptopia

Beyond this time period, there are two types of AI drawing products that have been popular throughout 2022. One is text-to-photo, which generates images by inputting text, and representative products include dream by WOMBO, Wonder, etc. The other is to upload images to generate derivative images, and representative products include Lensa.ai and Dawn AI, etc.

According to Diandian data, the downloads of these four apps increased significantly from November to December last year, and then fell back quickly in less than half a month. The change in downloads here is most likely caused by the popularity of AI. Among them, after Wonder and Lensa.ai fell back, the global average daily downloads were about 40,000, while the global average daily downloads of dream by WOMBO and Dawn AI were only about 10,000. According to Diandian data, the app revenues of Wonder and lensa.ai are also relatively higher among the four apps.

Lensa.ai's global revenue changes in the past six months | Source: Diandian Data

Wonder's global revenue changes in the past six months, although the fluctuations are not obvious,

But the income level is much lower than Lensa.ai | Source: Diandian Data

In my opinion, the richer functions of Wonder and Lensa.ai are probably the reason why the two apps still maintain relatively high download volumes after the popularity of AI drawing has faded. Although Wonder has more functions, AI drawing itself is not a rigid demand of users, so its revenue is difficult to be on the same level as Lensa.ai, which added photo editing and video editing functions after Avatar became popular. This also directly shows that AI drawing gives users a temporary "stimulation", but the supply does not correspond to the rigid demand of users.

Features of 4 AI drawing apps

Before AI mapping models such as stable diffusion and DALL-E were open sourced, most users had never played with text/image mapping. So when AI mapping became popular, a large number of users downloaded popular products to find out more. Apps with good mapping effects and proper publicity and delivery may get higher downloads. However, when the user's novelty wears off, it is not the AI ​​mapping gameplay that can retain users, but whether the functions provided by the App really meet user needs. Just like the rise of TikTok, the way of browsing short videos by swiping up is the reason why users are attracted to TikTok, but what really makes users stay is the short, addictive entertainment content.

The same is true for Nothing, Forever and the AI ​​mapping app. Although both are products that emerged with the help of open source AI models, Nothing, Forever is essentially a drama. Whether the popularity of Nothing, Forever can be sustained may depend on whether it can really attract people in terms of plot, character style, and setting. However, there is indeed a difference here, that is, whether users will have a "nurturing" emotion for such content. Just like the AI ​​Internet celebrity Lia, who has 160,000 followers on Instagram and 200,000 followers on Twitter. Judging from the comments, people did not regard her as an Internet celebrity, but rather gave opinions, paid attention to iterations, and paid attention to her like "raising a daughter."

This is also where the imagination space of Nothing, Forever lies. Although the plot content is still very rough and the jokes uttered by the characters are sometimes repetitive, it cannot resist the audience's investment of time and emotion.

But after many twists and turns, on February 7, the Nothing, Forever experiment did not end because of the decline in popularity, but was temporarily stopped by Twitch's community rules. Nothing, Forever was removed from the platform due to violation of community guidelines, on the grounds that the protagonist Larry's talk show jokes were suspected of discriminating against transgender people. In fact, if you look closely at this joke, you will know that the joke is not about transgender people, but AI is not smart enough and seems to have not learned to avoid group discrimination, so it fell into the trap. The founder said in an interview that he would solve the problem and resume the broadcast of Nothing, Forever as soon as possible.

Unfortunately, we can’t continue to observe for the time being, but after Nothing, Forever has paved the way, entrepreneurs will continue to try.

Author: Zhao Siyao; Editor: Yin Guanxiao; Produced by: White Whale Overseas Editorial Department

Source public account: Baijing Overseas (ID: baijingAPP), a pan-Internet overseas service platform.