Where has the AIGC dream at station B come to?

Source: "Silicon-based Research Laboratory" (ID: gh_cef05ac13977), author: Hickory

Image source: Generated by Unbounded AI

“Although Station B did not participate in the Hundred Models Competition, our large models have been actually applied to specific businesses.”

In the second-quarter earnings conference call of Bilibili (NASDAQ: BILI, HKEX: 9626; hereinafter referred to as Bilibili), the management of Bilibili said so.

While many major manufacturers are using AI to tell the dream of AIGC, the attitude of station B seems a bit "ambiguous".

Compared with the situation of players' "big refinement model", Station B did not reveal much about the layout in the AIGC field. At the beginning of July this year, Bilibili launched the "Search AI Assistant" and conducted a grayscale test on a small scale. At the end of that month, Bilibili announced the launch of a large model of the bilibili index and provided technical support for the "Search AI Assistant". How much spray.

In subsequent public occasions, Station B’s discussion of AIGC mostly focused on “lowering the threshold for content creation” and “the highest quality and richest AI content on the entire network is at Station B”.

As a content community, Station B was previously controversial due to its commercialization path of "difficult to break the circle", but since 2022, from organizational design to the improvement of a series of commercial infrastructure, Station B has gradually let go of its obsession and generously side to make money. In the financial report conference call in March this year, Chen Rui, CEO of Station B, said that improving the efficiency of commercialization is the most important task for the company right now.

On August 17, Station B released its financial report for the second quarter of 2023, with revenue of 5.3 billion yuan, which was lower than the market expectation of 5.273 billion yuan, an increase of 8% year-on-year. In terms of profit, losses are also narrowing, with an adjusted net loss of 964 million yuan, a year-on-year decrease of 51%.

In addition to financial results, under this wave of AIGC, the content, advertising and e-commerce business of Bilibili is undoubtedly one of the areas where AI is most likely to be implemented and the most effective. **So, what we are curious about is, what are the possible actions of Station B? What are the challenges? **

The bonus of station B: video search and optimization supply

As mentioned above, the relevant layout of the AIGC field of station B is rare in public information. There are two main actions. One is that at the beginning of July this year, Bilibili launched the "Search AI Assistant" and conducted a small-scale internal test. At present, the test scope has been expanded. The other is the bilibili index large model announced by Station B at the end of the month.

In addition to specific products, the senior management of Bilibili mainly attributed AIGC's empowerment to Bilibili into three aspects. The first is the optimization of search experience, helping users to extract key information and saving decision-making costs. The second is to improve the productivity of creators and use AI tools to improve creative efficiency. The third is to optimize the image of a large number of virtual anchors at Station B. **

Apart from these "correct nonsense", what bonus does **AIGC have for Station B? In fact, it is also in two aspects of commercialization and community ecology. **

**The first is to step on the dividends of video search and improve the performance of advertising business. **

Different from pure text search, video search can place advertisements more "softly" and "natively" to attract users to click. The reason is that video search itself has its own purpose, and at the same time, it is also knowledge-oriented in terms of content. At the same time, unlike short videos, the content of medium and long videos is also easier to accurately capture user demands, extract more keywords, and then match the needs of advertisers.

In fact, YouTube has launched video search ads, and the conversion effect is also good. This quarter, station B’s revenue from advertising reached 1.6 billion yuan, an increase of 36% compared to the same period in 2022. In the past year, many organizational design and commercialization actions of station B have been carried out around improving the advertising business. Therefore, if more attention is paid to video search, its advertising business may usher in new dividends.

**The second is to optimize the supply and drive more UP owners to make money. The UP master is the core resource of Bilibili. It is not only the main supplier of community content, but also the main force of Bilibili’s commercial exploration (live streaming, advertising). ** Station B announced a major change in its 14th anniversary celebration, replacing the current playback volume with the number of minutes the user actually watches the video. The implementation of AIGC technology has two advantages for UP owners:

**One is that under the user search logic, high-quality content will be easily seen by users, and then seen by more "financial fathers". **

Previously, the relevant person in charge of YouTube mentioned: "There are billions of video views generated by keyword searches every week, and many videos with a more practical style are discovered in this way."

**Secondly, if Station B can provide UP owners with more out-of-the-box creation tools, it will improve UP production efficiency and encourage more creators to enter Station B. **

Difficulties of station B: experience, cost and content supervision

But dividends are part of it. How to make good use of AI at station B to form a synergy between content and commercialization is still facing a lot of challenges.

**First of all, it is a problem on the experience side. **From the user's search behavior, search means precision.

"Silicon Research Lab" actually tested the "Search AI Assistant" at Station B, and found that the current functions are mainly focused on simple question answers and video recommendations. Previously, according to the evaluation of "Guoke" and "Ray Technology", the little assistant still has the phenomenon of "talking nonsense", which shows that the current ability is still limited.

The current function is still relatively simple, mainly focusing on simple question answering and video recommendation Source: author

In addition, in May of this year, Bilibili launched the AI video assistant "Xiaoben". After three months of iterations, in addition to the single video content summary function, there are currently high-energy moments, comment summary and other functions. At the same time, on the summon assistant, there are currently three forms of summoning in the comment area, sharing summons in private messages, and sending bv numbers in private messages. **

However, perhaps limited by computing power and technical reasons, the video summary assistant also has problems such as low feedback response. Regarding the summary function of "Little Stupid", in a user survey of more than 10,000 users released on August 10, more than half of the users think that it needs to be optimized.

Image source: Screenshot of station B

**Experience is linked to user quality, Bilibili must find a better way of interaction. **The financial report of station B shows that the average daily usage time of station B was 96 minutes in the last quarter, but this quarter dropped to 94 minutes, and the average monthly active number was the same as the previous quarter, at 315 million.

** Followed by cost difficulties. ** Judging from Google's action on YouTube, its way of integrating AI large language models into the content community can be described as "strengthening miracles". But for Station B, which is currently burdened with profitability KPI and is trying to learn how to save money, this will be a considerable investment.

** More critical is the pressure of content regulation. **This comes after a new study by NewsGuard revealed that dozens of websites around the world are using artificial intelligence to generate low-quality "clickbait" articles in order to make money from advertising. The previous "AI Stefanie Sun" and "AI Face Changing" also faced copyright and ethical disputes. In the conference call, the management of Station B mentioned that the current large language model has been applied to content review, which has greatly improved the efficiency of review.

There is no doubt that under the wave of AIGC, the video content platform undoubtedly ushered in a good era. The management of Station B stated in the conference call that AIGC is highly valued internally, which is a major opportunity for a content platform like Station B. "It's like the mobile Internet ten years ago."

But for Station B, which is facing multiple pressures at the moment, it may be more critical to take the next step under the opportunity.

References:

1. Outside and inside: Bilibili’s ambition, Baidu’s crisis