Can Baidu Netdisk Library become the “Swiss Army Knife” of the AI ​​era?

1. What sparks will “camera + AI” create?

Before we talk about how to modify the camera on today's mobile phone, let's review a little history of camera development:

In June 1913, German engineer Oskar Barnack designed a lightweight camera that creatively used 35mm movie film as a negative.

Before that, cameras used huge 8*13 cm glass film. Cameras using this kind of film were huge in size and needed to be placed in a bulky wooden box and fixed on a large tripod.

So, before Barnack's invention, photography was an indoor activity, where people took stiff portraits in photo studios, and Barnack's invention made the camera convenient for the first time.

For the first time, humans could freely take vivid, real pictures outdoors, and this camera was named "Leica".

Yes, in 1913, Leica changed the way people use cameras. Based on the pragmatic product philosophy, it truly increased the frequency of camera use.

Surprisingly, his approach was not to improve the quality of the camera's photos, but to completely revolutionize the camera's usage scenarios.

Today, AI has given cameras a "Leica"-level opportunity, allowing cameras that act as "eyes" to truly connect to the "brain."

This is what Baidu wants to do by releasing the "AI Camera" on Wenku and Netdisk at this AI Day - to create a full-modal super portal that integrates storage, search, photo editing, scanning, translation, and creation, comprehensively covering the capabilities of storage, management, use, creation, and sharing.

2. What exactly is an “AI camera”?

What does “a full-modal entrance covering storage, management, use, creation and sharing” mean?

In other words, compared with the camera that comes with a mobile phone, what is special about Baidu's "AI camera"?

A common operation for most people when using a camera is to first take a photo with the phone's native camera, then use a photo editing app to edit it, use a scanning app to recognize it, use a translation app to translate it, and then sync it to a cloud disk for storage management.

This experience is fragmented, decentralized, costly, and frictional.

The "AI Camera" is designed to solve this problem - using a one-stop, All in One approach.

Intuitively speaking, taking pictures is just the starting point, and the series of AI actions after taking pictures are the core - a series of AI functions such as photo retouching and beautification, photo recognition, problem solving, text extraction, translation, scanning, contract checking, etc., which can solve 99% of users' needs for photo storage, shooting, processing, and management in one stop.

Whether it is scenic spot explanation, plant identification, and product identification in life scenes, document scanning and ticket archiving in work scenes, or photo solving and wrong question explanation in learning scenes, AI cameras can solve all problems in one go.

More importantly, it can also “shoot and save”, with all shots stored directly in the cloud, freeing up a large amount of mobile phone memory, allowing users to completely say goodbye to memory anxiety and truly help users achieve “shoot, store, and manage” in one .

Its long-term management functions, such as sharing, smart classification, and smart image search, further increase its practicality - multi-dimensional intelligent classification by location, people, things, photo type, data type, etc.

Even many celebrity studios and celebrity fan clubs use the relevant functions of Baidu Netdisk to stimulate interaction with fans.

Design of AI Camera All in One

3. Why is Baidu Wenku Cloud Disk said to be creating the Swiss Army Knife of the AI ​​era?

In fact, AI camera is a microcosm of Wenku Network Disk's full-modal input and processed output at the input level.

In my opinion, what Baidu Wenku Cloud Disk is doing now is to create the Swiss Army Knife of the AI ​​era.

What are the essential characteristics of a Swiss Army Knife?

Because it embodies a fundamental principle of technology design: improving practicality through rational integration.

In an era of tool scarcity, we pursue the ultimate performance of each tool, but in an era of tool surplus, we pursue the integration, portability and practicality of tools - users do not want to jump between many different AI tools.

Baidu Wenku and Cloud Disk hope to achieve the following: when you need AI help, it is always there, and you don’t need to think about which tool to use.

Their goal is to make AI no longer a scattered tool, but a "super productivity" that can solve problems in one stop.

Just like more than a hundred years ago, what a soldier needed was not the sharpest knife, the sturdiest screwdriver and the most precise pair of scissors, but a Swiss Army knife that could solve all problems at any time.

Specifically, Baidu has achieved complete delivery of full-modal input, processing and output through the method of "marking dots, connecting lines and building networks".

How to understand?

Let’s look at them one by one –

Let’s first look at omnimodal input – “it can handle anything”.

Behind the input is demand, and omnimodal input can lock in user needs to the greatest extent - accepting and responding to user needs and launching tasks around the clock and in all directions, supporting omnimodal input methods such as keyboards, AI microphones, AI cameras, AI videos, etc. for various scenarios, regions, clients, and devices.

For example, many AI products do not support users entering URLs. That is, when a user throws a link starting with http into the product, it only knows that it is a string of characters and has no way to parse the content behind it, which will severely limit the user's input bandwidth.

After all, any demand of a user does not arise out of thin air, but has a starting point. This starting point is most likely based on some existing material: a video, an article, a podcast, and then AI demand is derived. Omnimodal demand can meet this demand to the greatest extent.

In order to handle input of every mode, it actually relies on a very systematic infrastructure. On the one hand, handling these demands consumes a large number of tokens, and on the other hand, it also relies on the optimization of the product at the detail level.

Let’s look at holomodal processing—“a complete and robust nervous system.”

The core of the full-modal processing here is actually implemented based on Cangzhou OS. The core of Cangzhou OS is three modules - first, an intelligent and efficient scheduling system, second, massive public and private domain content data, and finally, comprehensive AI capabilities.

An intuitive metaphor is that Cangzhou OS is a complete "nervous system". It does not create directly, but it perceives, coordinates and schedules.

It connects various functional modules - hundreds of AI agents, some are "eyes" (AI cameras, responsible for visual input), some are "brains" (large models, responsible for thinking and planning), some are "hands" ( PPT agents, document agents ), and some are huge "memory" (databases of document libraries and network disks), and finally form a joint force to respond to different omnimodal needs.

Finally, let’s look at omnimodal delivery – “Reject semi-finished products: open up the last mile of delivery”.

"What users want is not a drill, but a hole in the wall."

That’s right, the same goes for AI products. What users want is not the delivery of a point, but the delivery of a complete product; what they want is not ingredients, but a dish.

Taking the new feature “GenFlow Super Partner” as an example, you only need to enter the following requirements:

"I recently saw a lot of stores selling anime and anime peripherals while shopping in the mall. Is this very popular now? I also studied design. Can I start a business with this? Please help me write a market research report."

A few minutes later, a detailed research report of 27 pages, 23,000 words and 8 charts was completed. This report also supports viewing in code mode. This is the importance of full-modal delivery.

Yes, most users usually need articles with both text and pictures, PPTs with professional charts, videos with sound, and a combination of these contents.

However, for most AI products, users only receive raw materials or semi-finished products, and a lot of follow-up work needs to be done.

It's like giving someone a bunch of Lego bricks and saying, "Look, you can build anything with them!"

But what the user really wants to say is: "I don't want to build it, I want the castle that's already built."

That's right, through dot-by-dot, connection and networking, Baidu Wenku Cloud Disk organically integrates the tool infrastructure into a whole, and users only need to enjoy the one-stop service conveniently.

4. The underestimated Baidu Library cloud disk

Many people working at the forefront of AI have an intentional or unintentional tendency to pursue novelty, believing that newly born AI products are naturally more AI Native, while products transformed from traditional tools to AI are not that sexy.

I do not support this argument. In my opinion, the vote cast by users with their feet is the core of judging an AI product.

In this sense, I would like to share two points that I think Baidu Library and Baidu Cloud do very well:

1. Make good use of the inertia of trust

Indeed, in the pre-AI era, a large number of users had already handed over their very valuable digital assets—family photos, work documents, and learning materials—to Baidu Wenku and Baidu Netdisk.

For example, Baidu Wenku has 1.4 billion professional documents and 100 billion GB of storage capacity - behind these numbers are real user needs and usage scenarios.

This long-term relationship based on trust is the starting point for Baidu Netdisk and Baidu Wenku to directly use AI to reconstruct existing products - not starting from scratch, but continuing the inertia of trust.

The cleverness of Baidu lies in that they did not require users to learn completely new tools. Instead, they seamlessly integrated AI functions into the document library and network disk that users are already familiar with to solve practical problems. This greatly reduced the user's psychological threshold and the difficulty of cold start.

2. Unique data is an important moat

That’s right, OpenAI’s Sam Altman has also repeatedly emphasized the importance of “Memory” to ChatGPT recently.

In his opinion, this is the key point to keep users in the ChatGPT product for a long time. Google also emphasized the importance of data such as search history in the Gemini product.

The public domain knowledge base of the library, the private domain database of the network disk and the user memory base constitute more three-dimensional data. Its deeper significance lies in that it can provide personalized context for AI, which is its unique and non-replicable asset.

As Baidu Wenku Cloud Disk continues to deepen its research in the fields of AI content creation, content consumption, personal knowledge base, etc., and explores the big model industry from deep thinking to "deep delivery", the three-dimensional data upgraded through massive interactions with users will become a deep moat for the product.

Indeed, the data from users voting with their feet is the most honest. Currently, Baidu Wenku AI has 97 million monthly active users.

Baidu Netdisk's AI has over 80 million monthly active users. In the AI ​​product rankings on June 3, Baidu Netdisk APP MAU exceeded 150 million, ranking first in the domestic application list and second in the global list, second only to ChatGPT.

Conclusion

Baidu Wenku Cloud Disk's " full-modal input, processing, and output system " is actually a promise, and this promise is -

"Tell me your idea and leave the rest to me. I will use everything I know (the library) and everything you have (the online disk) to get it done."

For knowledge workers, this is a very attractive promise that means less friction, more certainty in creation, and the promise of reshaping “super productivity” through complete delivery.