DingTalk is becoming more and more "unlike" DingTalk.
This "office software", which is in the eyes of many people, is becoming the most ruthless super application for AI transformation. On January 9, at the press conference held by DingTalk in Hangzhou, President Ye Jun released the latest version 7.5, with a clear focus, one is one: AI assistant.
Loading...
AI assistant is actually one of the most important concepts in the technological innovation brought by large models - AI Agent. After the mega model was completely ignited by ChatGPT, DingTalk's self-AI transformation actions were one after another. In April last year, DingTalk announced that all its products would be reshaped with large models, and then it took only more than 100 days to transform 17 product lines, and by August 22, DingTalk will open the intelligent base (AI PaaS) to ecological partners and customers, and launch an innovative product based on AI PaaS "Digital Employee" to continue the transformation to the ecological layer, and in November, DingTalk AI was launched, and various AI functions were fully open to users for testing.
All of this seemed to be a temptation on January 9, and Ye Jun finally put DingTalk's AI ambitions on the AI Agent platform.
"AI Agent is the best AI application entrance, and DingTalk should become a low-threshold, high-frequency, and open AI intelligent assistant platform. Ye Jun said at the press conference. "DingTalk will build an AI intelligent assistant platform and application market, so that users, developers, and ISVs can develop personalized AI assistants on DingTalk with a low threshold, so that more excellent AI applications can emerge on DingTalk. Let's be the birthplace of the next Midjourney, the next Pika. ”
DingTalk 7.5 has a major update for this.
According to DingTalk, the DingTalk AI assistant is built on a large language model and will have the ability to perceive, remember, plan and act. Moreover, DingTalk said that this capability is not only internal to DingTalk, it also has cross-application execution capabilities, and you can create an AI Agent that "shuttles freely" between DingTalk's internal, third-party and enterprise-built applications according to your own ideas and needs.
According to the press conference, DingTalk currently presets high-frequency official AI capabilities such as intelligent Q&A, image generation, content creation, and data statistics for the creation of AI assistants, which users can use with only simple configuration. For developers and IT teams, it supports the rapid development of customized AI capabilities through DingTalk AI PaaS, and connects with original systems such as SaaS applications and on-premise systems through DingTalk's open APIs and connectors.
If these statements are too professional, perhaps these capabilities can be more intuitively understood through the DingTalk personal version fully launched by DingTalk before this press conference. This is a brand new,Even look completely different from the previous DingTalkApp(currently built-in DingTalk,You can switch at any time,Avoid the complexity of downloading),It providesAIThe dialogue,AI drawing and other functions,These functions are also the ability of AI Agent but full of to C flavor。 DingTalk has also recently held an agent development competition full of to C flavor, and the final winner is a "part-time job partner (companion version)" Agent. It can help you summarize weekly reports, edit code, record to do lists, and play games with your part-time job partners in your spare time.
"If you look at these agents, many of them are unthinkable to technicians, they are all thought of by young people, they know the scene better, and they are all really improving efficiency. Ye Jun said in a conversation with us before the press conference. He also paid attention to the many Agents created by this internal Agent competition, and also used the personal version of the AI-generated picture function to make himself look like Zhen Huan.
"We embraced AI, and the move was big and fast because AI aligned with our strategy. Ye Jun said. "AI has really made DingTalk very different from before. ”
Why can DingTalk?
In the attempt to implement the potential of large model technology into practical application, the route of AI Agent has gradually been verified as the most suitable one. But why did DingTalk become the platform that allowed Agents to prosper?
This is the first question we threw at Ye Jun.
"Cognition is key, whether you have chosen this direction or not, and you are firm when others are watching, behind it is largely your opinion on the matter. For AI, I think it's something that can make DingTalk completely different," Ye said. And he described this confidence as coming from accumulation, from DingTalk's own attributes.
"Generally, a new technology is easier to implement on the production side, on the tool side, and on the B side. He said. "It's not a 180-degree strategic adjustment for us to do AI, in fact, the complexity of our products has always been a problem we want to solve, and now we just happened to encounter AI, which can not only help us solve the complexity but also enrich our ecology, which is completely consistent with our main strategy. ”
In other words, the weakness of "the scene is too scattered and too complex", which used to be DingTalk, has become an advantage in the AI era.
Judging by the characteristics of the Agent, it seems that this is indeed the case. In a sense, the agent is a powerful technical solution to solve complex scenarios, or even to "survive" by relying on complex scenes. The power of an agent is maximized when a complex but well-defined problem is presented in front of you.
On the DingTalk platform, these scenes are simply not too many.
"If you chat with GPT alone, the information you can perceive is the input you give him, unless your history is called history, nothing else, but the ability to perceive the environment on DingTalk is very strong, because you are dealing with who you are communicating with, what group you pull the assistant into, and he is dealing with sales problems, procurement problems, organizational goal management problems, or companionship problems. Including the role of the user, the position, the personal preference is on the DingTalk, including the organizational culture, the system can be perceived. DingTalk CTO Cheng Caohong said to us.
And he also revealed that DingTalk is also further trying to improve the ability of logical reasoning, feeding the huge number of processes on OA approval to AI, which can do reasoning planning, and then independently call various materials and low-code applications that are already on DingTalk on demand.
If the core competence of the agent is simply understood as "automatic processing of tasks", then each agent can also be regarded as an individual who can be networked, although it is not as scary as general artificial intelligence, but it also has the meaning of "replace" people. Compared with the process of technology really popularizing to every human individual, the "Internet" has played a key role - from the reduction of the technical threshold to the reduction of the threshold for technology to be used by everyone, the infrastructure represented by a "Internet" to open up the interconnection of resources has played a key role, and today Agent also needs its own "Internet", rich resources that can be reached, and a low threshold for everyone to use the method. DingTalk is trying to play this role.
"From the definition of the real Agent, we don't have an accurate definition, but there is a consensus that it is indeed different from the original functional process expression of the software form, its autonomy, environmental perception ability, reasoning and planning ability, action ability, we want to make it more specific, in the current stage to be able to steadily promote these links, these links one by one in the DingTalk system has been combined. Cheng Caohong said.
Judging from the technical adjustments that DingTalk is making today, the core is to minimize the cost of building these agents on DingTalk, the cost of release, and the cost of use. An agent can be developed through natural language dialogue, and after development, it can be pulled to the DingTalk group for use.
Dingding's ambition
Another meaning of "Internet" is to let the super individuals who participate in it get rewards, which is also seen in DingTalk's products.
Ye Jun believes that after the blessing of AI, the value of individuals will become more and more obvious. "We also need to provide some weapons, for example, to a certain extent, a personal assistant will also be sold in the market, he represents a certain ability, I have trained this assistant, it is extremely capable, this is the so-called super individual. ”
"To a certain extent, DingTalk is a toB platform, but it maximizes personal value. You're not just here as a tool, you're also playing a role in more situations, whether you're negotiating contracts within the organization, or you're going to another organization, and you're training this ability to keep iterating and updating it and even making it hire by more companies. ”
In the personal version of DingTalk, some similar shadows have been seen. In the upper right corner of DingTalk's personal version, a "+" is actually a foreshadowing. Click to enter, there are only a few AI functions that can be added, but the name of the "Inspiration Store" is obviously reminiscent of the GPT store, and at the bottom of the page, there is also a guide link for developers to publish their own AI applications, which may be where the greater imagination lies.
These imaginary spaces point to a radical change in the synergistic relationship between people and between people and systems in production. DingTalk is decisively becoming an "AI Agent platform", as a national-level App that actually affects and changes the way people work collaboratively, and its transformation is also destined to affect the collaboration itself.
After the ability of AI is released in the form of an agent and the individual is liberated, how will collaboration change? Ye Jun has his own imagination, and his ambition is not small.
"I think the essence of what we're doing is not to reduce jobs, it's to reduce working hours. It is good to make people's output efficient, and the future of 8 hours of work will really become 1 hour, and the rest of the time will be good to drink tea, chat, and think. ”
He believes that this is not imaginary, but possible, based on a radical change in the relationship between people and processes.
"In the past, people were regarded as a part of the process, and all information systems were automated processes. He said. "DingTalk used to do this job, but today it's not, we're changing perspectives. In the past, the internal system was to stuff people into the mechanically designed system, and the system software you used was very uncomfortable because you had to passively adapt to these processes, but now we have to adapt people to the external environment and be truly human-centric. ”
In other words, DingTalk is "to remove the soul from automation and let the body do these things for others".
"We help these people, and each one of them eventually becomes a boss, teaching himself to work, teaching his assistant to work, and he actually becomes a boss. Ye Jun said.
"In the past, everyone always said that DingTalk was the boss, but now everyone can be the boss. ”