Four products help you understand what AI Agent is.

Author: Mu Mu

If AIGC is a typical application of the new AI era in 2023, then in 2024, AI Agent (AI intelligent agent) will truly productize the capabilities of AIGC.

A sentence that arranges schedules, writes text messages, and sends emails for GPT is just a piece of cake. Recently, GPT can even generate long videos with text. Compared with large models like GPT, AI Agent is more like a more concrete all-around employee, which is the primary form of artificial intelligence robots. The intelligent body that takes care of both software and hardware can observe the surrounding world like a human, make decisions, and take corresponding actions automatically.

As Microsoft co-founder Bill Gates said, "Whoever can dominate AI Agent, that is the big deal. Because you will never have to search websites or Amazon."

More and more companies are focusing on the AI Agent field. According to a set of data from "Entrepreneur", as of November 2023, there have been 13 financing events in the domestic Agent track, with a total financing amount of approximately 73.5 billion RMB, and the average financing amount per company is 5.654 billion RMB.

In China, the former vice president of DingTalk has entered the AI Agent track and established BanTouYan Intelligent Technology, which has received hundreds of millions of RMB in financing. In foreign countries, this field is even more popular. Data shows that "at least 100 projects are dedicated to commercializing AI intelligent agents, and nearly 100,000 developers are building autonomous agents."

In terms of products, the appearance of AI Agents such as AI wearable device Rabbit R1, which can help you call a taxi at any time, and Samsung robot Ballie, which can detect the owner's emotions and take care of pets, all indicate that AI Agent will become the new frontier of intelligent technology. The following four applications will help you understand which internet experiences AI Agent will subvert.

ARC Search Reshapes Traditional Search Experience

ARC Search is the latest AI Agent launched in January 2024, claiming to subvert the traditional search experience. Typically, when we open a search engine to find a topic of interest, we need to enter keywords, browse web pages, and search for useful information from hundreds of relevant links.

In the era of information explosion, obtaining useful information is no longer easy. For browsers such as Chrome and Safari, ARC Search's entry point is speed, that is, to allow users to obtain the most useful information in the shortest time and with the most convenient interaction.

Arc Search has two browsing modes: traditional mode, which is accessed through keywords or URLs, and AI mode, in which users can input questions or prompts in the form of natural language dialogue.

In AI mode, the search results obtained by users are no longer a chaotic and fragmented mass of information. The retrieval not only includes AI-generated answers, but also has a unified layout. The results contain key points, the best search results, specific content, and related web links.

For example, when searching for "how to make perfect omelets," Arc Search will provide a graphic and textual cooking tutorial, including "ingredients and tools," "cooking process," and "notes," accompanied by relevant photos or videos.

ARC Search automatically summarizes search content (Image from TechCrunch)

Based on the traditional search method, Arc Search not only has the ability of AIGC (AI-generated content), but also allows AI to act as an Agent, automatically analyzing and making decisions, organizing, summarizing, and coming up with the "best answer."

Voyager Acts as a Human Trainee in "Minecraft"

Voyager is a large model-driven game intelligent agent launched by NVIDIA. In May 2023, NVIDIA open-sourced Voyager and applied it in "Minecraft."

In the game "Minecraft," apart from survival and freely exploring the open world, no specific goals are set for players. In this highly free game, Voyager can autonomously explore for hours, adjust decisions based on the environment, and develop skills to fight monsters and find food as needed.

Voyager can self-learn in "Minecraft"

The Voyager applied in the game can learn from mistakes until it finds the correct way and stores the execution program for future use.

The capabilities demonstrated by Voyager in "Minecraft" are just NVIDIA's initial exploration of AI Agent, proving that Voyager does not need to run according to a script, but can continuously learn, make decisions, and take action like a human. According to NVIDIA's expectations, they hope to introduce Voyager into various scenarios such as virtual space, drone operations, and humanoid robots.

DingTalk's "AI Super Assistant" Enters Workflow

Since its establishment in 2014, DingTalk has quickly entered the enterprise management system and become an office software that employees cannot escape from. With the increasing diversity of DingTalk's functions, the operation interface has become more and more cumbersome. The complex work system has led to the problem of "many functions that are not needed, and the functions that are needed cannot be found."

Ten years later, DingTalk has 700 million users. In the era of AI, this office software will also use AI to increase efficiency for users. On January 9, at the DingTalk 7.5 product launch event, AI Agent became the main highlight.

DingTalk releases AI Super Assistant

According to the demonstration, in the natural language dialogue window, the AI assistant can complete more than 30 tasks in the DingTalk system, including setting functions, scheduling, handling tasks, work overview, message summaries, leave approvals, content creation and editing, and intelligent drawing.

It is worth mentioning that the DingTalk AI assistant has the ability to take cross-system actions. After obtaining open APIs from external platforms, it can connect to various third-party platforms such as shopping, video, and websites. It is reported that DingTalk will launch the AI Assistant Market, SkyAgents, in April this year and invite developers to join this ecosystem.

SkyAgents: Zero-Code Creation of AI Intelligent Agents

Currently, there are not only AI intelligent agents tailored for specific scenarios on the market, but also enterprises developing tools to create and deploy intelligent agents.

On December 1st of last year, Kuanwu Wanwei released a tool for developers to build AI Agents called "SkyAgents" (天工). On December 25th, the beta version of SkyAgents was officially open for testing.

SkyAgents is built on Kuanwu Wanwei's large-scale model and is claimed to have the ability to learn autonomously and think independently from perception to decision-making and execution. It can achieve codeless operation, allowing users to design their own large-scale model AI assistants even without coding experience.

SkyAgents testing page

On the SkyAgents platform, users can deploy their own AI Agents in minutes through natural language and simple operations without the need for coding, including industry research reports, document filling, trademark design, and even personalized needs such as fitness plans and travel flight reservations.

Conclusion

Existing AI Agents are built on the capabilities of AIGC, with text generation being the mainstream. As large models continue to develop, multimodal generation capabilities such as voice, image, and video will also appear in AI Agents, optimizing the experience of using internet applications with artificial intelligence.

So, what exactly are AI Agents disrupting?

Bill Gates believes that AI Agents will disrupt the software industry in the coming years. He predicts this disruption as "a shock wave to the tech industry" and "the biggest change in the computer field after the transformation from entering commands to clicking icons."

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。