Source: Market Information

Wall Street Vision

This agent can perform multi-step complex tasks with minimal instructions and has the ability to learn from past actions and self-improve. This marks DeepSeek's shift from a reasoning model to a more autonomous AI system, competing with OpenAI, Anthropic, and others in the next stage of AI development.

Is DeepSeek challenging OpenAI again?

Today, Bloomberg revealed a major insider news:

DeepSeek has been holding back a super big move - the real AI agent (AI Agent)!

DeepSeek R1, the previous open-source king

In January this year, the DeepSeek R1 model achieved performance comparable to or even surpassing OpenAI's top products at a cost of only a few million dollars in multiple benchmark tests.

The industry was shocked worldwide: The annual salary of Meta's GenAI department executives is not less than one million US dollars.

After that, DeepSeek continuously disclosed the underlying technology for five days, promoting the advancement of human AGI through open source.

On the sixth day, it released the technical architecture of DeepSeek V3, as well as a potentially more important open-source business model for AI.

Subsequently, both domestic and international AI communities vied for the open-source AI throne, which changed hands several times.

Even due to DeepSeek's open-source influence, OpenAI once again opened up AI: After many years, OpenAI provided an open-source weight model again.

But with GPT-5 from OpenAI coming, DeepSeek hasn't released the next generation of reasoning models or the next major update, only releasing a few minor updates.

This time, the secret is revealed: DeepSeek is building a new AI model, designed to execute multi-step complex tasks with minimal user instructions.

Releasing the agent by the end of the year, and enabling AI self-improvement

Bloomberg reported that DeepSeek's founder Liang Wenfeng is pushing the team to launch a new AI in the fourth quarter of this year - creating a true AI agent.

Additionally, the new AI system can learn and self-improve based on past actions.

DeepSeek is aiming directly at the next stage of AI development.

Bloomberg believes this demonstrates Liang Wenfeng's determination to maintain leadership in the competitive industry.

Many chatbots can only provide simple text responses. In contrast, agents aim to handle more complex tasks, with broader applications, covering everything from planning holidays to writing and debugging code.

For DeepSeek and the entire industry, the goal is to build increasingly autonomous AI systems that can initiate and perform complex real-world tasks with minimal human intervention.

Over the past few months, OpenAI, Anthropic, and Microsoft have all launched their own agents to simplify personal and work tasks.

However, so far, AI agents still require a lot of human supervision - agents are just beginning.

Perhaps, according to DeepSeek, AGI has not yet been achieved, and the open-source community still needs to strive.

This article is sourced from Xin Zhi Yuan, original title: "DeepSeek's Big Move Exposed? Liang Wenfeng Urges the Super Intelligent Agent: Capable of Autonomous Learning, Possibly Released by Year-End"

Risk Notice and Disclaimer

Market risks exist, and investment should be cautious. This article does not constitute personal investment advice, nor has it considered individual users' special investment goals, financial conditions, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific situations. Investing based on this is at your own risk.

Original: https://www.toutiao.com/article/7546393428817363475/

Statement: This article represents the views of its author. Welcome to express your opinion by clicking the [Like/Dislike] button below.