New Milestone in AI Development: Breakthroughs and Challenges of the Manus Model

Recently, the Manus model has achieved groundbreaking results in the GAIA benchmark test, outperforming other large language models in the same category. This achievement means that Manus can independently handle complex tasks, such as multinational business negotiations, involving contract analysis, strategy formulation, and team coordination.

The advantages of Manus are mainly reflected in three aspects: dynamic goal decomposition, cross-modal reasoning, and memory-enhanced learning. It can break down complex tasks into hundreds of executable sub-tasks, handle various types of data, and continuously improve decision-making efficiency and reduce error rates through reinforcement learning.

This development has again sparked discussions within the industry about the path of AI development: should it evolve towards a single model of Artificial General Intelligence (AGI), or should it be dominated by multi-agent systems (MAS) collaboration?

The design concept of Manus suggests two possibilities:

1. AGI Path: Continuously enhancing the capabilities of a single intelligent system to approach the comprehensive decision-making level of humans.

2. MAS Path: Use Manus as a super coordinator to direct multiple intelligent agents in various professional fields to work collaboratively.

The choice between these two paths actually reflects the balance between efficiency and safety in the development of AI. The closer a single agent gets to AGI, the harder its decision-making process becomes to explain; while multi-agent collaboration can disperse risks, it may miss critical decision-making opportunities due to communication delays.

The development of Manus also highlights the inherent risks of AI systems:

1. Data privacy issues: In fields such as healthcare and finance, Manus needs to access sensitive data.

2. Algorithmic bias: Unfair decisions may occur in areas such as human resources.

3. Security Vulnerabilities: Hackers may interfere with Manus's judgment in specific ways.

These issues reflect a concerning trend: the smarter AI systems become, the broader their potential attack surface.

In response to these challenges, the industry is exploring various security solutions:

1. Zero Trust Security Model: Strict authentication and authorization for every access request.

2. Decentralized Identity (DID): Achieve verifiable and persistent identity recognition without centralized registration.

3. Fully Homomorphic Encryption (FHE): Allows computation on data in an encrypted state, protecting privacy.

Among them, FHE is considered a key technology to solve security issues in the AI era. It can play a role in the following aspects:

• Data Layer: User information is processed in an encrypted state, and even the AI system itself cannot decrypt the original data.

• Algorithm Level: Achieve "Encrypted Model Training" through FHE to ensure the privacy of the decision-making process.

• Collaborative Level: Communication between multiple agents uses threshold encryption to enhance the overall security of the system.

In the Web3 space, some projects have begun to explore the application of these security technologies. For example, a certain project has launched a decentralized identity solution on the Ethereum mainnet, while another project focuses on the implementation of a zero-trust security model. Additionally, there is a project that has become the first FHE project to go live on the mainnet and has partnered with several well-known institutions.

As AI technology continues to approach human intelligence levels, establishing a strong security defense system becomes crucial. Advanced encryption technologies like FHE not only address current security issues but also pave the way for more powerful AI systems in the future. On the journey towards AGI, these security technologies will play an increasingly important role.