A New Peak in AI Development: The Manus Model and Its Implications

Recently, the Manus model has made breakthrough progress in the GAIA benchmark tests, outperforming other large language models of its class. This achievement means that Manus is capable of independently handling complex tasks such as multinational business negotiations, involving multiple aspects like contract analysis, strategy formulation, and proposal generation.

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

This development has once again sparked discussions within the industry about the path of artificial intelligence: should it develop towards artificial general intelligence (AGI), or should multi-agent systems (MAS) take the lead in collaboration?

From the design concept of Manus, it suggests two possible directions for development:

1. AGI Path: By continuously enhancing the capabilities of a single intelligent system, it gradually approaches the comprehensive decision-making ability of humans.

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

The discussion of these two paths actually touches upon a core issue in the development of AI: how to strike a balance between efficiency and safety? As monolithic intelligent systems get closer to AGI, the risks associated with the opacity of their decision-making processes also increase. While multi-agent collaboration can disperse risks, it may miss critical decision-making opportunities due to communication delays.

The progress of Manus also highlights some inherent risks in the development of AI:

1. Data privacy issues: In fields such as healthcare and finance, AI systems need to access a large amount of sensitive information.

2. Algorithmic Bias: Unfair decisions may arise in fields such as human resources.

3. Security vulnerabilities: The system may be subject to malicious attacks, leading to erroneous judgments.

These issues highlight a fact: the smarter the AI system, the broader its potential attack surface.

When addressing these issues, Fully Homomorphic Encryption (FHE) technology has demonstrated great potential. FHE allows computations to be performed on encrypted data, providing possible solutions to security issues in the AI era:

1. Data layer: All information entered by users is processed in an encrypted state, and even the AI system itself cannot decrypt the original data.

2. Algorithm Level: Achieve "encrypted model training" through FHE to protect the decision-making process of AI.

3. Collaborative Level: Communication between multiple agents uses threshold encryption to enhance the overall system's security.

As AI technology continues to approach human intelligence levels, establishing a robust security defense system becomes increasingly important. FHE not only addresses current security issues but also paves the way for more powerful AI systems in the future. In the journey toward AGI, security technologies like FHE will play an increasingly important role, becoming an indispensable guarantee for the development of AI.