Manus achieves breakthrough results in GAIA Benchmark, sparking debates on the path of AI development

Recently, Manus achieved groundbreaking results in the GAIA benchmark, outperforming large language models of the same tier. This achievement indicates that Manus has the capability to independently complete complex tasks, such as multinational business negotiations, including contract clause breakdown, strategy prediction, plan generation, and even coordinating legal and financial teams.

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

This development has once again sparked discussions within the industry about the evolutionary path of AI: will the future lead to a world dominated by General Artificial Intelligence (AGI), or will it be a collaborative dominance of Multi-Agent Systems (MAS)?

The design concept of Manus implies two possibilities:

1. AGI Path: Continuously improve the level of individual intelligence to approach human comprehensive decision-making ability.

2. MAS Path: As a super coordinator, directing thousands of agents in vertical domains to work together.

On the surface, this is a debate about different development paths, but it actually reflects the core contradiction in AI development: how to balance efficiency and safety. As single intelligence approaches AGI, the risk of decision-making opacity increases; while multi-agent collaboration can disperse risks, it may miss critical decision-making opportunities due to communication delays.

The progress of Manus has inadvertently amplified the inherent risks of AI development, including:

1. Data privacy issues: In medical scenarios, there is a need for real-time access to sensitive patient data; in financial negotiations, there may be undisclosed financial report information related to the company.

2. Algorithmic Bias: In recruitment negotiations, there may be unfair salary suggestions for specific groups; during the review of legal contracts, the misjudgment rate for clauses in emerging industries is relatively high.

3. Adversarial Attack Vulnerability: Hackers may implant specific voice frequencies, causing the system to misjudge the opponent's price range during negotiations.

These issues highlight a grim reality: the more intelligent the system, the broader its attack surface.

In the Web3 space, security has always been a topic of great concern. Around this theme, various cryptographic solutions have emerged:

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

2. Decentralized Identity (DID): Achieving identity recognition without a centralized registry.

3. Fully Homomorphic Encryption (FHE): Allows computation on encrypted data without decrypting it.

Among them, fully homomorphic encryption is considered a key technology to solve security issues in the AI era. It can provide protection at several levels:

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

• Algorithm level: Achieving "encrypted model training" through FHE, ensuring that even developers cannot peek into the AI's decision-making path.

• Collaborative level: Communication between multiple agents uses threshold encryption to prevent single point leaks from exposing global data.

Although Web3 security technologies may not have a direct connection to ordinary users, they are crucial for protecting user interests. In this unknown field, strengthening security measures is a necessary means to avoid becoming "chives."

Historically, multiple projects have explored the field of Web3 security:

• uPort was launched on the Ethereum mainnet in 2017 and is one of the earlier decentralized identity projects.
• NKN launched its mainnet based on a zero-trust security model in 2019.
• A certain network is the first FHE project to launch on the mainnet and has reached partnerships with several well-known institutions.

As AI technology continues to approach human intelligence levels, non-traditional defense systems are becoming increasingly important. Secure technologies such as FHE not only address current issues but also lay the foundation for the future era of strong AI. On the road to AGI, these security technologies are no longer optional but are essential for survival.