Looking at the next-generation AI infrastructure paradigm from the Computing Power alliance between Flock and Alibaba.

Yesterday, the DeAI training platform Flock in the Web3AI field officially announced a partnership with Alibaba Qwen, a large language model under Alibaba Cloud. If I remember correctly, this should be considered the first integration cooperation initiated by web2 AI towards web3 AI. It not only allowed Flock to truly break out of its circle but also revitalized the morale of the web3 AI track under the pressure of a downturn.

1. I have elaborated in the pinned tweet that the previous web3 AI Agent has been trying to stimulate the implementation of Agent applications through Tokenomics and has also engaged in a rapid deployment competitive paradigm. However, after the FOMO frenzy of asset issuance, everyone found that when it comes to practicality, innovation, etc., web3 AI has almost no chance of winning compared to web2 AI.

The emergence of web2 innovative AI technologies such as Manus, MCP, and A2A has directly or indirectly pierced the bubble in the Web3 AI Agent market, resulting in a moment of bloodshed in the secondary market.

2. How to break the deadlock? The path is actually quite clear. Web3 AI urgently needs to find a complementary ecological niche to Web2 AI, in order to address the high cost of computing power, data privacy issues, and the fine-tuning of vertical scene models that Web2 centralized AI cannot solve.

The reasons are that pure centralized AI models will inevitably face explosive problems in the end regarding the channels and costs of obtaining computing resources, as well as issues related to data resource privacy. In contrast, the distributed architecture attempted by web3 AI can reduce costs by utilizing idle computing resources, and it will also protect privacy based on hardware and software technologies such as zero-knowledge proofs and TEE. Additionally, it aims to promote the development and fine-tuning of models in vertical scenarios through data ownership and incentive contribution mechanisms.

No matter how criticized, the decentralized architecture and flexible incentive mechanisms of web3 AI can have an immediate effect on solving some of the problems existing in web2 AI.

3. Speaking of the collaboration between Flock and Qwen. Qwen is an open-source large language model developed by Alibaba Cloud, which has become a popular choice among some developers and research teams due to its outstanding performance in benchmark tests and the flexibility it allows developers to locally deploy and fine-tune.

Flock is a decentralized AI training platform that integrates AI federated learning and AI distributed technology architecture. Its main feature is to protect user privacy through distributed training while ensuring that "data does not leave the local environment," allowing for transparent and traceable data contributions, thus addressing the fine-tuning and application issues of AI models in vertical fields such as education and healthcare.

Specifically, Flock has three key components, which I will briefly share here:

1. AI Arena, this is a competitive model training platform where users can submit their own models to compete against others for optimization results and vie for rewards. Its main purpose is to encourage users to continuously fine-tune and improve their local large models through a gamified mechanism design, thereby selecting better benchmark models.

2. FL Alliance, to address the cross-organizational collaboration issues present in traditional sensitive vertical scenarios such as healthcare, education, and finance, has achieved a collaborative framework through localized model training and distributed collaboration, enabling multiple parties to enhance model performance without sharing raw data.

3. Moonbase, it serves as the neural hub of the Flock ecosystem, equivalent to a decentralized model management and optimization platform, providing various fine-tuning tools and computing power support (computing power providers, data annotators). It not only offers a distributed model repository but also integrates fine-tuning tools, computing resources, and data annotation support, empowering users to efficiently optimize local models.

4. So, how should we view the cooperation between Qwen and Flock? In my personal opinion, the extended significance of their collaboration is even greater than the current substance of the partnership.

On one hand, against the backdrop of web3 AI being continuously overwhelmed by web2 AI technology, Qwen, representing the tech giant Alibaba, has already gained a certain level of authority and influence within the AI circle. Qwen's active choice to collaborate with a web3 AI platform fully demonstrates web2 AI's recognition of the Flock technology team. At the same time, the series of research and development efforts between the Flock team and the Qwen team will deepen the interaction between web3 AI and web2 AI.

On the other hand, the previous web3 AI once had the shell of Tokenomics, and its performance in the actual Utility landing was very unsatisfactory, although it tried a variety of AI Agent, AI Platform, and even AI Framework and many other directions, but it was not possible to come up with a real solution to the problem in terms of DeFai and Gamefai. To a certain extent, the brand from the web2 technology giant has set the tone for the future development path and force of web3 AI;

The most crucial thing is that web3 AI, after experiencing a pure Fomo frenzy of "asset issuance," needs to regroup and focus on a goal that can deliver real results.

In fact, web3 AI has never been just an easier and more efficient channel for deploying AI agents to issue assets, nor is it a game of raising funds by issuing assets. It is necessary to seek collaboration with web2 AI and complement each other's ecological niches, so that web3 AI can truly demonstrate its indispensable role in this wave of AI trends.

I am glad to see more cross-border collaborations like web2AI and web3AI being achieved.