New Trends in the AI Industry: Local Small Models and Edge Computing

Recently, the AI industry has shown an interesting development trend: from the previous mainstream direction focused on large-scale computing power concentration and large models, a new branch has gradually emerged that emphasizes local small models and Edge Computing. This trend can be evidenced from multiple aspects, such as Apple Intelligence covering 500 million devices, Microsoft launching a dedicated small model Mu with 330 million parameters for Windows 11, and Google's DeepMind developing robots that can run "offline".

What differences does this transformation bring? Cloud AI primarily relies on large parameter scales and massive training data, making financial strength a key competitive factor. In contrast, local AI focuses more on engineering optimization and scenario adaptation, with clear advantages in protecting user privacy and improving reliability and practicality. This is especially important because the "hallucination" problem that often occurs when applying general models in specific fields severely limits their promotion in vertical scenarios.

For Web3 AI projects, this trend may bring more opportunities. Previously, when the industry focused on "generalization" abilities including computing, data, and algorithms (, traditional tech giants naturally dominated. In this context, merely applying the concept of decentralization to compete with industry giants is akin to dreaming. After all, compared to these giants, Web3 projects are at a disadvantage in terms of resources, technology, and user base.

However, with the rise of localized models and Edge Computing, the application prospects of blockchain technology in the AI field have become even broader. When AI models run on users' own devices, how can we ensure the authenticity of the output results? How can we achieve collaboration between models while protecting privacy? These questions are precisely the areas where blockchain technology excels at solving.

Some new projects have emerged in the industry to address these challenges. For example, a project that recently secured $10 million in funding launched a data communication protocol aimed at solving the issues of data monopoly and black box operations present in centralized AI platforms. Another project collects real human data through brainwave devices to construct an "artificial verification layer" and has already achieved $14 million in revenue. These projects are all trying to tackle the "credibility" issues faced by local AI.

Overall, decentralized collaboration can only transform from a concept into a real demand when AI technology truly "sinks" into every user device. For Web3 AI projects, rather than continuing to compete in the already crowded generalized track, it is better to seriously consider how to provide the necessary infrastructure support for the upcoming wave of localized AI. This might be a more promising development direction.