Author: Haotian
Recently observed the AI industry, and found an increasingly 'sinking' change: evolving from the original mainstream consensus of centralized computing power and 'large' models, a branch leaning towards local small models and edge computing has emerged.
This is evident from Apple Intelligence covering 500 million devices, to Microsoft's launch of the Windows 11 dedicated 330 million parameter small model Mu, and Google's DeepMind robot's 'offline' operations, etc.
What difference will it make? Cloud-based AI competes on parameter scale and training data, with burning money capability as the core competitiveness; local AI competes on engineering optimization and scene adaptation, making further advancements in privacy protection, reliability, and practicality. (The illusion problem of major general models will severely affect the penetration of vertical scenarios.)
This actually presents a greater opportunity for web3 AI. Originally, when everyone was competing on 'generalization' (computing, data, algorithms), they were naturally monopolized by traditional giant companies. Trying to compete with Google, AWS, OpenAI, etc., by applying the concept of decentralization is simply wishful thinking, as there is no resource advantage, technical advantage, and certainly no user base.
However, in the world of localized models + edge computing, the situation faced by blockchain technology services is quite different.
When AI models run on user devices, how can we prove that the output results have not been tampered with? How can we achieve model collaboration while protecting privacy? These questions are precisely the strengths of blockchain technology...
Have noticed some new projects related to web3 AI, such as the recently launched data communication protocol Lattica by @Gradient_HQ, which received zero investment of 10M from Pantera, aimed at solving the data monopoly and black box issues of centralized AI platforms; @PublicAI_'s brainwave device HeadCap collects real human data to construct a 'human verification layer' and has already achieved 14M in revenue; in fact, they are all trying to address the 'trustworthiness' issue of local AI.
In a nutshell: only when AI truly 'sinks' into every device will decentralized collaboration shift from being a concept to a necessity?
#Web3AI projects might as well seriously consider how to provide infrastructure support for the localized AI wave, rather than continuing to compete in the generalized track.
