When Trump Invests Trillions in AI, Who Provides Credible Data?
When Trump invests trillions of dollars in AI, it appears to be a competition of models, chips, and data centers, but it also raises deeper questions: How are the data relied upon by AI models verified, traceable, auditable in the training and reasoning processes, and can models collaborate or are they fighting alone?
In simpler terms, when we use AI to obtain information, who can ensure that the information provided by AI is correct? Data pollution is no longer just a casual term. An AI application once claimed to be a ChatGPT killer has already been deeply entrenched in a data pollution environment. When data sources are incorrect, how can the answers be right?
Is current AI intelligent? Perhaps, but even the smartest AI needs model training, yet we cannot know which data was used for training, verify if GPUs truly completed an inference process, or establish inter-model trust logic.
To truly advance AI to the next generation, these three problems might need to be solved simultaneously:
1. Training data must be credible and verifiable.
2. Inference processes must be auditable by third-party models.
3. Models must be able to coordinate computing power, exchange tasks, and share results without platform mediation.
This cannot be solved by a single model, API, or GPU platform, but requires a system truly built for AI. This system should be able to store data cost-effectively and permanently, give data itself the right to review and be reviewed, enable inter-model reasoning verification, and support models' autonomous discovery of computing power, task coordination, and step-by-step auditing.
This is difficult to achieve on centralized platforms, so could it be possible on a decentralized platform? Why use a decentralized approach?
I believe only blockchain can truly integrate "data storage, data execution, and data verification" into the same underlying network. This is one of blockchain's greatest attractions: immutability and transparency. However, not every chain is suitable for AI's underlying infrastructure.
While IPFS protocol exists for storage, mere storage is insufficient. It needs to allow smart contracts to directly call data, audit inference results, and even coordinate GPU resources to complete computational tasks - features that most L1 or AI applications cannot yet achieve.
If there's any relevance, <@irys_xyz> might have a chance. Irys is not a traditional storage chain but aims to build a data execution network for AI, treating data as programmable assets. Models can read data, verify inference, call computing power on-chain, and implement pricing, authorization, profit sharing, and verification through smart contracts.
Of course, Irys still has some immature aspects, but the development direction seems correct. Whether centralized or decentralized AI, if data sources are not credible, all computing power is like building a castle on sand - even the strongest model is but a moon in water, a flower in a mirror.
