The "decentralized model marketplace" of Model Hub is described in section 8.1 of the @OpenGradient whitepaper like a developer's paradise: upload models, earn OPG royalties, deploy with one click. But that monster called the "Model Review Committee" in section 8.3 turns paradise into approval hell.
The whitepaper states that all models listed on Model Hub must undergo quality checks by this committee to prevent malicious models from contaminating the ecosystem. Sounds reasonable until you find out that this review power is held by a group of anonymous "initial guardians" with no public identities, no selection criteria, and no term limits. The on-chain governance proposal doesn’t even outline how to impeach them. Whether your model can earn $OPG in the OpenGradient network depends not on a clean piece of open-source code but on the subjective scores from a few individuals.
Even more absurd is the bundling of review with tokens. Section 6.9 stipulates that model uploaders must stake OPG as a review deposit; if the review fails, part of the deposit is deducted and given to the committee members. This creates a terrifying feedback loop: committee members are economically incentivized to reject models because approval only yields minimal token rewards, while a rejection allows them to pocket the deposit immediately. Human nature can’t resist this kind of temptation, yet the whitepaper pretends it’s a perfect game balance.
Section 1 shouts "resist centralized review power" but then hands all the life-and-death power of AI service providers to a centralized committee. Whether your model can reach users, and whether your OPG stake is safe, entirely depends on the breakfast mood of these guardians on a given day. $BTC
The harsh reality soon sets in. Developers are already complaining on Discord that their models were rejected for using "undeclared geophysical data training sets," with no specific information on which data violated the rules. The review reports are cryptographically hashed and stored on-chain, with section 3.4 boasting that it is a "permanent and traceable transparent record." Transparent to what extent? You only see a string of hash values with no human explanation.
If you want to strike gold by creating AI models in the #OPG ecosystem, you first have to cater to a bunch of anonymous folks holding veto power. Is this still decentralized? It’s just swapping the Apple App Store reviewers for mysterious individuals wearing Web3 masks, who even hide the appeal button deep within long DAO proposal queues.
#OPG $OPG @OpenGradient
The whitepaper states that all models listed on Model Hub must undergo quality checks by this committee to prevent malicious models from contaminating the ecosystem. Sounds reasonable until you find out that this review power is held by a group of anonymous "initial guardians" with no public identities, no selection criteria, and no term limits. The on-chain governance proposal doesn’t even outline how to impeach them. Whether your model can earn $OPG in the OpenGradient network depends not on a clean piece of open-source code but on the subjective scores from a few individuals.
Even more absurd is the bundling of review with tokens. Section 6.9 stipulates that model uploaders must stake OPG as a review deposit; if the review fails, part of the deposit is deducted and given to the committee members. This creates a terrifying feedback loop: committee members are economically incentivized to reject models because approval only yields minimal token rewards, while a rejection allows them to pocket the deposit immediately. Human nature can’t resist this kind of temptation, yet the whitepaper pretends it’s a perfect game balance.
Section 1 shouts "resist centralized review power" but then hands all the life-and-death power of AI service providers to a centralized committee. Whether your model can reach users, and whether your OPG stake is safe, entirely depends on the breakfast mood of these guardians on a given day. $BTC
The harsh reality soon sets in. Developers are already complaining on Discord that their models were rejected for using "undeclared geophysical data training sets," with no specific information on which data violated the rules. The review reports are cryptographically hashed and stored on-chain, with section 3.4 boasting that it is a "permanent and traceable transparent record." Transparent to what extent? You only see a string of hash values with no human explanation.
If you want to strike gold by creating AI models in the #OPG ecosystem, you first have to cater to a bunch of anonymous folks holding veto power. Is this still decentralized? It’s just swapping the Apple App Store reviewers for mysterious individuals wearing Web3 masks, who even hide the appeal button deep within long DAO proposal queues.
#OPG $OPG @OpenGradient
谁在掌控模型生杀大权❌
0%
为何他们倾向否决你质押金是不是被盯着坑如何反制主观审核
0%
0 votes • Voting closed