AI-Based Crypto Tokens: The Illusion of Decentralized AI?

Table of Contents

1 Gabatarwa

Haɗewar blockchain da hankalin wucin gadi (AI) sun haifar da bayyanar alamun AI, waɗanda su ne kadarorin sirri da aka ƙera don ƙarfafa dandamalin AI da ayyuka. Waɗannan alamun suna nufin canza ikon sarrafa fasahar AI daga ƙungiyoyin masu zaman kansu zuwa buɗaɗɗen tsarin mulki na al'umma. Babban abin da ya sa ake yi shi ne haɓaka ayyukan AI waɗanda ke nuna ka'idodin blockchain: ƙaddamarwa, cin gashin kai, da mallakar mai amfani akan bayanai da hanyoyin lissafi.

Bayan fitar da ChatGPT a ƙarshen 2022, kadarorin crypto masu alaƙa da AI sun sami babban riba mara kyau, tare da kololuwar ribar da ta wuce 41% a cikin makonni biyu. Wannan martanin kasuwa yana tayar da muhimman tambayoyi game da ko waɗannan alamun suna wakiltar amfani na fasaha na gaske da ƙaddamarwa ko kuma kawai suna amfani da labarun da suka shafi AI don ribar kuɗi.

2 Tsarin Fasahar AI Tokens

2.1 Token Utility Models

AI tokens suna cika ayyuka da yawa a cikin yanayinsu:

• Biyan Kuɗi don Ayyuka: Tokens kamar RENDER da AGIX suna sauƙaƙe biyan kuɗin lissafin AI da samun damar samfur
• Haƙƙin Gudanarwa: Masu riƙe Token suna shiga cikin yanke shawara na dandamali
• Staking Mechanisms: Masu amfani suna saka token don samun damar albarkatun cibiyar sadarwa da kuma samun lada
• Data MonetizationProtocols like Ocean Protocol suna ba da damar raba bayanai da samun kuɗi

2.2 Consensus Mechanisms

Ayyukan AI token daban-daban suna amfani da hanyoyin yarwa daban-daban:

• Proof-of-Stake variants: Used by platforms like Fetch.ai for network security
• Federated Learning ConsensusBittensor's approach combining AI model performance with consensus
• Hybrid ModelsCombining traditional blockchain consensus with AI-specific validation

3 Iyakoki da Kalubale

3.1 Iyakokin Fasaha

A halin yanzu AI token aiwatarwa na fuskantar manyan kalubalen fasaha:

• Dogaron Ƙididdiga na Kashe-sarkar: Yawancin sarrafa AI yana faruwa a kashe-sarkar, yana iyakance fa'idodin rarrabawa
• Matsalolin Ƙarfin GudanarwaAyyukan AI akan sarkar suna fuskantar iyakancewar kayan aiki
• Iyakancewar Hankali akan SarkarKayayyakin sarkar na yanzu ba za su iya tallafawa aiwatar da hadadden samfurin AI ba

3.2 Business Model Concerns

Yawancin ayyukan AI token suna kwafa tsarin tsakiya:

• Ƙananan sassa na biyan kuɗi na token da aka ƙara zuwa tsarin sabis na al'ada
• Hanyoyin gudanarwa waɗanda ba sa canza yanayin iko sosai
• Ƙayyadadden sabon ƙima fiye da waɗanda suka tsakiya AI ayyuka

4 Experimental Results

Binciken Ayyukan Kasuwa

Bincike ta [11, 12] ya rubuta muhimman martanin kasuwa ga sanarwar token na AI:

Za a iya ƙirƙira motsin farashin ta amfani da samfurin farashin kadarorin babban birnin (CAPM):

$R_{it} - R_{ft} = \alpha_i + \beta_i(R_{mt} - R_{ft}) + \epsilon_{it}$

A inda $R_{it}$ yana da dawowar AI token i a lokacin t, $R_{ft}$ shine ƙimar rashin haɗari, kuma $R_{mt}$ shine dawowar kasuwa.

5 Aiwarta Fasaha

Misalin Kwangila na Yarjejeniya

A nan wani sauƙaƙan kwangilar wayo don kasuwar samfurin AI:

pragma solidity ^0.8.0;

contract AIModelMarketplace {
    struct Model {
        address owner;
        string modelHash;
        uint256 price;
        bool isActive;
    }
    
    mapping(uint256 => Model) public models;
    uint256 public modelCount;
    
    event ModelListed(uint256 modelId, address owner, uint256 price);
    event ModelPurchased(uint256 modelId, address buyer, uint256 price);
    
    function listModel(string memory _modelHash, uint256 _price) public {
        modelCount++;
        models[modelCount] = Model({
            owner: msg.sender,
            modelHash: _modelHash,
            price: _price,
            isActive: true
        });
        emit ModelListed(modelCount, msg.sender, _price);
    }
    
    function purchaseModel(uint256 _modelId) public payable {
        Model storage model = models[_modelId];
        require(model.isActive, "Model not available");
        require(msg.value >= model.price, "Insufficient payment");
        
        payable(model.owner).transfer(model.price);
        emit ModelPurchased(_modelId, msg.sender, model.price);
    }
}

Haɗin Koyo na Tarayya

Haɗin toshe shinge tare da koyo na tarayya za a iya wakilta ta hanyar lissafi:

$\min_{w} \sum_{k=1}^{K} \frac{n_k}{n} F_k(w) + \lambda R(w)$

Ina $F_k(w)$ shine ma'ana aikin haƙiƙa na gida don abokin ciniki k, $n_k$ shine adadin wuraren bayanai a abokin ciniki k, kuma $R(w)$ shine lokacin daidaitawa.

6 Aikace-aikace na Gaba

Ci Gabobi na Tasowa

• Tabbatarwa akan SilsilaZero-knowledge proofs for AI output verification
• Blockchain-enabled Federated LearningSecure aggregation of AI models without data sharing
• Robust Incentive Frameworks: Improved tokenomics for sustainable ecosystems
• Cross-chain AI Services: Interoperable AI models across multiple blockchains

Technical Roadmap

Ci gaba na ci gaba suna mai da hankali kan magance iyakokin na yanzu:

• Ai watsi da lissafin kwamfuta don ayyukan AI
• Haɓaka ƙwararrun blockchains na AI
• Haɗin kai tare da binciken amincin AI da daidaitawa

7 Bincike na Asali

Fitarwar da'awar tsabar kuɗi na AI na wakiltar haɗin gwiwa mai ban sha'awa na fasahohi biyu masu canzawa, duk da haka bincikenmu ya bayyana manyan gibubbe tsakanin alkawurransu na ka'ida da aiwatar da su. Yin kwatankwacin ci gaban hanyoyin sadarwa masu adawa (GANs) kamar yadda aka rubuta a cikin takardar CycleGAN na asali (Zhu et al., 2017), muna lura da irin wannan tsarin inda ɗimbacin fasaha sau da yawa ya wuce ƙwararrun ƙira. Yayin da ayyuka kamar SingularityNET da Bittensor ke nufin ƙirƙirar kasuwannin AI marasa tsari, tsarinsu na yanzu ya dogara sosai akan lissafin kashe-kashe, yana haifar da matsalolin matsugunan tsakiya waɗanda ke ɓata ka'idodin blockchain.

Daga fasahar fasaha, iyakokin haɓaka suna da damuwa musamman. Kamar yadda aka lura a cikin sabunta hanyar Ethereum da bincike daga cibiyoyi kamar Cibiyar Blockchain ta Stanford, tsarin tushen blockchain na yanzu ba zai iya sarrafa buƙatun lissafi na hadaddun samfuran AI yadda ya kamata ba. Tushen ilimin lissafi na hanyoyin yarjejeniya da yawa, yawanci bisa bambance-bambancen hujjar mallaka tare da $text{Pr}(text{zaɓi}) \propto \text{stake}^{\alpha}$, yana gwagwarmaya don haɗa ma'auni mai ma'ana na ingancin samfurin AI ba tare da gabatar da sabbin hanyoyin tsakiya ba.

Yanayin kasuwa da ke kewaye da tokens na AI bayan fitowar ChatGPT yana bayyana mafi zurfin matsaloli game da ƙimar ƙima a cikin yanayin crypto. Bisa ga bayanai daga CoinGecko da binciken ilimi daga dandamali kamar SSRN, hauhawar farashin kashi 41% da aka gani a cikin tokens na AI ya bayyana galibi ya rabu daga ci gaban fasaha na asali. Wannan tsari yayi kama da kumfa na crypto na farko inda hasashe mai kunnawa labari ya rufe cancantar fasaha. Koyaya, ci gaba mai ban sha'awa a cikin injin ilimin sifili (zkML) da ingantaccen tunani, kamar yadda ƙungiyoyi a Berkeley da MIT suka bincika, suna ba da hanyoyin yuwuwar zuwa ga AI na gaske ta hanyar ba da damar tabbatar da kan layi na lissafin kashe layi.

Our critical evaluation suggests that while current implementations may represent an "illusion of decentralization," the underlying vision remains valid. The integration of blockchain's trustless verification with AI's predictive capabilities could eventually yield novel applications that transcend what either technology can achieve independently. However, achieving this potential requires more rigorous technical foundations and honest assessment of current limitations, moving beyond the AI-themed financial speculation that currently dominates the space.

8 References

1. Zhu, J. Y., Park, T., Isola, P., & Efros, A. A. (2017). Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks. IEEE International Conference on Computer Vision.
2. Buterin, V. (2014). Ethereum White Paper: A Next-Generation Smart Contract and Decentralized Application Platform.
3. Goodfellow, I., et al. (2014). Generative Adversarial Networks. Neural Information Processing Systems.
4. Render Network Whitepaper (2023). Decentralized GPU Rendering Platform.
5. SingularityNET Foundation (2021). SingularityNET Whitepaper and Protocol Documentation.
6. Ocean Protocol Foundation (2022). Ocean Protocol: Kayan Aikin Tattalin Arzikin Bayanai na Web3.
7. Fetch.ai (2023). Fetch.ai Whitepaper: Tsarin Wakilan Tattalin Arziki Masu Cin Gashin Kansu.
8. Numerai (2022). Numerai Tournament Documentation and Tokenomics.
9. Bittensor (2023). Bittensor Protocol: Internet-Scale Neural Networks.
10. Stanford Blockchain Center (2023). Research on Blockchain Scalability and AI Integration.
11. Cryptocurrency and AI Research Group (2023). Market Impact of ChatGPT on AI Tokens. SSRN Electronic Journal.
12. MIT Digital Currency Initiative (2023). Verifiable Computation for AI on Blockchain.