Frequently Asked Questions


  • What is TuringNet?

    TuringNet is a public chain that creates world’s first open and trustable Artificial Intelligence (AI) platform. Our solution sets future AI applications and development free from centralized organizations, serving the best of interests of the community that reshapes AI. We are also reinventing how the public participates in AI learning, eliminating duplicated AI model construction by centralized organizations, at the same time ensuring data privacy and security.

  • Is there a ICO?

    Short answer is no for now. Instead of collecting funds through the traditional ICO, where a significant portion of the tokens are pre-mined, then distributed to investors and community, we are taking a bold new approach. Not saying the ICO model doesn’t work well, but the true value of tokens in our ecosystem, should play a critical role to incentivize real users who help to grow our eco system in the long term. We want to ensure the massive public receives maximized benefits on our self-gover ning platform.

    Based on all these beliefs, we are introducing a new model — Community Mining Offering (CMO). In this CMO model, users can obtain TNET Tokens through accomplishing a number of simple mining tasks on TuringNet’s plat form operated by smart contract. This way, legal and financial risks are both mitigated as users won’t have to put in real money to purchase these tokens, instead they get rewards by providing valuable of work on our platform. In the earlier stage, our aim is to lower the bar for the public to participate and start mining through some light tasks. We are exploring and developing a few different ways that will allow the public to enter our platform and start mining with minimum learning curves.

  • How does TuringNet compare to others in the market?

    TuringNet’s new infrastructure design and incentivizing LBFT consensus mechanism can significantly reduces repetitive model training and fully utilizes idle computing resources, at the same ensuring data security.

    TuringNet’s solution has the following key advantages

    ● Reaching Consensus on Useful Work: To incentivizes computing devices for useful work, we innovates a novel and groundbreaking consensus protocol: LBFT protocol (TuringNet patent pending).

    ● Building Trust in Decentralization: AI models trained on TuringNet’s decentralized platform are guaranteed trustworthy.

    ● Unleashing Unlimited Computing Power for AI: Computing power capacity available for model training and inferencing is far beyond any centralized platform could ever provide.

  • When will the platform go live?

    We will release our main chain test network in 2018 Q4.

  • What’s TuringNet’s release schedule?

    2018 Q4

    ● Launch main chain test network

    ● Collaborate with initial partners for preliminary AI models training/inferencing

    ● Build and run communities with early developers and partners


    ● Launch subchain test network

    ● Build global community and expand ecosystem partners 2019Q2

    ● Launch version 1 main network with both main chain and subchains

    ● Continue to promote for global adoption with world-wide partners 2020 and after

    ● Launch version 2 main network world-wide

    ● Expand training tasks for consumer devices


  • How does TuringNet ensure scalability as AI training is a quite heavy task?

    TuringNet incorporates the next-generation multi-layer blockchain architecture, in order to meet the requirements of fast speed and large scale data for AI tasks, both in training and inference. The design idea is to keep the main chain secure and decentralized, while implementing fast transactions and extensibility to subchains or sidechains.

    TuringNet also implements multi-layer architecture with more than 10 thousand transactions per second, and in addition, supports the extensibility for AI as another key innovation. The sub-chains are responsible for training and inferencing trusted AI models, while the main chain as a public ledger guarantees the security of transactions.

  • What’s unique about the LBFT consensus?

    Compared to the current Proof-of-Work (PoW) consensus which calculates useless hashes, LBFT(Learning Byzantine Fault Tolerance) is a next-generation type of consensus, or proof of useful work. The useful work is measured by the contribution to AI DModel optimization in TuringNet.

  • Do users upload data onto the chain?

    TuringNet supports various storage form of the data and parameters.

    ● Training data are often stored off-chain, such as a separate data storage platform (IPFS 17 for example), or even locally in all or some of the trainer nodes.

    ● Testing data are the key for the consensus on the subchain, so either the entire testing data for small or medium size problems or the hash of the testing data are stored in the block. For public testing data, cross validation technique is employed for each iteration to calculate the loss. For testing data stored off-chain, such as IPFS, it is possible to use certain encryption technique to ensure only verifiers can access the testing data.

    Our Machine Learning models are designed to be collaborative and decentralized. This approach enables mobile devices to keep all the training data local on device, while working concurrently to learn a shared model, truly getting rid of the need to store data in a centralized platform. Turingnet allows complete ownership data to its sources.

  • How does TuringNet ensure data security and privacy?

    In a decentralized environment, the blockchain is secured against attacks from malicious nodes and data. The main chain of TuringNet as a public ledger guarantees the security of transactions. Privacy issues are mitigated since user data was never taken out of user devices for training purposes, but stored in local.

  • Does the platform build AI models from scratch?

    As deep learning is becoming the mainstream and is capable of solving increasing numbers of AI problems, the success of TuringNet as an open and trusted AI platform should be supported by its intriguing and world-class deep learning virtual machine, named, Graph Virtual Machine (GVM). GVM adopts graphs as AI models, one of the key components in TensorFlow. The abstraction of a graph can best describe a variety of deep learning network models, with nodes as operations and edges as data. The team is thriving to port all types of deep learning models to GVM, while optimizing the model storage and loading in a decentralized environment. Therefore, users and the community can easily migrate their TensorFlow models into TuringNet and grow TuringNet into one of the world’s largest AI platforms.

  • Does the platform charge fees or tokens for accessing AI models?

    1.  An AI user who wants to inference a model, invokes the main chain with a inferencing task start transaction. Certain amount of TN tokens are submitted as a deposit for incentives for the miners.

    2.  The main chain freezes such amount of TN token deposit, and creates a inferencing block in the subchain of related DModel for the requested inferencing task.

    3.  The subchain starts for inferencing, miners execute the task in the order of higher fee paid and package the result of inferencing in to the block. The verifier check the result data in this block and sign and broadcast the block.

    4.  The main chain pays TN token deposits to the nodes in the subchain.

    5.  The main chain writes a transaction stating that the inferencing task is finished.

    6.  and notify the user, the user could get the result from the subchain block.

  • How does the platform verify the results after each training iteration?

    The innovation of LBFT in TuringNet lies in the integration of AI task execution within every iteration of consensus. In each iteration, there is a well-defined value regarding the performance, particularly in training AI models, i.e., the loss. By voting the minimum loss in each iteration, the consensus is able to select the best performing node and confirm it with token rewards.The LBFT consensus mechanism is the hallmark of TuringNet and empowers it as world’s first blockchain with seamless integration between AI tasks and the blockchain consensus. Therefore we call it the world’s first true AI consensus.


  • How is CMO different from ICO?

    The CMO model is different from the traditional ICO mode. In this mode, the project team can not pre-mine the tokens, which means 100% tokens are by mined by the community. The project team obtains the tokens through donations, and the investor obtains a certain percentage of the tokens in proportion of the equity ratio.

    In this way, there will be a strong consensus between the project team, investors and the community, while the traditional ICO causes conflict of interest. In the CMO mode, community users will be encouraged to get tokens through mining effort. For community users, the Token is only mined through the workload. We won’t be issuing the tokens at certain price for users to purchase at the first place, and our model significantly mitigate financial and legal risks. The workload value is the basis of the token price. In order to motivate users to mine, the project team is motivated to work very hard to build community consensus and generate higher overall value.

  • How can I mine from genesis?

    Our aim is to lower the mining bar, so that everyone who’s interested in helping to advance AI industry by contributing their resources (including local data, or computational resources) will be able to participate from early on. We are designing a set of light and fun mining tasks for users. Users who pass our initial KYC will be whitelisted and informed before our alpha launch. We will provide you detailed instructions on how to participate in mining from the very beginning.

  • What can I do to mine tokens on TuringNet?

    Please subscribe to our news channels on Telegram and follow our Twitter account for more updates. We’ll make announcements on details closer to our alpha launch.

  • How does the platform ensure that I will receive a fair amount of tokens?

    Since all of TNET tokens will be mined through completing various tasks on our platform, we will reward users based on their contribution, whether you have provided data sets on your local device for our models to learn, or computing resources to train or inferencing our models, or being one of our verifier or confirmer nodes, our AI algorithms will calculate your contributions based on your role, and output a parameter from the subchain to the mainchain, to distribute rewards accordingly. All transactions are AI verifiable on our platform, thus guaranteeing transparent and fair rewards to all participants that help to advance AI models.