去中心化AI：应用场景及主要项目盘点

I believe that openness brings innovation. In recent years, artificial intelligence has developed by leaps and bounds, which has global utility and influence. As the computing power increases with the integration of resources, artificial intelligence will naturally lead to centralization, and the party with stronger computing power will gradually occupy the dominant position, which will hinder our innovation. I think decentralization and decentralization are strong competitors for pre-training and fine-tuning. The crowdsourcing resource model of crowdsourcing computing support opinions may be extended to the computing field, and idle computing resources will be aggregated into a market, which may solve the following problems, provide lower-cost computing resources for some use cases to deal with some downtime and delay failures, and use computing resources with anti-censorship characteristics to train models that may be supervised or banned in the future. Objection Crowdsourcing computing cannot achieve economies of scale. Most high-performance decentralized computing is not owned by consumers. Paradox It actually stands on the opposite side of high-performance computing. If you don't believe it, you can ask any example of infrastructure machine learning engineer's project. Decentralized reasoning runs open source model reasoning in a decentralized way. Support opinions. In some ways, open source model is getting closer and closer to closed source model and getting more and more use of most people or such centralized services to run model reasoning, thus introducing privacy and censorship problems. One solution is to decentralize or distribute supply. There is no need to decentralize reasoning, and local reasoning will become the ultimate winner. Now we are releasing a special chip that can deal with parametric model reasoning. Edge computing is our solution project example in privacy and anti-censorship. The application that supports the use of opinion agents in the chain using machine learning needs a coordination layer to conduct transactions. For agents, it is logical to use cryptocurrency to pay because it is digital technology itself. Moreover, it is obvious that agents are decentralized people who can't open bank accounts through authentication. There is no platform risk. For example, they can suddenly decide to change their plug-in architecture, which will destroy my plug-in, but without prior notice. This is a real occurrence. Agents created on the chain have no such platform risk. Opposition agents are not ready for production, and they are all toys. In addition, entities that create artificial intelligence agents can use them without encryption to pay for them. Debate on platform risk This is a cliche use case of cryptocurrency. We haven't seen it come into play yet. Why do different project sample data and model sources support the independent management and value collection of data and machine learning models? The ownership of data should belong to the users who generate the data, not the companies that collect the data. Data is the most valuable resource in the digital age, but it is monopolized by large technology companies and the highly personalized network with poor financialization performance is coming. This requires portable data and models. We will take our data and models from one application to another through the Internet, just as we let our encrypted wallets flow between different applications. The data source is a huge problem, especially the phenomenon of counterfeiting is getting more and more serious. Even Biden admits this. Blockchain architecture is probably the best solution to solve the data source puzzle. Objection, no one cares whether they have their own data or privacy. We have seen this again and again from the user's preference. Look at the registration volume. In the end, people will trust to provide their machine learning data. Let's face it. Project examples Token incentives, such as companion ideas, are very effective in starting and guiding networks and behaviors. We will see a large number of applications centered on artificial intelligence adopting this mechanism. Companion is a compelling market. We believe that this field will be trillions of dollars. In the primary market of scale, Americans spent more than hundreds of millions of dollars on pets. Companion class is pets. We have seen that companion class has achieved the market fit of products, and the average session length is more than hours. If we see an encryption incentive platform occupying market share in this field and other vertical fields of applications, we will not be surprised. Objection, this is just an extension of cryptocurrency speculation fanaticism, and it will not last forever. Have we not learned the customer acquisition cost from tokens? Training example project Machine learning operation of token incentive such as training reasoning, it is assumed that encrypted token incentive supports opinions. Encrypted incentive can be used in the whole machine learning workflow to encourage behaviors such as optimizing weight fine-tuning to judge the output of the model by human beings to further fine-tune objections. Machine learning operation is a bad use case of cryptocurrency incentive because quality is too important. Although encrypted token is good at motivating consumers' behaviors under the condition of no problem in entropy, it is of great quality and accuracy. In critical cases, they are not conducive to coordinating behavior projects. Verifiability on the chain proves which models can run effectively on the chain and insert them into the encrypted world. Model verifiability on the chain will unlock composability, which means that you can use the combination output in the field of encryption. In the future, when we have agents running doctor models to check our bodies without going to the hospital to see a doctor, we will need some methods to verify their knowledge and diagnose what is used. The verifiability of the model is like the reputation objection of intelligence. No one needs to verify what model is running. This is what we don't care about the most. We are putting the cart before the horse, but we are afraid that other models are running in the background. This is encryption technology. Zero-knowledge proof is intended to find a problem to solve and zero-knowledge proof is the consequence of hype to get a lot of venture capital. Example project. 比特币今日价格行情网_okx交易所app_永续合约_比特币怎么买卖交易_虚拟币交易所平台

注册有任何问题请添加 微信：MVIP619 拉你进入群

文章转载注明 网址：https://netpsp.com/?id=61295