As ChatGPT becomes more and more powerful, more and more AI upstream and downstream companies rely on it. After ChatGPT released its latest features a few days ago, due to the further surge in users, ChatGPT experienced long-term downtime in the past two days. Even the founder of OpenAI personally apologized. The reason is actually the computing power behind ChatGPT. Resources are still far from enough, resulting in downtime events if users make too many requests in a short period of time.
If a centralized approach is used to solve the problem, Microsoft should further increase the supply of computing power to OpenAI, that is, expand the hardware scale of Microsoft's artificial intelligence cloud computing center as soon as possible and purchase more of servers and GPUs. However, whether it is from the perspective of capital investment (although ChatGPT already has paid services, it is still losing money to replace users so far) or from the perspective of the construction speed of physical computer rooms, it is difficult to support the increasing user demand.
Now, UtilityNet, a public chain project focusing on decentralized AI computing power distribution, has proposed an eye-catching solution, which is to encourage everyone in the world through blockchain incentives. Ordinary people can purchase specialized AI chips to provide computing power for this distributed AI computing power network, and thus receive UtilityNet token rewards based on the amount of computing power they contribute. Sounds like a great idea, right? But many people must ask, after so many years of blockchain development, it seems that many public chain projects are related to distributed computing, but there are no particularly successful cases, so will UtilityNet also be a flash in the pan?
This starts with UtilityNet’s new consensus mechanism-POCI (Trusted Computing). In fact, almost all distributed computing projects in the past essentially used the POW (Proof of Work) consensus mechanism. The essence of this consensus mechanism is that all equipment participating in the incentive layer (commonly known as mining), in order to obtain the corresponding tokens Incentives are all doing corresponding calculation tasks all the time. Once the calculation stops, you will lose the qualification to receive token rewards. In other words, these projects simply cannot make the computing power of the equipment participating in the incentive truly redundant, and then transfer or lease it to other customers who need computing power. In other words, the project itself has consumed all the computing power of all participating devices during the mining process, and it is impossible to sublease the computing power to others. What’s even more fatal is that the computing tasks specified by these projects are repetitive garbage computing tasks that will not progress or promote human society. Therefore, the POW consensus mechanism has been criticized for wasting the earth through ineffective calculations. Valuable energy (electricity and generation resources).
Now, UtilityNet has proposed a new POCI consensus mechanism in order to solve the problem of ineffective calculation of the POW mechanism and the problem of all computing power being lost during the mining process. Its essence is that it no longer uses mathematics to prove the computing power of the chip, but uses the security engine module inside the chip and combines the cryptography principles on the chain to allow the chip to self-certify. What this sentence means is that in the current world, behind the computing power is the chip. A mature mass-produced chip often has a recognized computing power measurement standard. Then, from the perspective of the sharing economy, as long as the authenticity and online status of the hardware chip contributed by the user online can be verified through the chain during the process of the user contributing computing power, the user can theoretically be considered to have contributed the chip to the platform. The corresponding computing power value can give users corresponding token rewards. At this time, the computing power of the chip can be redundant and ready to be rented to real computing power demanders at any time, and the computing power demanders can rent the computing power they want by going to the secondary market or purchasing relevant tokens off-site. Required computing power.
In this way, valuable AI computing power can truly be transferred to the customers who need it under the action of the incentive layer, instead of being incentivized by miners to obtain tokens. Forced to do invalid calculations to waste. If UtilityNet can realize their vision, this should be a very promising blockchain + AI project!
In terms of project attributes, UtilityNet also belongs to the recently popular Depin (Decentralized Physical Infrastructure Networks, Chinese full name: decentralized physical infrastructure network) track. At the same time, UtilityNet's token distribution mechanism is decentralized enough. 97% of the tokens will be produced by miners, which is very friendly to miners.
UtilityNet is currently still in the closed testing phase. According to official information, they plan to launch the test network of the main network around February 2024. In other words, we may be able to see it in two or three months. How is this new POCI consensus mechanism verified? We look forward to the success of UtilityNet. Perhaps in the near future, various AI companies and users around the world will no longer have to worry about insufficient computing power.
YouQuan
Aaron
Hui Xin
Joy
Alex