Dialogue with io.net: Apple chip controversy, Layer1 plan and token issuance
Author: Xiaobai Navigation coderworld
In April, which crypto projects are you most looking forward to launching?
io.net must be on the list.
As a distributed computing network in the Solana ecosystem, io.net was born with a golden key and occupiesAI+DePIN, Sonala two popular narratives, inAIAt a time when GPUs are in short supply, io.net combines 1 million GPUs to form the world's largest GPU cluster and build the "GPU Internet".
As more and more users connect GPUs to the io.net network, controversies arise, such as frequent disconnection issues and why Apple's M series chips, which do not have an advantage in AI computing power, can also connect to the network and gain revenue, and even gain higher revenue than other chips in the early stages.
With io.net ChineseCommunityTo answer some of the questions, Coderworld interviewed io.net CMO Garrison and obtained a lot of exclusive information.
The core information is organized as follows:
io.net provides flexible and low-cost computing resources. On the one hand, it is committed to providing computing power for the AI field. On the other hand, it also hopes to become a truly decentralized infrastructure in the crypto industry. After all, the crypto industry now relies heavily on centralized cloud services.
io.net plans to develop a computing-centric DeFi ecosystem around computing resources. This ecosystem will allow trading, staking, borrowing, lending, and leverage, creating a dynamic and accessible financial model based on computing resources.
The goal of io.net is not to replace AWS, and decentralized cloud services will not replace centralized cloud services, but whenCentralized services cannot provide the necessary flexibility when unexpected demand surges.io.netThe system allows for rapid deployment of resources without the need for contracts, permissions or KYC, you can immediately access the computing cluster.
Token IO is designed as a computing power currency and is planned to be regarded asSafetycollateral and develop a complete ecosystem around the mortgage function.
As the computing network expands, io.net plans to move towards building its own L1 or L2, are particularly interested in using the Solana Virtual Machine (SVM), which will provide significant flexibility and a more decentralized infrastructure, with IO asBlockchain网络的原生代币。
Exploring Apple's M1 ARM chip and others in the network was an experimental attempt, and later found that some older M-series chips did not contribute significantly to the efficiency of the network. As a result, io.net is shifting its focus to the more advanced Pro and Ultra series, exploring their uses beyond traditional AI tasks. These chips may be more suitable for low-risk operations such as model inference rather than model training, or for supporting verifiers, ZK provers, and sequencers.
io.net has always viewed the inclusion of APPLE chips as an experiment, and io.net will ensure that contributors of high-end GPUs, such as users of 4090, 8600, 4810, and A100 models, will not feel their rewards reduced due to the influx of Apple chips. Therefore, we have set up a dedicated separate incentive pool for Apple chips. As more Apple chips participate, they will only affect the allocation within their specific pool, while the rewards of other contributors will not be affected.
io.net still plans to launch its token on April 28, which is the right time to coincide with Nvidia’s earnings release.
io.net will implement some anti-cheat measures, including detecting virtual private servers (VPS) and working with Galxe to remove bots.
The following is the original interview:
Xiaobai Navigationcoderworld: Please introduce yourself and what is your journey in the crypto world like?
Garrison:
My background is in chemical engineering and I started my career working in distributed systems for the electric grid. I later joined Meta where I focused on advertising and then Avalanche where I led market growth and played a key role in brand development.
I have always been very interested in the DePIN space, looking for projects that solve real-world problems. While many DePIN projects seem more experimental or quirky, io.net caught my attention because it provides practical applications in areas where there is a huge demand but a lack of solutions. This also makes io.net different from other projects.It is designed to solve real challenges and add real value.
I have been active in the crypto space since 2018, focusing on investments and closely following market movements, and later my interests expanded to the Web3 gaming sector, where I work on developing games.
Xiaobai Navigation CoderWorld: Can you elaborate on the specific challenges that the platform aims to solve in the current decentralized environment?
Garrison:
We at io.net are solving several key problems, with the primary goal of providing a decentralized alternative to traditional computing resources, essentially exploring the feasibility of creating a decentralized AWS.
This move stems from our understanding that cloud services, although widely used, are expensive and inflexible. When users choose long-term services, they can usually enjoy lower unit prices. However, this approach often requires multi-year contract commitments, which may become impractical when there is a sudden need to scale up the service.
By introducing decentralized options,We aim to provide flexible and low-cost computing resources.We are intentionally focused on the AI space, and we believe that decentralization can significantly accelerate the development of AI by reducing costs and circumventing censorship. Given the expanding role of AI in consumer products and the regulatory restrictions that various jurisdictions may impose on AI models, a decentralized framework can promote innovation and prevent censorship.
In addition, we will also work on truly "decentralizing Web3". Currently, most of the infrastructure supporting Web3, includingBlockchainVerification nodes and the like all rely on centralized cloud services, which undermines the basic principle of decentralization.
Our solution aims to be decentralizedBlockchainValidators and decentralized applications (DApps)’s hosting infrastructure, restoring true decentralization to the Web3 ecosystem.
As the company grows, we will not only focus on decentralized cloud computing, but will also develop a DeFi ecosystem centered around computing resources, centered around computing power. We compare computing power to "digital oil," the driving force behind the new technological industrial revolution. However, there is currently no mechanism to gain financial exposure to this "commodity." io.net's vision includes building a tokenized asset that represents the value of computing power, facilitating transactions and enabling a comprehensive DeFi ecosystem. This ecosystem will allow trading, staking, borrowing, lending, and leverage, creating a dynamic and accessible financial model based on computing resources.
Xiaobai Navigation CoderWorld: Why choose to develop on Solana?
Garrison:
It’s fast, cheap, and effective.
Running model inference on Ethereum is impractical because of the high cost. Each inference requires four on-chain transactions, which is very expensive. Therefore, we decided that io.net needs to run on a platform like Solana.We are particularly interested in using SolanaVirtual Machines(SVM) since we also plan to eventually develop our ownBlockchain.
Currently, our payments and token infrastructure is hosted on Solana. However, the inference infrastructure is on Aptos. We are exploring Move and other modular Move languages on Aptos, as well as testing the capabilities of SVMs. This exploration is critical because after our token launch, as the company and network scale, the plan will shift to building our own L1 and L2。这将提供显著的灵活性和更去中心化的Blockchain和计算基础设施。
Xiaobai Navigation CoderWorld: We all know that the competition in the field of decentralized computing is quite fierce. So, what are the key differentiating features of io.net that make it stand out among traditional Web2 providers and emerging Web3 platforms?
Garrison:
In the context of Web2, our goal is not to replace AWS. Obviously, due to their reliability, high capacity, speed, maturity and dataSafetyDecentralized cloud services will not replace centralized cloud services.
However, centralized cloud services like AWS are like nuclear power in that they provide base load for companies’ computing needs when demand is predictable. These services become more cost-effective when capacity is purchased multiple years in advance.The challenge arises when unexpected surges in demand occur and centralized services cannot provide the necessary flexibility.
This is where decentralized cloud computing comes into play, providing on-demand resources to manage fluctuating demand. This is particularly beneficial for researchers, individuals, or universities that need short-term computing resources. The system allows for rapid deployment of resources and immediate access to computing clusters without contracts, permissions, or KYC.
io.net is great for several reasons. First, we are the largest decentralized computing network, with 190,000 nodes in 139 countries, providing a wide range of computing power from MacBook chips to A100s. This diversity allows us to customize clusters to each user's specific needs, ensuring that they are neither over- nor under-provisioned with the computing power they need.
In addition, our expertise in resource orchestration and networking also gives us an advantage. While we have one of the largest first-party computing networks, our advantage lies in the ability to integrate partners like Filecoin and Render. This collaboration enhances our ability to form clusters that meet customer needs, spanning a variety of applications from gaming to AI, optimizing scale and minimizing network latency.
Our ability to efficiently aggregate nodes across regions while minimizing latency solves the core challenges of decentralized computing. We offer “superclusters” for those who need high-speed, high-capacity clusters, and provide the ability to create clusters of varying sizes andSafetyThe flexibility of the specification cluster makes io.net a versatile platform that can meet a wide range of computing needs.
Essentially, io.net aims to be the "Expedia" of decentralized cloud (Note: Expedia is a well-known American travel website that provides air tickets,Xiaobai NavigationWe aggregate resources from specialties like storage, gaming, and rendering. When GPUs in these areas are idle, we make them available to AI companies.
Xiaobai Navigation CoderWorld:io.netWhat are the main functions of the tokens issued in the future?
Garrison:
正如我之前提到的，这个代币被设计为算力货币。我们的目标是使 IO 成为在 io.net 上进行交易和解锁高级功能的主要方法。此外，我们也设想，将 IO 视为Safety的抵押品，将围绕抵押开发一个生态系统，这与熟悉的 Filecoin 的代币经济学非常相似。一旦我们启动我们的区块链，IO 将作为原生代币。
An important aspect of computing power tokens is the ability to facilitate the trading of computing resources, which is part of building a dynamic DeFi ecosystem. As a commodity, computing power is something that ordinary people generally find difficult to invest in directly.cryptocurrency, we aim to make this commodity available for financialization. Tokens not only secure and operate the network, but also provide utility and enable people to access commodities that would otherwise be out of reach.
When IO becomes the native token of both the decentralized network and the blockchain, it will create scenarios we have not seen yet. Most tokens are dedicated to either securing a blockchain network or a dApp.Our goal is to build the first token that encompasses both, ensuring cryptoeconomic security and incentives for blockchains and decentralized networks.
Now, let’s talk about your actual business. GPU Increased participation of miners in the network,io.net How to ensure that participants obtain actual value and real profits?
Garrison:
We usually see tokens being inflated and distributed to validators as compensation. Validators and miners play a vital role as they provide hardware and storage in exchange for tokens.
However, especially for AI-focused decentralized networks, cryptoeconomics around computing consumption has its own unique characteristics. AI computing resources are in high demand. When launching and operating our network, we do not rely solely on token economics to compensate operators. Operators typically want to recoup the cost of investing in hardware within 18 months. In the traditional model, this involves deploying hardware in data centers and selling computing power directly to end users.
Through io.net,Operators can not onlyAICompanies generate revenue by providing services and can also earn tokens as rewards. This dual income stream enhances the ability to offer competitive rewards and payments to partners and individual contributors on the network.
In addition, we emphasize performance-based compensation. Availability and reliability are critical. Nodes that fail to provide the computing resources they promised will face penalties. We will continuously adjust the network mechanism to ensure that high-performance nodes receive larger rewards, while nodes with poor hardware or performance receive smaller rewards. This strategy prevents abuse of the system and ensures that only contributors who provide real value are rewarded.
Xiaobai Navigation CoderWorld: Among them, Apple chips arecomputing networkWhat role does it play? Garrison:
Exploring Apple’s M1 ARM chips in our network is an experimental endeavor. These chips have Neural Engines, which provide us with new opportunities to evaluate their capabilities in a cluster.
However, we found that some older M-series chips did not contribute significantly to the efficiency of the network.As a result, we are shifting our focus to the more advanced Pro and Ultra series, which offer larger onboardMemory, and is better suited to making the operation of Web3 infrastructure more robust.
We remain keenly interested in exploring different chip architectures for AI applications. Nvidia currently leads the industry with AI-optimized chips, but we believe other processors with comparable processing power and RAM have potential if equipped with the appropriate software layer.Nvidia's solution, while effective, is costly.So unlocking the potential of higher-end MacBook chips for AI could present significant value.
In addition, given the large user base of M1 chips, we are exploring their use beyond traditional AI tasks. These chips may be more suitable for low-risk operations,Such as model inference, rather than model training, or for supporting verifiers, ZK-certifiers, andSortingIdentifying appropriate workloads for these chips can unlock considerable value.
So you’re not even sure if Apple’s M-series chips can provideAIHow much computing power is needed?
Garrison:
The M-series chips are delivering computing power, and given their novelty, it's of great interest to explore their capabilities. The premise is that if we can use these chips effectively, we can unlock a lot of computing power. The suitability of these chips for AI tasks is still being studied. The M1 chip and its successors, such as the M1 Pro and M1 Ultra, have different specifications. In particular, the M3 Ultra represents a leap in power, and the cost of purchasing such a device is about six thousand to seven thousand dollars.
io.net is at the forefront because we embrace innovation and experimentationAlthough the M series chips are suitable for AI There were skeptics, but we believed it was important to explore their potential ourselves. We had the ability to cluster thousands of M-series chips and test their performance in AI applications.Success in this endeavor will make us pioneers, however, failure is equally valuable because it tells us what to do next. Innovation requires this experimental approach.
Xiaobai Navigation CoderWorld: Back to the topic of rewards, such as rewards for Apple device users, can you provide more detailed information?
Garrison:
Apple chips are categorized as a completely separate group within our system. Specifically for the airdrop, we have one set of rewards for the general pool and another set specifically for Apple chips.We have always viewed the inclusion of Apple silicon as an experiment. Therefore, our goal is to ensure that we use high-end GPU Contributors to the 4090, 8600, 4810, and A100 models won't feel their rewards reduced by the influx of Apple chips.Therefore, we set up a dedicated incentive pool for Apple chips. As more Apple chips participate, they will only affect the allocation within their specific pool, while the rewards of other contributors are not affected and come from a separate pool.
Xiaobai Navigation CoderWorld: Considering the various participants, such as large studios and individual users using laptop GPUs for mining. Which type of "miners" do you prefer? Are there any restrictions or incentives?
Garrison:
Our operations are driven by demand, and we expect AI to be one of the main use cases, in parallel with running Web3 infrastructure. We have identified the nodes best suited for these tasks, and our system dynamically adjusts to incentivize the most valuable nodes.Generally, GPUs in the 4090 series and above are expected to benefit the most and, therefore, receive the highest rewards.However, consumer-grade graphics cards still play a critical role, especially for validator workloads and certain Web3 infrastructure tasks, and will be incentivized accordingly.
Xiaobai Navigation CoderWorld: How does this restriction work? For example, if a user’s mining behavior is deemed to be ineligible, will it be directly banned?
Garrison:
We will never ban them outright, as this goes against the principles of decentralization. Instead, we adjust incentives to reduce rewards for resources that are less in demand. For example, if an Nvidia 3016 GPU is in less demand due to its limited use, its associated rewards in the network will also decrease accordingly. This does not mean that they are completely out of the game, and some users may still choose to participate using these GPUs, accepting lower contribution returns. Our approach relies on market dynamics to naturally incentivize or disincentivize certain activities based on their value to the network.
Xiaobai Navigation Coderworld: Many miners are also speculative airdrop hunters. How can we attract GPU owners to continue to provide computing power in the long term, especially in a bear market?
Garrison:
Filecoin and Bitcoin are the best examples of the resilience of the crypto market. Despite the challenges brought by the bear market, there are always people willing to mine thesecryptocurrencyMining during such times is also advantageous because competition is less intense and you may receive a larger share and rewards.
As far as market dynamics are concerned, volatility is to be expected. Even in a bear market, working with io.net will bring benefits. If you sign up at market prices through io.net, the benefits you receive may exceed other potential uses of your GPU.
Furthermore, contributors to our network benefit from a dual income model. They receive USDC for providing computing services to AI companies, and they also earn IO tokens. This ensures that contributors’ participation is worth more than what they would receive from other businesses, making it a more attractive option in less favorable market conditions.
Xiaobai Navigation Coderworld: Previously, Ahmad Shadid (CEO) mentioned that the token is expected to be launched on April 28. Has this plan changed??
Garrison:
We are currently planning to launch the token on April 28th. However, there is some suggestion that it might be better to launch on a Monday to avoid the weekend. If necessary, we can move the launch date back a day to ensure it falls on a weekday. The decision date is imminent,And it coincided with Nvidia's earnings release, which was appropriate timing given all the events happening at the same time.
Xiaobai Navigation CoderWorld: Speaking of token airdrops, are there any anti-cheating measures you can disclose?
Garrison:
We have implemented some countermeasures to ensure the integrity and authenticity of our network, and this approach has several aspects. First, we are able to identify when an actor is operating a fake node. We can also detect when an actor is running a virtual private server (VPS) because this activity is visible to us on the backend. Those who engage in such behavior will face penalties.
Our goal is to have participants contribute working, usable nodes. Bot operations have been prevalent in Galaxy, and we are working with the Galaxy team to eliminate bots. This effort is designed to ensure that real participants receive their fair share of rewards without being diluted by bots. For us, detecting these activities is relatively straightforward.
Xiaobai Navigation CoderWorld: How did Galaxy and you remove the robot BOT?
Garrison:
We implemented two methods within the Galaxy platform to ensure the authenticity of participants. Galaxy utilizes “human scoring” to evaluate user activities and profiles to measure authenticity. In addition, a programmatic approach is used to analyzewalletActivities to judgewalletIs it true.
具体来说，包括观察进入的交易，如交易费用所需的 SOL 代币，并追踪这些代币的来源，以确定钱包是由真实个人操作还是由单一实体控制网络的一部分。虽然一些机器人可能最初绕过这些措施，但我们的目标是尽可能地识别并消除它们，以最小化它们的影响。
This strategy is critical to me because it ensures that weCommunityThe real members of the project, who have supported the project for a long time, have received the rewards they deserve, rather than having their rewards diluted by BOT.
The article comes from the Internet:Dialogue with io.net: Apple chip controversy, Layer1 plan and token issuance
Pendle currently relies on the development of the LRT track, but the LRT track alone still has the potential for multiple growth. Written by: MIIX Capital *This research report was completed in early March 2024 and revised in early April 2024 1. Research report highlights 1.1 Investment logic and narrative Cryptocurrency income trading may be lucrative…
相关文章
