从数字稀缺到丰富:加密货币和人工智能如何互补

币界网Опубліковано о 2024-08-02Востаннє оновлено о 2024-08-02

币界网报道:

乍一看,加密货币和人工智能可能看起来像是正交技术,每种技术都建立在根本不同的原则之上,并服务于不同的功能。

然而,更深入的探索表明,这两种技术有机会平衡彼此的权衡,每种技术的独特优势可以相互补充和增强。

Balaji Srinivasan在SuperAI会议上雄辩地提出了这种互补能力的概念,激发了对这些技术如何相互作用的详细比较。

来源:IOSG Ventures(该表的灵感来自Balaji在SuperAI会议上的演讲)

加密货币采用自下而上的方式运作,源于匿名网络朋克的去中心化努力,并在十多年来通过全球众多独立实体的协调努力不断发展。相比之下,人工智能是通过由少数科技巨头主导的自上而下的方法开发的。这些公司决定了该行业的步伐和动态,进入壁垒更多地取决于资源强度,而不是技术复杂程度。

这两种技术也有不同的性质。本质上,加密货币是产生不可变结果的确定性系统,例如哈希函数的可预测性或零知识证明。这与人工智能的概率性和往往不可预测性形成鲜明对比。

同样,加密技术在验证方面表现出色,确保交易的真实性和安全性,并构建无需信任的流程和系统,而人工智能则专注于生成和创建丰富的数字内容。然而,在创造数字财富的过程中,存在着确保内容来源和防止身份盗用的挑战。

幸运的是,加密货币与数字丰富的概念相反——数字稀缺。它提供了相对成熟的工具,可以推断到人工智能技术,以确保内容来源,避免身份盗用问题。

加密货币的一个显著优势是它们能够将大量硬件和资本吸引到服务于特定目标的协调网络中。这种能力可能对人工智能特别有益,因为人工智能消耗了大量的计算能力。调动未充分利用的资源来提供更便宜的计算可以显著提高人工智能的效率。

通过将这两个技术巨头并置,我们不仅可以欣赏他们各自的贡献,还可以欣赏他们如何共同打造技术和经济的新途径。两者相互抵消,创造了一个更加整合、创新的未来。在这篇博客文章中,我们的目标是探索新兴的加密货币x人工智能行业地图,重点介绍这些技术交叉的一些新兴垂直领域。

来源:IOSG Ventures(最初由Momir于6月21日在X上发布)

计算网络

行业地图始于计算网络,该网络正试图解决GPU供应受限的挑战,并试图以不同的方式降低计算成本。值得强调的是以下几点:

    非统一GPU互操作性:这是一项非常雄心勃勃的尝试,具有很高的技术风险和不确定性,但如果成功,它将有可能产生巨大的规模和影响,使所有的计算资源都可以互换。从本质上讲,我们的想法是构建编译器和其他先决条件,这样在供应端,你可以插入任何硬件资源,在需求端,所有硬件的不均匀性都将被完全抽象出来,这样你的计算请求就可以路由到网络中的任何资源。如果这一愿景成功,它将降低CUDA软件的障碍,CUDA软件是当今人工智能开发人员完全主导的解决方案。同样,技术风险很高,许多专家对这种方法的可行性持高度怀疑态度。高性能GPU聚合:将全球最受欢迎的GPU集成到一个分布式、无需许可的网络中,而不必担心跨非均匀GPU资源的互操作性。商品消费者GPU聚合:指向聚合一些性能较差的GPU,这些GPU可能在消费者设备中可用,并且在供应端提供了最未充分利用的资源。它迎合了那些愿意为了更便宜、更长的培训过程而牺牲性能和速度的人。

训练和推理

计算网络被用于两个主要功能:训练和推理。对这些网络的需求来自Web 2.0和Web 3.0项目。在Web 3.0领域,像Bittensor这样的项目利用计算来执行模型微调。在推理方面,Web 3.0倡议强调过程的可验证性。这种关注导致了可验证推理作为一个垂直市场的出现,项目正在探索将人工智能推理整合到智能合约中的方法,同时保持去中心化的原则。

代理平台

接下来是代理平台,该图概述了此类初创公司必须解决的核心问题:

    代理互操作性以及发现和相互通信的能力代理构建集体和管理其他代理的能力人工智能代理的所有权和市场

这些功能强调了灵活和模块化系统的重要性,这些系统可以在各种区块链和人工智能应用程序之间无缝集成。人工智能代理有可能彻底改变我们与互联网的互动方式,我们相信代理将利用加密基础设施为其运营提供动力。我们设想AI代理以以下方式依赖加密基础设施:

    利用分布式爬行网络访问实时网络数据,使用加密支付渠道进行代理对代理支付,要求经济利益不仅能够在行为不端的情况下进行惩罚,而且能够提高代理的可发现性(即在可发现性过程中利用利益作为经济信号),利用加密共识来确定哪些事件应该导致削减,开源互操作性标准和代理框架,以实现构建可组合的集体,依靠不可变的数据历史来评估过去的性能,并实时选择正确的代理集体。

数据层

加密人工智能融合的一个核心组成部分是数据。数据是人工智能竞争中的战略资产,与计算一起是关键资源。然而,它往往是一个被忽视的类别,因为该行业的大部分注意力都集中在计算层上。加密原语在数据采集过程中提供价值的角度有很多,其中两个高级方向是:

    访问公共互联网数据访问围墙花园中的数据

前者是关于建立一个分布式剪贴器网络,该网络可以在互联网上爬行,并在几天内访问大量数据集,或者提供对互联网上非常具体数据的实时访问。然而,为了能够抓取互联网上的大量数据集,网络要求非常高,至少从一些有意义的工作负载开始,需要几十万个节点。幸运的是,Grass是一个由废弃节点组成的分布式网络,已经有超过2M个节点主动向网络共享互联网带宽,目的是废弃整个互联网。它显示了加密货币经济激励在吸引宝贵资源方面的巨大潜力。

虽然Grass在访问公共数据方面创造了公平的竞争环境,但仍然存在挖掘潜在数据潜力的问题——专有数据集。也就是说,由于其敏感性,仍有大量数据以隐私保护的方式保存。一些初创公司正在利用一些加密和密码学工具,使人工智能开发人员能够利用专有数据集的底层数据结构来构建和微调大型语言模型,同时保持敏感信息的私密性。

联邦学习、差分隐私、可信执行环境、完全同态加密和多方计算等技术提供了不同程度的隐私和权衡。Bagel在研究文章中总结了这些技术的概述。这些技术不仅可以在机器学习过程中保护数据隐私,还可以在计算层面实现全面的隐私保护人工智能解决方案。

数据x模型来源

数据和模型来源技术旨在建立流程,向用户保证他们正在与预期的模型和数据进行交互。此外,这些技术提供了真实性和来源的保证。以水印为例。水印是模型来源技术之一,它将签名直接嵌入到机器学习算法中,更具体地说,直接嵌入到模型权重中,这样在检索时,您就可以验证推理是否来自缩进模型。

应用

当涉及到应用程序时,设计领域是无限的。在上面的行业地图中,我们列出了一些用例,我们特别兴奋地看到,随着人工智能技术在Web 3.0领域的实施而发展。由于这些用例大多是自我描述的,我们目前不会提供额外的评论。然而,值得注意的是,人工智能和Web 3.0的交叉有可能重塑加密货币领域的许多垂直领域,因为这些新的原语为开发人员创造创新用例和优化现有用例提供了更多的自由度。

结论

加密货币和人工智能的融合展现了一个充满创新和潜力的成熟景观。通过利用每种技术的独特优势,我们可以应对各自的挑战,并开辟新的技术途径。当我们驾驭这个新兴行业时,加密货币和人工智能之间的协同作用可能会推动重塑我们未来数字体验和网络互动方式的进步。

数字稀缺与数字丰富的融合,为提高计算效率而调动未充分利用的资源,以及建立安全、隐私保护的数据实践,将定义下一个技术进化时代。

然而,至关重要的是要认识到,该行业仍处于起步阶段,目前的行业地图有可能在短时间内过时。创新的快速步伐意味着,今天的尖端解决方案可能很快就会被新的突破所超越。尽管如此,所探索的基础概念——如计算网络、代理平台和数据协议——突显了人工智能和Web 3.0交叉点的巨大可能性。

Пов'язані матеріали

Interview with NDV Founder Jason Huang: Piercing the AI Bubble and the MicroStrategy Myth, Seeking the Ultimate Edge in the Crypto Market

In a podcast with WuBlockchain, NDV founder Jason Huang discusses recent market dynamics, expressing a bearish outlook on crypto in the near term. He attributes Bitcoin's recent decline to a combination of cyclical selling pressure, the start of a US stock market correction, and liquidity tightening. A key catalyst is the emerging financial strain on MicroStrategy (MSTR). Huang explains that MSTR's model of borrowing to buy Bitcoin created a positive "flywheel" in a bull market. However, with falling BTC prices turning its stock premium into a discount, the model is now under severe stress. While MSTR only sold 32 BTC recently, the market is "front-running" the fear of its massive 80,000+ BTC holdings potentially being liquidated to meet debt obligations. He believes a true market bottom requires a major, capitulation-level event similar to the FTX collapse. Regarding investments, Huang states his fund is up over 20% this year, outperforming Bitcoin by 50-60%. The strategy involves crypto assets and commodities like oil, gold, and silver, but avoids AI stocks due to a perceived lack of trading edge. He is cautious of crowded trades in semiconductors and sees bubbles in the broader market, citing the hype around a potential SpaceX IPO. Despite short-term pessimism, Huang remains long-term bullish on one crypto innovation: stablecoins. He views them as the clearest example of a "faster, better" financial tool with significant room for global adoption. For the future, he is very bearish on Ethereum. For Bitcoin, he anticipates potential for a significant drop below $48,000 before a eventual rebound, but stresses the need to wait for a true panic-driven bottom marked by widespread despair and disinterest in the market.

marsbit44 хв тому

Interview with NDV Founder Jason Huang: Piercing the AI Bubble and the MicroStrategy Myth, Seeking the Ultimate Edge in the Crypto Market

marsbit44 хв тому

Conversation with Jason Huang, Founder of NDV: Puncturing the AI Bubble and the MicroStrategy Myth, Searching for the Ultimate Trump Card in the Crypto Market

In a podcast interview, NDV founder Jason Huang discusses the recent crypto market downturn, attributing the initial phase to typical Bitcoin cycle selling pressure, now compounded by a US stock market correction, tightening liquidity, and MicroStrategy's financial strain. He argues the market hasn't bottomed yet, noting true bear market lows often require a major, despair-inducing event like FTX's collapse. Huang details MicroStrategy's precarious position: its debt-and-equity fueled Bitcoin buying model has reversed into a negative cycle as prices fell. He interprets its sale of just 32 BTC as a signal prioritizing creditors over shareholders, sparking market "front-running" of its larger potential sell-off. A true bottom may arrive only after MicroStrategy resolves its looming debt payments, possibly via a large, private Bitcoin sale. His fund is up ~20% this year, outperforming Bitcoin by 50-60%, by shorting crypto and trading commodities like oil and gold. He avoided AI stocks despite being a heavy user, citing a lack of trading edge in the crowded semiconductor hardware trade, which he views as ripe for a significant correction. Long-term, Huang remains bullish on stablecoins as crypto's clearest, most practical innovation with high growth potential. He is very bearish on Ethereum and skeptical that Bitcoin has found its floor, suggesting $48,000 may not hold. He expects a sharp decline followed by a strong recovery within a year, but only after a major panic event leads to widespread capitulation and despair—the true hallmark of a market bottom.

链捕手50 хв тому

Conversation with Jason Huang, Founder of NDV: Puncturing the AI Bubble and the MicroStrategy Myth, Searching for the Ultimate Trump Card in the Crypto Market

链捕手50 хв тому

U.S. Stocks Trend (June 24): Korean Stock Plunge Ripples Global Chip Sector, Micron Tumbles Over 10%, Long-Term Supply Certainty Faces a 'Hard Test'

US Stock Market Trend (June 24): South Korean Market Plunge Disrupts Global Chips, Micron Drops Over 10%, Long-Term Supply Certainty Faces Hard Test On Monday, the South Korean KOSPI index plunged 10%, with SK Hynix and Samsung dropping over 12%, triggered by rumors that SK Hynix might slow its HBM4 production expansion. This shock quickly spread to the U.S. semiconductor sector. Micron plummeted 13.18% to $1,051.77, SanDisk fell 13.64%, and Marvell declined 8%. The Philadelphia Semiconductor Index closed down 7.87%, while the Nasdaq fell 2.21% to 25,587.04 points. The sell-off particularly hit memory chip stocks. Defensive sectors showed relative resilience, with gains in stocks like IBM and Johnson & Johnson. Market volatility spiked, with the VIX index jumping 12.79%. Commodities weakened, with WTI crude oil hitting a near three-month low and gold falling below $4,100. The core issue is not a challenge to AI demand itself, but a market reassessment of overly optimistic capacity expectations for memory chips, especially HBM. The rumor about SK Hynix undermined perceived certainty in the AI infrastructure cycle. Key upcoming events include Thursday's PCE inflation data, which will influence Fed rate expectations, and Micron's earnings report. The market will focus on Micron's HBM gross margins and its long-term capacity guidance. The shift indicates the AI investment cycle is moving from euphoria to rational pricing. Large institutions are questioning the sustainability of AI-related capital expenditure growth. Micron's repricing from an "AI infrastructure staple" to a more cyclical stock highlights this change. Thursday's data and earnings represent a critical juncture for assessing long-term supply certainty, which has now significantly decreased.

marsbit57 хв тому

U.S. Stocks Trend (June 24): Korean Stock Plunge Ripples Global Chip Sector, Micron Tumbles Over 10%, Long-Term Supply Certainty Faces a 'Hard Test'

marsbit57 хв тому

Торгівля

Спот
Ф'ючерси
活动图片