谷歌加大力度打击搜索结果中的AI Deepfakes

币界网Published on 2024-08-01Last updated on 2024-08-02

币界网报道:

这家技术和搜索巨头周三表示,在遏制未经授权的人工智能生成的深度伪造的最新举措中,谷歌正在采取新措施,删除和降级据报道包含非法图像的搜索网站。

AI deepfake是使用生成式AI创建的媒体,用于制作看起来真实的视频、图片或音频片段。许多这些假照片描绘了女演员斯嘉丽·约翰逊等名人、美国总统乔·拜登等政客,以及更阴险的儿童。

谷歌在一篇博客文章中表示:“多年来,人们一直可以根据我们的政策要求从搜索中删除未经同意的虚假露骨图像。”。“我们现在已经开发了使这一过程更容易的系统,帮助人们大规模解决这个问题。”

谷歌发言人进一步向Decrypt解释说,此类报告将影响网站在其搜索结果中的可见性。

发言人说:“根据这项政策,如果我们收到一个网站的大量删除请求,这将被用作向我们的排名系统发出信号,表明该网站不是一个高质量的网站——我们将把它纳入我们的排名体系,以降级该网站。”。“从广义上讲,这不是我们限制搜索内容可见性的唯一方法。”

随着谷歌的新更新,当收到删除搜索中发现的未经同意的deepfake网站的请求时,谷歌还将努力过滤包含被冒充者姓名的类似搜索结果。

发言人说:“这意味着,当你根据我们的政策从搜索中删除结果时,此外,我们将对任何包含你的名字的查询——或者可能从搜索中显示该页面的查询——所有明确的结果都将被过滤。”。“因此,并非所有显式结果都会被删除,但所有显式的结果都会在这些搜索中过滤,这会阻止它们出现在可能出现的搜索中。”

除了过滤搜索结果外,谷歌表示,它还将降级那些“因虚假露骨图像而被大量删除”的网站

谷歌表示:“这些保护措施已被证明在处理其他类型的非自愿图像方面是成功的,我们现在也为虚假露骨图像建立了同样的功能。”。“这些努力旨在让人们更加安心,特别是如果他们担心未来会出现类似的内容。”

谷歌承认,新政策的一个挑战是确保双方同意或“真实内容”,如电影中的裸体场景,不会与非法的人工智能深度伪造一起被删除。

谷歌表示:“虽然区分这些内容对搜索引擎来说是一个技术挑战,但我们正在不断改进,以更好地展示合法内容,并降低显性虚假内容的排名。”。关于CSAM,谷歌发言人表示,该公司非常重视这个问题,并专门成立了一个团队来打击这种非法内容。

发言人说:“我们有散列技术,在技术上我们有能力主动检测CSAM。”。“这是一种全行业标准,我们能够阻止它出现在搜索中。”

今年4月,谷歌与Meta、OpenAI和其他生成性人工智能开发人员一起承诺实施护栏,防止各自的人工智能模型生成儿童性虐待材料(CSAM)。

随着谷歌努力删除并使deepfake网站更难找到,音频取证公司Media Medic的首席执行官本·克莱顿等deepfake专家表示,随着技术的发展,这种威胁仍将存在。

克莱顿告诉《解密》杂志:“打击深度假货是一个不断变化的目标。”。“虽然谷歌的更新是积极的,但它需要持续的警惕和对算法的改进,以防止有害内容的传播。在这与言论自由的需求之间取得平衡是棘手的,但保护弱势群体至关重要。”

Clayton表示,虽然深度造假会影响隐私和安全,但该技术也可能对法律案件产生影响。

他说:“Deepfakes可能被用来伪造证据或误导调查,这对我们的法律客户来说是一个严重的问题。”。“deepfakes干扰司法的可能性是一个关键问题,突显了先进检测技术和媒体道德标准的重要性。”

政策制定者还采取措施打击深度假货。7月,华盛顿州民主党参议员Maria Cantwell提出了《编辑和深度伪造媒体内容来源保护和完整性法案》(COPIED),该法案呼吁采用标准化的方法对人工智能生成的内容进行水印处理。

民主党参议员克里斯·库恩斯在一份声明中说:“每个人都有权拥有和保护自己的声音和肖像,无论你是泰勒·斯威夫特还是其他任何人。”。“生成型人工智能可以被用作培养创造力的工具,但这不能以未经授权利用任何人的声音或肖像为代价。”

娱乐业领袖和科技公司庆祝谷歌对其政策的更新。

SAG-AFTRA在一份对该措施表示赞赏的声明中表示:“从工作室和主要唱片公司到工会和艺术家倡导团体,整个娱乐行业都支持《禁止伪造法案》。”。“将所有这些团体聚集在一起实现同一个紧迫目标是一个里程碑式的成就。”

“游戏结束了,人工智能骗子,”SAG-AFTRA总裁Fran Drescher补充道。“将对未经授权的数字复制品的保护作为联邦知识产权,将使我们在这个勇敢的新世界中受到保护。”

由Ryan Ozawa编辑。

Trending Cryptos

Related Reads

DRAM ETF Issuer: Samsung, SK Hynix, Micron All Surpass $1 Trillion, the AI Era of Memory Chips Has Only Just Begun

Authors: Dave Mazza, Thomas DiFazio | Source: Deep Tide TechFlow The article, written by Roundhill Investments (issuer of the DRAM ETF), responds to Morningstar's caution about investing in memory chip stocks. Morningstar warns of the sector's history of boom-bust cycles, a lack of economic moats, and potential momentum-driven overvaluation. Roundhill argues the current situation is structurally different due to AI. Key points in Roundhill's rebuttal include: * **Changed Demand & Supply Dynamics:** AI infrastructure, not consumer electronics, is now the primary growth driver for memory demand. New, strict long-term supply agreements with hyperscalers reflect the high capital intensity of advanced manufacturing. * **Existence of a Moat:** High-Bandwidth Memory (HBM), essential for AI, has extremely high manufacturing barriers. The market is dominated by Samsung, SK Hynix, and Micron, with new entrants blocked by technological complexity and long lead times for equipment like ASML's EUV machines. * **Strong Fundamental Outlook:** Analyst consensus projects the three companies will rank among the world's most profitable by 2027, with combined profits of $704 billion on over $1 trillion in revenue. Their operating margins have already reached record highs. * **Valuation Re-rating:** Despite significant stock price gains, memory stocks trade at attractive valuations (e.g., a median NTM P/E of 8.37x for the DRAM ETF) relative to projected explosive EPS growth. Roundhill suggests historical valuation frameworks may no longer apply given the new profitability paradigm. Conclusion: Roundhill contends the rally is justified by fundamentals, marking a structural shift for the memory industry into a new era of sustained, AI-driven demand against constrained supply, rather than a repeat of past cycles.

marsbit10m ago

DRAM ETF Issuer: Samsung, SK Hynix, Micron All Surpass $1 Trillion, the AI Era of Memory Chips Has Only Just Begun

marsbit10m ago

EF's Epic Reorganization: 20% Layoffs, Budget Halved, Is Ethereum Gearing Up for a Leaner Future?

The Ethereum Foundation (EF) has announced a major organizational restructuring, involving a 20% staff reduction (approx. 54 employees) and a division into functional clusters like Protocol, Access, User, Community, and Institutional layers. Co-founder Vitalik Buterin further revealed plans to cut the EF's budget by around 40% over the coming years, aiming to reduce its annual spending rate from about 15% to roughly 5% by 2030, transitioning to an endowment-driven model. This overhaul is seen as a long-overdue correction to the EF's ambiguous role. As Ethereum grew, the foundation faced persistent criticism over ETH sales, perceived lack of execution, and unclear strategy, often becoming a focal point for community frustration amid ETH's price stagnation. The reform aims to redefine the EF's boundaries, narrowing its focus to core protocol research, public goods funding, and ecosystem coordination, while offloading more applied development work to the broader market. Concurrently, ecosystem forces like the newly formed Ethlabs (founded by ex-EF researchers) and other independent groups are stepping in to fill the space, signaling a shift from a centralized model to a more distributed, collaborative ecosystem structure. The move was notably praised by Solana co-founder toly, who viewed a "leaner" EF as potentially more decisive and agile.

Odaily星球日报51m ago

EF's Epic Reorganization: 20% Layoffs, Budget Halved, Is Ethereum Gearing Up for a Leaner Future?

Odaily星球日报51m ago

Dragonfly Partner Haseeb: The Fastest-Growing Companies of the Future May All Get Stuck at 149 Employees

Dragonfly partner Haseeb explores the distorted economics of AI model pricing, drawing parallels to tax policy. He notes that startups and small teams (under 150 users) enjoy heavily subsidized, fixed-price AI subscriptions (like Claude Code), where the marginal cost of an additional token is effectively zero. This creates a powerful incentive for them to maximize token usage ("token-maxxing") and innovate aggressively with AI automation. In contrast, large enterprises (over 150 users) are forced onto "Enterprise" plans, paying per-token API fees with high (~75%) markups. This acts like a steep "tax" on AI-powered labor, disincentivizing marginal automation and experimental use, and encouraging them to retain more human workers. Haseeb argues this pricing creates a "150-person cliff," a regulatory notch similar to labor laws in France that discourage firms from growing past 50 employees. He predicts the fastest-growing future companies may deliberately cap their headcount at 149 to avoid the punitive enterprise pricing. This would foster an "AI-first" management philosophy obsessed with automation and outsourcing to stay lean. While not intentionally designed, this bifurcated pricing could become one of the most influential de facto tax policies, shaping how AI replaces labor—not through mass layoffs at big firms, but through agile, AI-native startups outcompeting them.

marsbit1h ago

Dragonfly Partner Haseeb: The Fastest-Growing Companies of the Future May All Get Stuck at 149 Employees

marsbit1h ago

How xBubble Breaks Through in the VC-Heavily-Backed OPC Economy

xBubble: Addressing the Structural Gap in the VC-Backed OPC Economy The concept of OPC (One Person Company) is evolving from a buzzword to a significant AI-driven market. While AI coding tools like Replit and Lovable have validated demand from non-technical users wanting to build applications, a key gap remains: the leap from creating a demo to running a stable, evolving business. These tools still require users to manage the development process, including technical judgments for integrations, modifications, and deployments—a major hurdle for OPCs. xBubble, by DAPPOS, tackles this by shifting from "Prompt-to-Code" to "SOP-to-Business." Instead of generating code from instructions, its core is a system of pre-organized SOPs (Standard Operating Procedures) that translate business goals—like "sell World Cup merchandise"—into complete, executable workflows. This includes generating cohesive assets, pages, payment systems, and backend logic. The platform is augmented by a network of third-party service providers who handle infrastructure (hosting, domains, payment setup), acting like "on-site service engineers." Users can pay for these services directly with xBubble credits, simplifying onboarding. This ecosystem aims to deliver not just an app, but a complete, modifiable business launch path. xBubble targets a clear OPC segment: small commercial nodes (e.g., creators, merchants) with existing products, customers, or channels, but for whom a full tech team is unjustifiable. Its potential lies in SOPs accumulating expertise from real cases, improving reliability and reducing delivery costs over time. Additionally, its native support for crypto payments caters to global or digital-native OPCs. In summary, as AI democratizes software creation, xBubble's opportunity is to prove that "SOP-to-Business" provides more immediate value for launching a real, operational business than a powerful but unstructured AI coding tool.

链捕手1h ago

How xBubble Breaks Through in the VC-Heavily-Backed OPC Economy

链捕手1h ago

Trading

Spot
Futures

Hot Articles

Discussions

Welcome to the HTX Community. Here, you can stay informed about the latest platform developments and gain access to professional market insights. Users' opinions on the price of AI (AI) are presented below.

活动图片