去中心化人工智慧宣言承諾保護用戶免受人工智慧濫用

來自互聯網電腦協議背後的非營利組織 Dfinity 基金會的人工智慧研究人員以及去中心化人工智慧開發商 Onicai 的高管最近發布了去中心化人工智慧宣言。

Artificial intelligence (AI) researchers from the Dfinity Foundation, the nonprofit organization behind the Internet Computer Protocol, and executives from decentralized AI developer Onicai have recently published the Manifesto for Decentralized AI. The manifesto presents seven points to ensure that AI benefits the end-user, including self-sovereign AI that serves the user and not large institutions.

來自網路電腦協定背後的非營利組織 Dfinity 基金會的人工智慧 (AI) 研究人員以及去中心化人工智慧開發商 Onicai 的高階主管最近發布了去中心化人工智慧宣言。宣言提出了七點，以確保人工智慧有利於最終用戶，包括服務用戶而不是大型機構的自主主權人工智慧。

In an interview with Cointelegraph, Onicai CEO Patrick Friedrich stated that the centralization of AI creates the potential for abuse by powerful corporations or state actors to control and manipulate populations in a near-absolute way. The Onicai CEO told Cointelegraph:

Onicai 執行長 Patrick Friedrich 在接受 Cointelegraph 採訪時表示，人工智慧的中心化有可能被強大的公司或國家行為者濫用，以近乎絕對的方式控制和操縱人口。 Onicai 執行長告訴 Cointelegraph：

A solution to the problem of information censorship or manipulation, according to the Onicai executive, is running decentralized AI through smart contracts on permissionless networks, which are immutable, open-source and highly transparent.

Onicai 高層表示，資訊審查或操縱問題的解決方案是透過在無需許可的網路上的智能合約運行去中心化人工智慧，這些網路是不可變的、開源的和高度透明的。

This transparent set of rules would govern and constrain AI behavior and allow users to run their AI with custom parameters using local storage methods, decentralized clouds or even hybrid models — giving users true control over the entire software stack running their AI.

這套透明的規則將管理和約束人工智慧行為，並允許用戶使用本地儲存方法、去中心化雲端甚至混合模型透過自訂參數運行人工智慧——讓用戶真正控制運行人工智慧的整個軟體堆疊。

The Manifesto for Decentralized AI. Source: Dfinity

去中心化人工智慧宣言。來源：Dfinity

Stifling innovation and ignoring the niche

扼殺創新並忽視利基市場

“I do not think we come from a direction of fear, but really from a direction of potential that AI has,” Onicai chief technical officer Arjaan Buijk said.

Onicai 技術長 Arjaan Buijk 表示：“我認為我們並非來自恐懼的方向，而是真正來自人工智慧所具有的潛力的方向。”

Buijk told Cointelegraph that AI companies focus on building general-purpose AI models instead of custom-tailored or niche models. The lack of profitability means niche and unique implementations of AI models never materialize. This stifles innovation and limits the growth potential of AI.

Buijk 告訴 Cointelegraph，人工智慧公司專注於建立通用人工智慧模型，而不是客製化或利基模型。缺乏獲利能力意味著人工智慧模型的利基和獨特實施永遠不會實現。這扼殺了創新並限制了人工智慧的成長潛力。

The Artificial Superintelligence Alliance (ASI) launched the Cortex decentralized AI in November 2024 for this reason. Cortex is an enterprise-grade artificial intelligence for industrial needs that allows businesses to tailor AI models to their needs and reduce the reliance on general-purpose, centralized AI.

為此，人工超級智慧聯盟（ASI）於 2024 年 11 月推出了 Cortex 去中心化 AI。 Cortex 是一種滿足工業需求的企業級人工智慧，可讓企業根據自己的需求客製化人工智慧模型，並減少對通用、集中式人工智慧的依賴。

Michael Casey, co-founder of the Decentralized AI Society, recently told Cointelegraph that the nonprofit group is exploring alternative funding methods for AI projects as a counterbalance to venture capital-funded projects.

去中心化人工智慧協會聯合創始人 Michael Casey 最近告訴 Cointelegraph，該非營利組織正在探索人工智慧專案的替代融資方法，以平衡風險投資資助的專案。

On Nov. 11, Anthropic CEO Dario Amodei told Lex Fridman that human-level AI may arrive in 2026 or by 2027. 

11 月 11 日，Anthropic 執行長 Dario Amodei 告訴 Lex Fridman，人類等級的人工智慧可能會在 2026 年或 2027 年到來。

Amodei’s prediction that artificial general intelligence (AGI) will materialize by 2027 is years ahead of earlier predictions made by artificial intelligence researchers, who predicted the advent of AGI in the 2030s or beyond.

Amodei 預測通用人工智慧 (AGI) 將在 2027 年實現，這比人工智慧研究人員之前的預測提前了數年，後者預測 AGI 將在 2030 年代或更長時間出現。

Magazine: AI drone ‘hellscape’ plan for Taiwan, LLMs too dumb to destroy humanity: AI Eye

雜誌：台灣人工智慧無人機「地獄景觀」計劃，法學碩士太愚蠢，無法毀滅人類：AI Eye

Explore more articles like this

探索更多類似的文章

Subscribe to the Cointelegraph Research Newsletter

訂閱 Cointelegraph 研究通訊

Select the Cointelegraph newsletters to which you would like to subscribe to receive the latest news and analysis directly from our team. Delivered on Wednesdays

選擇您想要訂閱的 Cointelegraph 時事通訊，以直接從我們的團隊接收最新新聞和分析。每週三交付

By subscribing, you agree to our Terms of Services and Privacy Policy

訂閱即表示您同意我們的服務條款和隱私權政策