人工智慧與區塊鏈的融合：15 年展望

人工智慧（AI）和區塊鏈有望徹底改變世界。然而，事情需要更長的時間才能顯現出來。

Artificial Intelligence (AI) and blockchain were touted to revolutionize the world, but things have taken a bit longer to manifest. Here's how they can converge and what challenges lie ahead.

人工智慧（AI）和區塊鏈被吹捧為世界的革命，但事情需要更長的時間才能顯現出來。以下是他們如何融合以及面臨的挑戰。

Artificial Intelligence (AI) and blockchain were expected to revolutionize the world by now, but things have taken a bit longer to manifest. Both technologies have advanced significantly, but their convergence and mainstream adoption still face several challenges and opportunities.

到目前為止，人工智慧（AI）和區塊鏈預計將徹底改變世界，但事情需要更長的時間才能顯現出來。這兩種技術都取得了顯著進步，但它們的整合和主流採用仍然面臨一些挑戰和機會。

In this article, we'll explore why AI and blockchain need to converge, the specializations forming amongst Large Language Models (LLMs), and why we expect 15 more years to see commercially viable applications go mainstream.

在本文中，我們將探討為什麼人工智慧和區塊鏈需要整合、大型語言模型 (LLM) 之間形成的專業化，以及為什麼我們預計再過 15 年才能看到商業上可行的應用程式成為主流。

The evolution of AI: Specialization and cost challenges

人工智慧的演進：專業化與成本挑戰

As AI continues to leapfrog expectations, we're witnessing a trend towards specialization in LLMs. Models like Claude, developed by Anthropic, are already becoming popular among developers for technical tasks and coding assistance. Others focus on specific industries or use cases (e.g., ChatGPT for more general audiences, Gemini for copywriting, and Perplexity for general research).

隨著人工智慧不斷超越預期，我們正見證法學碩士專業化的趨勢。像 Anthropic 開發的 Claude 這樣的模型已經在技術任務和編碼協助的開發人員中流行起來。其他則專注於特定行業或用例（例如，ChatGPT 適用於更廣泛的受眾，Gemini 適用於文案寫作，Perplexity 適用於一般研究）。

This natural specialization reflects the growing demand for precision in AI applications, particularly in enterprise settings. However, this progress also comes at a cost.

這種自然的專業化反映了人工智慧應用（特別是在企業環境中）對精度不斷增長的需求。然而，這種進步也是有代價的。

Despite ongoing efforts to optimize AI models, the financial burden of using LLMs at scale remains significant. OpenAI’s GPT-4, for instance, charges $0.03 per 1K tokens for input and $0.06 per 1K tokens for output. Their o1 (‘Strawberry’) model, which focuses on reasoning, is being priced at $15 per 1 million input tokens.

儘管不斷努力優化人工智慧模型，但大規模使用法學碩士的財務負擔仍然很大。例如，OpenAI 的 GPT-4 對輸入收取每 1K 代幣 0.03 美元的費用，對輸出每 1K 代幣收取 0.06 美元的費用。他們的 o1（「草莓」）模型專注於推理，定價為每 100 萬個輸入代幣 15 美元。

These costs can quickly become prohibitive for businesses looking to integrate AI across multiple departments. For instance, a large e-commerce company might use an LLM to personalize product recommendations, generate marketing copy, and even assist customer service representatives.

對於希望跨多個部門整合人工智慧的企業來說，這些成本很快就會變得令人望而卻步。例如，一家大型電子商務公司可能會使用法學碩士來個人化產品推薦、產生行銷文案，甚至協助客戶服務代表。

While using an LLM for each task might be ideal from a performance perspective, the costs could quickly become unsustainable, especially considering that LLMs require continuous fine-tuning and maintenance.

雖然從性能角度來看，為每項任務使用法學碩士可能是理想的選擇，但成本可能很快就會變得不可持續，特別是考慮到法學碩士需要持續的微調和維護。

To address this challenge and make AI more accessible to a broader range of applications, we need to explore alternative approaches that can reduce the overall costs of deploying and using AI models at scale.

為了應對這項挑戰並使人工智慧更容易被更廣泛的應用程式所接受，我們需要探索可以降低大規模部署和使用人工智慧模型的整體成本的替代方法。

One promising solution lies in converging AI with blockchain technology, specifically Scalable Blockchain Technology (what we call “SBT”), which offers several unique advantages for AI applications.

一個有前景的解決方案是將人工智慧與區塊鏈技術融合，特別是可擴展區塊鏈技術（我們稱之為「SBT」），它為人工智慧應用提供了多種獨特的優勢。

Blockchain: A potential solution for AI’s pain points

區塊鏈：人工智慧痛點的潛在解決方案

The convergence of AI and blockchain can pave the way for a new era of decentralized intelligence, where data privacy, security, and ownership take center stage. Here's how blockchain can address some of AI's most pressing pain points:

人工智慧和區塊鏈的整合可以為去中心化智慧的新時代鋪平道路，其中資料隱私、安全和所有權將成為中心舞台。以下是區塊鏈如何解決人工智慧最迫切的痛點：

Data privacy and ownership: By integrating AI models with decentralized blockchain networks, we can create a system where data is no longer centrally controlled or owned. Instead, individuals and organizations could securely contribute their data to a collective pool, ensuring that AI models have access to a diverse and privacy-preserving dataset.

資料隱私與所有權：透過將人工智慧模型與去中心化區塊鏈網路結合，我們可以創建一個資料不再被集中控製或擁有的系統。相反，個人和組織可以安全地將其資料貢獻到集體池中，確保人工智慧模型能夠存取多樣化且保護隱私的資料集。

In this scenario, data contributors would retain ownership and control over their data, and they could choose to opt out or revoke access at any time. This approach aligns closely with the principles of Web3 and decentralized data governance, empowering individuals to participate in the data economy without sacrificing their privacy.

在這種情況下，資料貢獻者將保留對其資料的所有權和控制權，並且他們可以隨時選擇退出或撤銷存取權限。這種方法與 Web3 和去中心化資料治理的原則緊密結合，使個人能夠在不犧牲隱私的情況下參與資料經濟。

Secure and immutable data input: Another key benefit of converging AI and blockchain is the ability to ensure the integrity and immutability of data used to train and operate AI models.

安全且不可變的資料輸入：人工智慧和區塊鏈融合的另一個主要好處是能夠確保用於訓練和操作人工智慧模型的資料的完整性和不變性。

In a decentralized AI system, data would be securely recorded on the blockchain, making it virtually impossible to tamper with or manipulate. This immutable data record would serve as a single source of truth, ensuring that AI models are always operating on the most accurate and up-to-date information.

在去中心化的人工智慧系統中，資料將被安全地記錄在區塊鏈上，幾乎不可能被竄改或操縱。這種不可變的數據記錄將作為單一事實來源，確保人工智慧模型始終基於最準確和最新的資訊運行。

By combining the strengths of AI and blockchain in this way, we can create a new generation of AI models that are not only powerful and efficient but also privacy-preserving, secure, and transparent.

透過這種方式結合人工智慧和區塊鏈的優勢，我們可以創建不僅強大、高效，而且保護隱私、安全、透明的新一代人工智慧模型。

Several initiatives are already exploring the potential of blockchain for secure and privacy-preserving AI applications. For instance, the European Blockchain Services Infrastructure (EBSI) is examining how blockchain can be used to create a trusted and secure environment for deploying AI models.

多項措施已經在探索區塊鏈在安全和隱私保護人工智慧應用的潛力。例如，歐洲區塊鏈服務基礎設施（EBSI）正在研究如何使用區塊鏈為部署人工智慧模型創建可信賴且安全的環境。

Similarly, projects like Ocean Protocol are developing decentralized data marketplaces that could revolutionize how AI models access and use training data. And projects like Teranode are showcasing what's truly possible at scale—something AI systems need since they deal with infinitely larger datasets than traditional ones.

同樣，像海洋協議這樣的項目正在開發去中心化的數據市場，這可能會徹底改變人工智慧模型存取和使用訓練數據的方式。像 Teranode 這樣的專案正在展示真正的大規模可能性——這是人工智慧系統所需要的，因為它們處理的資料集比傳統資料集大得多。

Roadblocks on the path to convergence

趨同之路上的障礙

Despite the potential for synergy between AI and blockchain, several significant roadblocks stand in the way of seamless integration:

儘管人工智慧和區塊鏈之間具有協同作用的潛力，但無縫整合仍面臨一些重大障礙：

Nascent regulatory frameworks: Both AI and blockchain are still emerging technologies that are rapidly evolving. As a result, regulatory frameworks governing their use and application are still nascent and vary widely across jurisdictions.

新生的監管框架：人工智慧和區塊鏈仍然是快速發展的新興技術。因此，管理其使用和應用的監管框架仍處於新生階段，並且在不同司法管轄區之間存在很大差異。

This lack of clear and consistent regulation poses a challenge for businesses and technologists seeking to converge AI and blockchain in a legally compliant manner.

缺乏明確和一致的監管給尋求以合法合規方式融合人工智慧和區塊鏈的企業和技術人員帶來了挑戰。

For instance, some jurisdictions might have strict data privacy laws that limit the use of certain AI techniques, while

例如，一些司法管轄區可能有嚴格的資料隱私法，限制某些人工智慧技術的使用，而