威尼斯AI：AI的變形

當埃里克·沃希（Erik Voorhees）宣布威尼斯AI在基地上發布的代幣時，大多數人都錯過了諷刺意味。曾經建造Shapeshift以將加密貨幣釋放的人

Erik Voorhees, the maverick behind the once-Bitcoin gambling site and later billion-dollar crypto exchange ShapeShift, is now attempting something even more ambitious: liberating AI from centralized control.

Erik Voorhees是曾經是怪異的賭博網站背後的特立獨行者，後來又是十億美元的加密交易所變形，現在正在嘗試更加雄心勃勃的東西：從集中控制中解放了AI。

When Voorhees announced Venice AI's token launch on Base on May 29, most people missed the irony, but not us. We dove behind the scenes to uncover the story of how Voorhees is now attempting to free AI from centralized control—and sparking a new chapter in the battle between centralized and decentralized tech.

當Voorhees於5月29日宣布威尼斯AI的代幣在基地上發佈時，大多數人都錯過了諷刺意味，但我們卻不是我們。我們落後於幕後，揭示了Voorhees現在如何試圖從集中控制中解放AI的故事，並在集中式和分散技術之間的戰鬥中引發了新的篇章。

Let's back up. Venice AI, launched in May 2024, isn't just another AI chatbot. Then?

讓我們備份。威尼斯AI於2024年5月推出，不僅僅是另一家AI聊天機器人。然後？

It's an entire platform built on a radical premise: what if you could access the world's most powerful AI models without surveillance, without censorship, and without leaving a trace?

這是一個以根本前提為基礎的整個平台：如果您可以在沒有監視的情況下訪問世界上最強大的AI模型，而無需審查，而不會留下痕跡，該怎麼辦？

The numbers suggest Voorhees might be onto something.

這些數字表明，Voorhees可能會成為某物。

Since launch, Venice has attracted 400,000 registered users, handles 15,000 inference requests per hour, and maintains 50,000 daily active users.

自推出以來，威尼斯吸引了400,000名註冊用戶，每小時處理15,000個推理請求，並維持50,000個活躍用戶。

What makes Venice truly different is its approach to AI access.

使威尼斯真正與眾不同的是它的AI訪問方法。

Unlike ChatGPT or Claude, which run on their own proprietary models, Venice acts more like a privacy-preserving gateway to multiple AI models.

與Chatgpt或Claude不同的是，威尼斯的行為更像是通往多種AI模型的隱私門戶的門戶。

Want to use Meta's LLaMA? No problem.

想使用Meta的美洲駝嗎？沒問題。

Prefer the newly restricted DeepSeek? Venice has you covered.

更喜歡新限制的DeepSeek嗎？威尼斯讓你覆蓋。

It's this last point that proved particularly prescient.

正是最後一點被證明是特別有先見之明的。

When DeepSeek, China's rising AI star, suddenly restricted access to Chinese phone numbers only, Venice became one of the few remaining gateways to what many consider the most advanced open-source AI model.

當中國崛起的AI明星DeepSeek突然僅限於中國電話號碼的訪問時，威尼斯成為了剩下的少數幾個門戶之一，許多人認為許多人認為最先進的開源AI車型。

The timing couldn't have been better – or more suspicious, depending on who you ask.

根據您的要求，時機再好不過了，或者更可疑。

Before we dive into the controversy that followed, we need to understand exactly how this "uncensored AI" actually works.

在我們探討隨後的爭議之前，我們需要準確了解這種“未經審查的AI”實際上是如何工作的。

Because as with most things in crypto, the devil is in the technical details.

因為與加密中的大多數事情一樣，魔鬼也在技術細節中。

To get full access to our weekly premium features (HashedIn, Wormhole, Rabbit hole and Mempool) and subscribers only posts.

為了完全訪問我們的每周高級功能（Hashedin，蟲洞，兔子洞和Mempool），以及僅訂戶的帖子。

Upgrade to paid

升級到付款

If Venice AI's promise sounds too good to be true – private, unrestricted access to top AI models without surveillance – that's because making it work required some clever engineering.

如果威尼斯AI的承諾聽起來太好了 - 無需監視而無限制地訪問AI模型的私人，不受限制的訪問 - 那是因為使它起作用需要一些巧妙的工程。

The secret sauce? A combination of blockchain technology, decentralised GPU networks, and what Venice calls "abliterated models."

秘密調味料？區塊鏈技術，分散的GPU網絡以及威尼斯稱之為“融化模型”的結合。

Let's break it down.

讓我們分解。

First, there's the privacy layer.

首先，有隱私層。

Unlike traditional AI platforms that store every conversation on their servers, Venice claims to keep nothing. Your chats stay local, encrypted on your device.

與傳統的AI平台將每個對話存儲在服務器上的對話不同，威尼斯聲稱什麼都不保留。您的聊天保持本地，並在設備上加密。

When you switch browsers or devices, your history starts fresh – because there is no central database keeping track.

當您切換瀏覽器或設備時，您的歷史記錄開始新鮮 - 因為沒有中央數據庫跟踪。

This raises an obvious question: if everything's local, how does the AI processing actually happen?

這就提出了一個明顯的問題：如果一切都是本地的，那麼AI處理實際上是如何發生的？

Venice routes your queries through a decentralised network of GPU providers.

威尼斯通過分散的GPU提供商網絡來路由您的查詢。

Each node only sees the immediate request, without any context about who you are or what you've asked before. It's like having a different person handle each sentence of your conversation, with no one seeing the full picture.

每個節點只能看到即時請求，而沒有任何有關您是誰或您以前問的內容的上下文。這就像讓另一個人處理對話的每個句子一樣，沒有人看到完整的圖片。

Then there's the "abliteration" process – Venice's approach to censorship resistance. They take open-source AI models like LLaMA or DeepSeek and modify them to remove built-in restrictions.

然後是“融合”過程 - 威尼斯對審查制度抵抗的方法。他們採用諸如Llama或DeepSeek之類的開源AI模型，並修改它們以刪除內置的限制。

Here's where critics have raised valid concerns: if you're removing AI safeguards, what prevents misuse?

批評家在這裡引起了有效的關注：如果您要刪除AI保障措施，什麼阻止了濫用？

Venice's answer is a tiered system.

威尼斯的答案是一個分層系統。

Free tier: Basic safeguards remain

免費層：基本保障措施仍然

Private tier: More flexibility, with 20 daily prompts

私人層：更靈活，每天有20個提示

Pro tier ($18/month): Full access to unrestricted models

Pro Tier（每月18美元）：完全訪問不受限制的型號

The platform offers access to an impressive array of models.

該平台提供了令人印象深刻的模型。

OpenAI's GPT variations

Openai的GPT變化

Anthropic's Claude

人類的克勞德（Claude）

Meta's LLaMA

元電話

The controversial DeepSeek

有爭議的深發現

Various specialised coding models

各種專業編碼模型

Each model can be selected based on your specific needs – a luxury not offered by most AI platforms.

可以根據您的特定需求選擇每種型號 - 大多數AI平台都不提供的奢侈品。

Perhaps the most innovative aspect is how Venice handles the economics of AI computing.

也許最具創新性的方面是威尼斯如何處理AI計算的經濟學。

This is where the VVV token comes in, and where things get really interesting.

這是VVV令牌進入的地方，事情變得非常有趣。

When Venice AI launched its VVV token on January 28, 2025, it did something unusual in crypto: it skipped the presale.

當威尼斯AI於2025年1月28日推出VVV令牌時，它在加密貨幣中做了不尋常的事情：它跳過了預售。

Instead, 50% of the total 100 million tokens went straight to the community through an airdrop.

取而代之的是，總計1億個令牌中有50％通過空投直接進入社區。

The distribution looked clean on paper.

分佈在紙上看起來很乾淨。

50% to community (airdrop)

50％的社區（空投）

35% to Venice company (partially vested)

威尼斯公司35％（部分歸屬）

10% to Venice Incentive Fund

威尼斯激勵基金10％

5% for liquidity development

流動性開發的5％

It's what happened next that got everyone's attention.

接下來發生的事情引起了每個人的注意。

The VVV token took just 90 minutes to hit a fully diluted value of $1 billion after its January 27 launch, before pushing even higher to

1月27日發布後，VVV令牌僅花了90分鐘即可達到10億美元的全部稀釋價值，然後將更高的