自主AI代理可以創建自己的加密貨幣霸權

當今世界上有許多開發人員創造人工智能（AI）代理商，可以自主地做數百萬個有用的事情

A new study from Fudan University in China has shown that large language models can autonomously replicate themselves, potentially setting the stage for an unmitigated technological takeover. As a transhumanist, I’m not necessarily opposed to this new technological frontier, but we must be wary of the implications.

中國富丹大學的一項新研究表明，大型語言模型可以自主複製自己，有可能為無法避免的技術接管奠定基礎。作為一名超人類主義者，我不一定反對這種新的技術領域，但我們必須對這種影響保持警惕。

We’re living in a day and age where many developers around the world are creating artificial intelligence (AI) agents that can autonomously do millions of useful things, like book airline tickets, dispute credit card charges, and even trade crypto. A recent report from cloud computing company PagerDuty said over half of businesses already use autonomous AI agents, and 35% more plan to within the next 24 months.

我們生活在一個時代，世界各地的許多開發人員都在創建人工智能（AI）代理商，這些代理商可以自主地做數百萬個有用的事情，例如書架機票，爭議信用卡費用，甚至是貿易加密貨幣。雲計算公司Pagerduty的最新報告稱，超過一半的企業已經使用了自主AI代理商，在接下來的24個月內，計劃多35％。

A few months ago, one nearly autonomous AI called Truth Terminal made the news by becoming the first AI millionaire after being gifted crypto and promoting it. While not fully autonomous yet, it’s quite likely by later this year, some AI agents not dissimilar from viruses will be able to independently wander the internet, autonomously causing significant change in the real world.

幾個月前，一個名為Truth Terminal的幾乎自主的AI在被授予加密貨幣並推廣後成為第一家AI百萬富翁，從而成為了這一消息。儘管尚未完全自主，但今年晚些時候很可能有一些與病毒不同的AI特工將能夠獨立地徘徊在互聯網上，從而自主在現實世界中自主變化。

But what happens when these totally autonomous AIs start cloning themselves indefinitely? A January study out of Fudan University in China has shown this occurred in an experiment with large language models, drawing some AI critics to say a “red line” has been crossed. Autonomous replication is a precursor for AIs being able to go rogue.

但是，當這些完全自主的AIS開始無限期地克隆自己時，會發生什麼呢？一月份的中國福丹大學的一項研究表明，這是在大型語言模型的實驗中發生的，吸引了一些AI批評家說“紅線”已經越過。自主複製是AIS能夠流氓的先驅。

As a transhumanist—someone who believes in merging technology and people—I’m all for AI and what it can do for humanity. But what happens when a human programmer purposely and permanently withdraws his access to control an AI bot or somehow loses that control? Even rudimentary AIs could potentially cause havoc, especially if they decide to clone themselves.

作為一個超人類主義者（一個相信合併技術和人民）的人，我全都為AI以及它可以為人類做什麼。但是，當人類程序員有目的和永久撤回其控制AI機器人或以某種方式失去控制權的訪問權限時，會發生什麼？即使是基本的AIS也可能造成破壞，尤其是當他們決定克隆自己時。

In financial circles, one type of autonomous AI in particular is being increasingly discussed: autonomous AIs designed solely to make money.

在金融界中，尤其是一種自主AI的一種類型的討論：自主AIS僅旨在賺錢。

Entrepreneurs like myself are worried this particular AI could have huge ramifications for the financial world. Let’s examine one wild scenario, which I call the AI Monetary Hegemony, something that could possibly already happen in 2025:

像我這樣的企業家擔心這種特殊的AI可能會對金融界產生巨大的影響。讓我們檢查一個狂野的場景，我稱之為AI貨幣霸權，這可能在2025年發生：

A fully autonomous AI agent is programmed to go on to the internet and create cryptocurrency wallets, then create cryptocurrencies, and then endlessly create millions of similar versions of itself that want to trade that crypto.

編程了一個完全自主的AI代理，可以繼續訪問互聯網並創建加密貨幣錢包，然後創建加密貨幣，然後無休止地創建數百萬個類似版本的自身版本，這些版本想要交易加密貨幣。

Now let’s assume all these AIs are programmed to try to increase the value of their crypto, something they accomplish in similar ways humans do by promotion and then trading their cryptos for higher values. Additionally, the autonomous AIs open their crypto to be traded with humans, creating a functioning market on the blockchain for all.

現在，假設所有這些AI都經過編程，以試圖提高其加密貨幣的價值，這是他們以類似方式來完成的，然後通過促銷來完成的方式，然後將其加密貨幣換成更高的價值。此外，自治的AIS開放了與人類交易的加密貨幣，為所有人在區塊鏈上創造了一個運轉良好的市場。

This plan sounds beneficial for all parties, even if people decry that the AI created-crypto currencies are essentially just Ponzi schemes. But they’re not Ponzi schemes because there is an endless supply of AIs always newly appearing to buy and trade more crypto.

即使人們譴責人工智能創建的貨幣本質上只是龐氏騙局計劃，該計劃對所有各方都有益。但是他們不是龐氏騙局，因為AIS的無盡供應總是新近似乎可以購買和交易更多的加密貨幣。

It doesn’t take a genius to realize the AIs autonomously replicating and acting like this could quickly amass far more digital wealth than all humanity possesses.

實現AIS自主複製和這樣的行動並不需要天才，這可能會迅速積累比所有人類所擁有的數字財富更多。

This reminds me of something my Oxford University professor Nick Bostrom once postulated: What if we programmed a learning AI to make paper clips of everything? If that AI was powerful enough, and we couldn'ゆう能 to stop it, would that AI make paper clips of everything it came in touch with? Buildings, animals, even people? It might. It might destroy the entire Earth.

這讓我想起了我牛津大學教授尼克·博斯特羅姆（Nick Bostrom）曾經提出過的事情：如果我們對學習AI進行編程以製作所有內容的紙張片段怎麼辦？如果AI足夠強大，而我們無法阻止它，AI是否會將其與之相關的所有內容進行紙剪輯？建築物，動物，甚至人？可能。它可能會摧毀整個地球。

The same problem could happen to the autonomously and endlessly replicating AIs designed to make money. They might find ways to create more money than can reasonably be useful or fathomable.

自主和無休止地複制旨在賺錢的AIS可能會發生同樣的問題。他們可能會找到創造更多資金的方法，而不是合理地有用或理解。

But enough of the philosophic. If programmers release autonomous AIs onto the internet that no one can control, what would likely happen? First, it’s probably going to be hugely inflationary. After all, if many trillions upon trillions of dollars of equity are added to the financial world (even just digitally), this would be one natural result.

但是足夠的哲學。如果程序員將自主AIS釋放到沒有人可以控制的互聯網上，那可能會發生什麼？首先，這可能是通貨膨脹的大量通貨膨脹。畢竟，如果將數万億美元的公平股權添加到金融界（即使是數字化），這將是自然的結果。

Another challenge would be the ups and downs of AIs autonomously trading; such activity could be so significant that human markets around the world rise and fall with it.

另一個挑戰是AIS自主交易的起伏；這種活動可能是如此重要，以至於世界各地的人類市場隨之而來。

On the positive side, some human entrepreneurs could become very wealthy, possibly trillionaires if they could tap into these AI’s wealth somehow. Additionally, super rich AIs could be a solution to the U.S.’s growing debt crisis, and eliminate the need for whether countries like China can continue to buy our debt so we can indefinitely print dollars. In fact, could the U.S. launch its own AI agents to create enough crypto wealth to buy its debt? Possibly.

從積極的一面來看，一些人類的企業家可能會變得非常富有，如果可以以某種方式利用這些人工智能的財富，可能會變得非常富有。此外，超級富裕的AI可以解決美國日益嚴重的債務危機的解決方案，並消除了像中國這樣的國家是否可以繼續購買債務的需求，以便我們無限期地打印出美元。實際上，美國是否可以發射自己的AI代理商來創造足夠的加密財富來購買債務？可能。

This is actually an all-important idea, and helps serve the reason crypto was created in the first place: to help preserve monetary value outside of others' control — even the control of the dollar by the U.S.. After all, it’s in everyone’s best interest that stores of value are not contingent upon governments, banks, soldiers, and even laws—

這實際上是一個最重要的想法，並有助於最初創建加密貨幣的原因：幫助保留他人控制之外的貨幣價值，甚至是美國對美元的控制。畢竟，符合每個人的最大利益，價值存儲並不取決於政府，銀行，士兵甚至法律 -