Who Is Buying All the Old Books? AI, Maybe
Welcome back to "Learn English with Podcasts"! Sarah, here is a strange shopping list. Someone buys a spy novel translated into Estonian, a special edition of an old novel by Anne Brontë, and a warship magazine from 1983. All in one order.欢迎回到"Learn English with Podcasts"!Sarah,这里有一份奇怪的购物清单。有人买了一本翻译成爱沙尼亚语的间谍小说、一本安妮·勃朗特旧小说的特别版本,还有一本1983年的战舰杂志。全部在一个订单里。
That sounds like the most confusing shopping list I have ever heard. Is this another AI story? I can feel it.这听起来像是我听过最让人摸不着头脑的购物清单。这又是一个AI故事吗?我能感觉到。
You guessed it. Used bookstores in the UK, Ireland, the US, and Australia have been getting huge, strange orders from mysterious buyers. Many people now believe AI companies are behind them.你猜对了。英国、爱尔兰、美国、澳大利亚的二手书店,最近都在收到来自神秘买家的大量奇怪订单。现在很多人相信幕后是AI公司。
Wait. Why would an AI company want old paper books? They live on the internet, not in dusty bookshops.等等。AI公司为什么要旧纸质书?它们活在互联网上,不是在灰扑扑的书店里。
That is exactly the mystery. Let us start with what the bookstores saw. A UK shop called Barter Books says these orders began about three months ago. Normally, book orders follow a theme, like sports or cars. These orders have no pattern at all.这正是谜团所在。我们先从书店看到的情况说起。英国一家叫Barter Books的书店说,这种订单大约三个月前开始出现。通常,买书的订单都会围绕一个主题,比如体育或汽车。但这些订单完全没有规律。
No pattern? So one order can be a cookbook, a poetry book, and a guide to fixing old clocks?没有规律?所以一个订单里可以有一本烹饪书、一本诗集,还有一本修老钟表的手册?
Exactly. The owner told reporters that he sold hundreds of books with no clear connection to three buyers. The total value was about four thousand pounds, or about 36,500 yuan.没错。店主告诉记者,他向三名买家卖出了数百本没有明显关联的书,总价值约4000英镑,大约合3.65万元人民币。
Four thousand pounds for random old books? That is a lot of money for a pile of surprises.花4000英镑买一堆随机旧书?为一堆"惊喜"花这么多钱,可不少啊。
And it is not just one shop. In Ireland, MW Books has received huge orders since May. The topics are all over the place, from 18th-century African farming tools to biographies of race car drivers from the 1950s.而且不只一家店。在爱尔兰,MW Books从5月起就收到了大量订单。主题五花八门,从18世纪非洲的农具,到20世纪50年代赛车手的传记。
African farming tools and old race car drivers? Those books have almost nothing in common. It really does look like a machine is choosing them.非洲农具和老赛车手?这些书几乎毫无共同点。看起来真像是机器在选书。
The owner thinks so too. He says the orders are piece by piece, probably machine-driven, and have almost no theme. But here is the strangest part. An anonymous British seller says different buyers sent their books to the same address near Heathrow Airport.店主也这么想。他说这些订单零零散散,很可能是机器驱动,而且几乎没什么主题。但最奇怪的是这里。一位匿名的英国书商说,不同买家把书都寄到了希思罗机场附近的同一个地址。
Different buyers, same warehouse? So maybe they are not different buyers at all. Maybe it is one buyer using many names.不同买家,同一个仓库?所以也许他们根本不是不同的买家。也许是一个买家用了很多名字。
Right. That seller has received orders for about 6,000 books since January. The buyers pay top prices and never ask for a discount. That is very unusual for the secondhand book trade.没错。那位书商从1月起收到了大约6000本书的订单。买家出最高价,从不要求打折。这在二手书交易里非常少见。
No discounts on six thousand books? These buyers are not trying to save money. They want the books fast and quietly.6000本书还不打折?这些买家不是在省钱。他们想要快速、安静地拿到书。
And in Australia, booksellers suddenly got so many orders this month that one said it was driving them crazy. So what is going on? Reports suggest AI companies are buying the books for training data.而在澳大利亚,书商们这个月突然收到海量订单,有人说这些需求快把自己逼疯了。那么到底发生了什么?报道认为,AI公司买这些书是为了训练数据。
Training data? I thought AI companies collected text from the internet. Why switch to old paper books?训练数据?我以为AI公司是从互联网上收集文字的。为什么要改用旧纸质书?
Here is the key reason. Books published before 2022 were written before AI-generated text became common. So their words are "clean" human data, not data polluted by AI.关键原因在这里。2022年之前出版的书,是在AI生成文本普及之前写的。所以书里的文字是"干净"的人类数据,没有被AI污染过。
Ah, so old books are like fresh water from a clean river, while some internet text is like recycled water.啊,所以旧书就像干净河流里的清水,而有些网络文本就像再生水。
Great comparison. A report by 404 Media says AI companies cut off the spines of these books, scan every page, and then often destroy the physical books. That way, the data cannot go back into the market.这个比喻很好。404 Media的报道说,AI公司会切掉这些书的书脊,扫描每一页,然后通常会把纸质书销毁。这样,这些数据就不会再流回市场。
Destroy the books? Even rare old ones? That feels a little painful. Some of those books could be museum pieces.销毁书?连珍贵古籍也销毁?这让人有点心疼。有些书可能是博物馆级别的藏品。
You are right to feel that way. The report says Anthropic, the company behind Claude, destroyed millions of physical books, including some very rare editions. They call this operation Project Panama.你这么想很正常。报道说,开发Claude的Anthropic公司销毁了数百万册实体书,其中还有一些非常稀有的版本。他们管这个行动叫"巴拿马计划"。
Project Panama. It sounds like a secret spy mission, but the target is old books. Did anyone stop them?"巴拿马计划"。听起来像秘密间谍任务,但目标是旧书。有人阻止他们吗?
Not exactly. A US federal court ruled that this kind of scanning is legal. Buying physical books and scanning them destructively was considered fair use. So the books are gone, and the data stays inside the AI.并没有。美国联邦法院裁定,这种扫描是合法的。购买纸质书并加以破坏性扫描,被认定为合理使用。所以书没了,数据留在了AI里。
Hmm. Legal, but still strange to think about. A library of old knowledge becomes invisible food for a machine.嗯。合法,但想想还是觉得奇怪。一座旧知识图书馆,变成了机器看不见的食物。
That is a beautiful way to say it. These buyers are not collectors. They are not readers. They are data hunters. Every old book they buy is another bite of clean human language.这个说法太妙了。这些买家不是收藏家,也不是读者。他们是数据猎人。他们买的每一本旧书,都是干净人类语言的又一口"食物"。
Wait. So a mystery buyer takes all the weird old books... and maybe there is no strange person at all. Just an AI, trying to get smarter from the past.等等。所以一个神秘买家把所有奇怪的旧书都买走……也许根本没有什么怪人。只是一个AI,想从过去中变得更聪明。
And if you notice your local used bookshop suddenly has empty shelves, just remember. Somewhere, an AI is reading very, very fast.如果你发现附近二手书店的书架突然空了,请记住。在某个地方,一个AI正在非常、非常快地阅读。
Thanks for listening to "Learn English with Podcasts". See you next time!感谢收听"Learn English with Podcasts"。下次见!
And maybe check your bookshelf. If one of your old books disappears, it might be learning English too. Bye!顺便看看你的书架吧。如果你的某本旧书不见了,它可能也在学英语。再见!