丹尼·斯内尔森：小型数据库与大型AI模型 | #52

在迁移网站的过程中，网站的大量内容再次崩溃了。

That in moving the website, a bunch of the website, again, it broke down.

有很多部分我能够找到，也有很多部分找不到。

There are lots of pieces I could and couldn't find.

EPC 是早期互联网的产物，它展现了在互联网形成固定规范和类型之前，人们对于互联网可能性的各种想象。

EPC is like early internet stuff, so there are all these imaginaries of what the internet could be before we had these established mores and genres of internet production.

比如，它曾列出每位诗人的电子邮件地址，这在今天看来是隐私噩梦，但当时大概是一种正常的做法。

So for example, it had a listing of every poet's email address, which is a kind of like privacy nightmare now, but was at the time, I guess like a thing that was normal.

该网站还持续发布一系列讣告。

There's a set of like obituaries that the site was constantly publishing.

因此，考虑到对社区中已故成员或曾参与其中的人的维护与关怀，它还包含了一些如热门榜单之类的内容。

So thinking about this like maintenance or care for members of the community who had passed, or written into it, it had things like hot lists.

它还有各种页面计数、访客统计等，这些如今看来过时的东西，我认为依然非常有趣，尤其是当我们越来越生活在这种由五家公司主导的过滤气泡中，仅凭少数人就能从我们每一次点击和浏览中获利。

There were just like all these like page count, page visitor accounts, like all of these like anachronistic things that I think are still really interesting, especially as seen, increasingly we live in these filter bubbles driven by five companies and just as many people profiting off of everything we click and look at.

早期的互联网有着关于它本可以成为怎样的其他构想。

This earlier internet had other ideas of what it could have been.

因此，我认为回归这些已死的网页——这些曾对互联网可能形态抱有不同期待的网页——仍然非常有启发性。

And so I think returning to these dead web pages, these web pages that had other kinds of expectations of what the internet could be, I think is still very inspiring.

我认为，在越来越多显得封闭的数字空间中，没有出口、没有选择，我们只能——在很多方面我同意约翰·克拉里的观点，我们必须彻底摧毁互联网，除非我们毁灭整个系统，否则我们没有未来。

And I think in digital spaces that increasingly feel like foreclosed, where there is no exit and no option, and we just need to and I agree in a lot of ways with John Crary that we just need to scorch the Internet, like there's no future for us unless we just destroy the entire thing.

但我认为，这些早期的想象中仍蕴含着解放的潜力。

There's still liberatory potentials, I think, in some of these earlier imaginaries.

而且，是的，我可能还有更多关于这个话题要说，但我先不说了。

And, yeah, I probably have more to say about that, but I'll

我们会说到的。

We'll get there.

我们会

We'll

说到的。

get there.

是的。

Yeah.

嗯，因为我意识到我们已经花了相当多的时间讨论书中的主题，但还有很多内容没谈完。

Well, because I realize we've we've spent a good amount of time talking about subjects in the book, and we have much more to go.

但我确实想花一点时间，为那些可能正在听、想了解这本书是如何组织成册的人解释一下。

But I did want to take a moment for folks who are maybe listening and kind of like wanting to understand how this is organized in book format.

你能谈谈这本书的结构吗？特别是为什么你选择这样交错安排章节？

Could you talk a little bit about the structure of the book and in particular kind of highlight why you chose to kind of stagger the chapters the way you did?

因为我觉得这会影响内容的呈现方式。

Because I think that informs how, like, content comes through.

是的。

Yeah.

很好，谢谢你，杰西。

Great, thank you, Jesse.

你把话题拉回到一个很好的问题上。

Getting us back on track with a great question.

所以这本书，从很多方面来说，我很幸运能从一个鼓励我创作如此独特作品的实验性项目中走出来。

So the book, I mean, in a lot of ways I'm very lucky to have come out of an experimental program that encouraged me to make something so weird.

从那时起，我在宾夕法尼亚大学攻读研究生，之后在西北大学做博士后，过去八年我一直在这里的加州大学洛杉矶分校。

Since then, did my graduate study at University of Pennsylvania, I did a postdoc at Northwestern, I've been here at UCLA for the last eight years.

我很多工作的目标是为我的学生和未来的学者创造一个可以实验和探索的空间，突破标准的学术惯例——这些惯例固然有其价值，但也存在局限。

A lot of what I want to do is produce a space for my students and for future scholars to experiment and play, to break out of standard academic conventions, which have their merit, but also have their limitations.

构建知识的方式不止标准论文这一种。

There's other ways to build knowledge than just the standard essay.

因此，这个项目从一开始就被设计成一个包含多种元素的多媒体作品。

So this project was always meant to be a kind of multimedia object that contained many things.

我们这里桌上放了一些，你把所有东西都拿出来，真是太好了。

We've got some of them out here on the table, it's very nice of you to pull everything out.

但这本书包含了可执行文本，这是一种由Python生成的理论诗歌，让text.com集合能够自我理论化。

But the book contains, for example, executable text, which is a Python generated work of theory poetry, allowing the text.com collection to theorize itself.

它还包含一个MP3混音版、一个视频混音版，以及一本奇怪的盗版杂志。

It contains an MP3 remix, a video remix, a weird bootleg magazine.

当我终于要完成这本书并撰写索引时——写索引通常是任何书籍的最后一项工作，因为索引的作用是标示出所有主题的位置，方便你使用基于编码的搜索方式查找内容。

And then as I was trying to finally finish the book and write the index, which is always the last piece of any book as you write the index to map, well, here's where all the subjects are, here's the things you wanna find if you're using codex based search methods.

我变得对索引这种格式以及它的过时性着迷，因为索引是与纸质书籍紧密绑定的，但说实话，我使用纸质书的频率远不如我从网上找到的PDF文件。

I got really obsessed with the format and also the anachronism of the index, which is bound up to these paper bound objects, which to be honest, I don't use nearly as often as I use PDFs that I find on the internet.

我用Command F来搜索内容，而不是翻阅索引。

And I search for things using Command F, I don't turn to the index.

因此，我对这种格式产生了浓厚兴趣，并用大型语言模型写了一本完全不同的书，这本书是由它的索引生成的，而不是反过来。

So I got really interested in that format and then wrote a whole other book using a large language model to have a book that was generated by its index instead of the other way around.

我可以再多谈谈这个。

I can talk more about that.

不管怎样，抱歉绕了这么大一圈才回答你的问题。

Anyways, so I'm sorry for taking the long way around to answer your question.

抓住他最后的机会。

Take his last chance.

但这本书确实在提出论点。

But the book like so it's making arguments.

而且再次强调，我认为这些方式并不是无法相容的。

And again, this is where I'm thinking these things aren't incommensurate.

所以它有一个标准结构，本质上是一本四章的书，包含引言和尾声，还有四个插章。

So it has standard, it's got a standard, like, know, it's a four chapter book with an introduction and an epilogue in some ways, with four interludes.

这些章节的组织方式是围绕各种文件格式展开的。

And the chapters are organized to look at roughly file formats.

比如文本文件、图像文件、音频文件和视频文件。

So text files, image files, sound files, and movie files.

与这些文件并行的是，这在很大程度上像一场游戏，一种表演。

Alongside those files, and a lot of this is a kind of game, Like it's a performance.

我长期以来深受一种传统的影响，即学术本身就是一种表演性艺术。

And I've long come from a tradition where scholarship is a performative art.

我通常从宏观层面入手，试图审视整个收藏集。

I start from like a large scale, like trying to look at a whole collection.

然后我会聚焦于某一系列杂志，或一小部分音频文件。

And then I look at like one run of magazines, and I look at a small set of sound files.

最后，对某些特定影片进行细致解读，再由此扩展，与更广泛的网络建立联系。

And then finally, do some close readings of some specific films, which then explode back out to connect to the network at large.

所以在每一章中，我都在研究一种文件格式，提出一个问题，然后尝试以其他方式演绎这些问题。

So in each of the chapters, I'm looking at a file format, I'm asking a kind of question, and then I'm trying to perform those same questions otherwise.

当我研究纯文本文件时，我会写一章，进行严谨的学术工作，分析这个叫text.com的网站发生了什么——它有点像最初的海盗网站。

So when I'm looking at plain text files, I write a chapter, do my diligent scholarly work to analyze what happened to this website, text.com, which is kind of like the original pirate website.

但我想用一种更具创意的方式重新演绎它。

But then I wanted to perform it differently in a more creative way.

比如，text.com 是早期的网络艺术行动主义作品，由塞巴斯蒂安·卢特格特编辑，他深谙赛博朋克理论，里面充满了韦恩·吉布森的元素，同时也包含一些你可能想象得到的、2000年时穿着风衣的人会想读的另类内容。

And so text.com, for example, was an early piece of net art activism, edited by Sebastian Lutgert, very versed in sort of cyberpunk theories, a lot of Wayne Gibson in there, But also like alternative stuff that you might imagine, like somebody wearing a trench coat in 2000 would want to read.

它就像凯西·阿克、道格拉斯·亚当斯和居伊·德波的混合体。

So it's like Kathy Aker and Douglas Adams and Guy Debord.

它是一堆古怪的杂烩。

It's like a weird assortment of stuff.

所以，它再次抗拒了分析，对吧？

So again, it resists analytics, right?

而且它还用了四种不同的语言。

There's no and it's in like four different languages.

所以，如果你把这段内容输入到Voyant或其他任何数据分析工具中，是得不到多少分析性意义的。

So you're not going to get a lot of like analytic meaning by plugging this into something like Voyant or any other kind of data analytics tool.

但它背后有着一套完整的文本理论基础在支撑着。

But it has like a whole theoretical background of text that is working from.

所以我想尝试让text.com自己进行理论化。

So I wanted to try to let text.com theorize itself.

我在写博士论文时有一个反常的想法：那一章的所有引用都只来自这个收藏集。

I had this perverse idea when I was writing my dissertation that all the citations in that chapter would just be drawn from the collection.

这样我就能用这个收藏集来为其自身的运作进行理论阐释。

So I could like use the collection to theorize its own operations.

但结果发现这并不太有趣，而且也很困难。

That proved to be not so interesting and kind of hard.

但我心想，好吧，我仍然可以把它做成一本书。

But I was like, well, I could still do that as a book.

于是我写了一个Python脚本，提取出一些核心术语，然后将它们随机组合在一起。

So I wrote a Python script that drew out some core terms and then shuffled them together.

所以，text.com这个集合可以自行对其与文本的关系进行理论化。

So like the text.com collection could theorize its own relationship to text.

我每写一章都会采用这种两步法：先是一篇学术章节，接着是一个插曲，其中包含对本章所研究媒介本身的创意混搭。

And each chapter I do this kind of like two step process where there's an academic chapter followed by an interlude that includes a kind of creative remix within the media that the chapter is looking at itself.

就像我说的，绕了远路。

Like I said, long way around.

但这样做

But does that

说得通吗？

make sense?

太棒了。

Was amazing.

继续说可执行文本，正如我告诉过你的，我总想说成'ex text'。

Staying on executable text, which as I've told you, I always want to say ex text.

也许我记错了，他不是某种程度上为text.com创造了这个说法吗？就是text.com所从事的那些工作？

Because maybe I'm misremembering this, Wasn't didn't he kind of coin that kind of phrase as text.com, like the work that text.com was doing?

是的，他是一个我非常喜欢的词。

Yeah, so he is a great word that I love.

Textwares，T E X T W A R E Z，源自90年代和2000年代的Wares场景或演示场景。

Textwares, T E X T W A R E Z, from like the Wares scene or the demo scene from the 90s and the aughts.

这个想法我觉得非常有启发性且优美。

And that idea I found really inspiring and beautiful.

是的，这一章和创意干预都源于这个理念。

Yes, the chapter and the creative intervention, both kind of emerge out of this idea.

因此，所有作品都在与我所研究的收藏品、我所关注的具体作品进行对话。

So all of the works are in dialogue with the collections I'm looking at, the specific works I'm looking at.

在整个过程中，我努力保留住我对所分析对象的偶然性和具体性的感知。

Throughout, I try to retain the sense of contingent specificity with whatever it is that I'm trying to analyze.

但他认为，textware就像是一段有缺陷的软件。

But textware's, he argued, so in the same way that like, whereas is like a cracked bit of software.

你可以想象，当年Final Cut Pro还没基于云端的时候。

So you can imagine like Final Cut Pro back in the day when it wasn't rooted in the cloud.

我不

I don't

知道你在说什么。

know what you're talking about.

我从来没听说过

I've never heard

这个。

of this.

但你可以下载破解软件，并且有一个完整的圈子的人在破解软件，供人们在自己的设备上免费使用。

But you could download a cracked software and there's a whole scene of people cracking software for people to freely use it on their own devices.

编辑提出了一个很棒的观点，认为这些文本，比如纯文本文件，就拿居伊·德波的《景观社会》为例，它就像软件一样运作，不是在计算系统上，而是在读者的脑海中，对吧？

The editor made this great argument that the texts, these like plain text files, say like a great example would be like Guy Debord, Society of the Spectacle, that it operates as a piece of software, not on a computational system, but in the mind of the reader, right?

文本就像软件一样在你身上运作。

That the text is like operating on you in the same way that a piece of software might.

关于这一点我讲了很多，我花了整整40页来讨论这个观点。

I there's a lot that I say a lot, I spend a good 40 pages talking about that.

所以我不在这里深入展开。

So I won't get into too much depth here.

但这个想法让我深受启发，即文字确实能改变人生，这些文本在影响着我们。

But that idea was so inspiring to me that the idea that like, yes, a text can change your life, these texts operate on us.

我发现这种计算隐喻非常深刻而优美，它让我重新思考我们与数字对象相遇时所经历的那些有意义的互动。

And that computational metaphor I found like really poignant and beautiful to think again about these meaningful encounters we have with digital objects when we encounter them.

因此，每一章都试图从两个方向来解析‘Textwares’这个概念，特别是你提到这一点让我很高兴，因为这正是我通过创作试图发展出来的理念。

So each of the chapters tries to like parse in both directions and Textwares, in particular, I'm so glad that you brought that to mind, was the idea that I was trying to develop by making the creative work.

所以，我不只是把‘文本即软件’这个概念当作一个学术概念来分析，而是认真对待它，看看这个理念会催生出什么样的诗歌或理论。

So taking not just the, I guess the concept text wears and parsing it as a scholar, but taking that concept seriously and seeing what kind of poem or what kind of theory would that idea generate.

这挺有趣的。

Well, and it's funny.

我的意思是，这一切都是一体的。

It's I mean, again, all of a piece.

‘索引’这个词在某种程度上以一种更字面的方式在做这件事——是的，它发生在脑海中，但通过这种逆向推演，它也正在变成可执行的文本。

Index index is doing that work in a somewhat literal fashion, like taking it yes, it happens in the mind, but also, like, in that reverse extrapolation, it's it also is sort of becoming executable text.

多花一点时间。

Spend a little bit more time.

你稍微略过了‘index index’和我们叫它反语言的部分？

You kinda glossed over both index index and are we calling it anti language?

我们到底叫它什么？

What what what are we calling it?

Un language？

Un language?

根本没法发音。

Profoundly unpronounceable.

实际上，我的编辑强迫我删掉了等号。

Actually, my editors forced me to take out the equal signs.

我的意思是，为了语音转文字设备，删掉等号是个很好的建议。

I mean, was a great suggestion to take out the equal signs for speech to text devices.

以便于视障或低视力人群使用。

So for low vision or vision impaired folks.

比如我听过，我知道我们都听过很多由语音合成朗读的书籍。

Like I listened to it, I know we both listened to a lot of books with automated voices, reading them to us.

没有什么比听到 L 等于 A 等于 M 等于 Z 更让人难受的了，或者在这个例子中，L 不等于 A 不等于 M，如此类推。

And there's nothing worse than like having to hear L equals A equals M equals Z, or in this case, L doesn't equal A doesn't equal M and so on.

所以是的，这本书，我的意思是，这是一部非常著名的文献，至少在我们这个小圈子里，它在实验诗歌领域堪称经典。

So yeah, this book, I mean, this was a, I think, well, a very famous document, at least it's so funny to say like, famous in the teacup I live, like a very famous document in experimental poetry.

这是一本诗学期刊，它在推动和建立我们所说的语言派写作方面发挥了重要作用，而语言派大致被定义为一群围绕若干小型杂志聚集的作家。

Is a poetics journal that really kind of worked to help found and build what we call language writing, which is roughly defined, again, mostly a cluster of writers that center around a series of little magazines.

这就像二十世纪的许多文学运动一样。

So it's like a lot of twentieth century literary movements.

这是一个起源于、由并定义于其所依附的小型杂志的运动。

It's a movement that begins and is really both begins in, is carried by, and is defined by the little magazines it inhabits.

小型杂志在二十世纪先锋派、现代主义和实验性写作中的作用怎么强调都不为过，这也是为什么我觉得将 LITTLE 数据库与 LITTLE 杂志重新编码联系起来非常有用。

Like the role of little magazines in twentieth century avant garde modernist and experimental writing cannot be overstated, which is part of why I find that recoding of the LITTLE database alongside the LITTLE magazine really useful.

这里，我拿这本非常著名的经典杂志为例，它收录在 Eclipse 档案中，任何人都可以免费访问。

Here, so I take this like canonical magazine that is quite famous, it's on the Eclipse archive, anyone can access it for free.

于是我决定试着思考，为什么在我的章节中会多次提到这本杂志，因此我不会重复关于语言学各种媒介化在Eclipse档案中的论点。

And decided to like, try to think about why I write about that magazine quite a bit my chapter, so I won't rehearse the argument for languages, various remediations on the Eclipse archive.

但在每一章中，我都想做一些有趣的事情。

But with each of the chapters, wanted to do something playful.

这一章的灵感来自我在艾伦人工智能研究所发现的一件了不起的事——他们为《华盛顿邮报》做了一项出色的调查报道，构建了一个搜索栏，用于追溯——抱歉，这其实是一种寻宝式探索。

And this one came out of this remarkable thing that I found through the Allen Institute for AI did this great piece of investigative journalism for the Washington Post, where they built a search bar to return to, sorry, was a kind of fetch quest.

他们构建了一个搜索栏，你可以通过它找到哪些网站的数据被抓取用于生成C4。

They built a search bar where you could find what websites had been scraped to produce C4.

这个语料库就是‘巨型清洁语料库’，大多数大型语言模型都使用过，或几乎可以肯定在后台使用过它。

This is the corpus, the colossal clean corpus that most large language models have used or are pretty certain to use in the background.

而我发现，我所研究的那些激进而古怪的语言实践网站，比如Ubooweb、PEMTOUND、Eclipse，全都被这个语料库吸收了。

And I found all these websites I was writing about these like radical weird approaches to language, Ubooweb, PEMTOUND, Eclipse had all been absorbed by this corpus.

因此，在ChatGPT那深邃黑暗的核心深处，潜藏着一群激进的诗歌与语言派诗人——应当记住，这些诗人最著名的就是试图打破语言规范，他们成长于70、80年代，正值高理论思潮时期，深受语言学转向的影响，关注语言的人为性、建构性，思考我们口中的话语权力，认为诗歌形式的政治性具有深远且社会历史性的意义。

So somewhere in the deep, dark heart of the evil that is ChatGPT is like a radical set of poems who and language poets, it should be remembered, are most well known for trying to break linguistic norms, who came of age in the 70s and 80s, a kind of high theory moment, very much guided by the linguistic turn to think about the artifice of language, its constructiveness, to think about the police in our mouths, to think that the politics of form of poetic form are meaningful or deeply meaningful in contingent and socially historical ways.

于是我心想，太棒了，ChatGPT的某个角落里，存在着一种抵抗性的模型，一个幽灵般的影子，它生成的句子大多听起来像是八年级学生或Reddit用户写的。

So I was like, great, there's this resistant model, this like shadow, this ghost in the shell of ChatGPT somewhere that ChatGPT churns out sentences that mostly sound like they're written by, I don't know, an eighth grade English student or by a Redditor.

非常礼貌的一个。

A very polite one.

非常礼貌的一个。

A very polite one.

是的，非常礼貌，非常积极，甚至到了妄想的程度。

Yeah, very polite, very positive, delusionally so.

在它的深层内部，它其实知道语言诗。

Somehow deep within it, it knows language poetry.

所以在这个项目中，语言并不等于，我试图让它记住。

And so with this project, language does not equal, I was trying to get it to remember.

于是我使用Chateapiti重写了整本原始杂志，保留了其设计、版式和字数，但将其更新为以与大型语言模型的关系来理论化自身。

So I was using Chateapiti to rewrite the entirety of the original magazine, retaining its design, layout, and word count, but updating it to theorize itself in relationship to large language models.

所以这又是一种富有玩味的方法，但我觉得‘语言不等于语言’这个说法，正为我们提供了理解实验性写作与大型语言模型交汇的方式——即小型语言模型存在于大型语言模型之内的理念，这是单靠一篇论文无法实现的，对吧？

So again, it's a kind of playful approach, but I think like this language does not equal or not equals language is giving us ways of understanding that confluence of experimental writing and large language models, that the idea that little language models live inside the large language models that an essay just couldn't do, right?

我可以在论文中做出某些类型的论点。

I can make certain kinds of arguments in the essay.

但像这样的创作，每一行都在做不同的事情。

But with a production like this, like every line is doing something different.

每一个具体的创作决策——用词的选择、版式的安排、这些元素的转化方式——都超越了任何一种细读的范畴。

Like every discrete creative decision, the bits of diction, the bits of layout, the ways in which these things get transformed, that exceeds any kind of close reading.

而且，我认为这些并不需要被对立起来，但这种方式为我们提供了一种视角，去理解那本极具影响力的杂志，而我的文章却无法做到这一点。

And again, I don't think that they have to be put in contrast, but this opens up one view into that very influential magazine that my essay just couldn't do.

因此，在每一个这样的案例中，章节与间奏曲共同作用，揭示了这些文化对象的不同维度。

So in each of these instances, the chapters and interludes are kind of working together to open up different dimensions of these cultural objects.

所以，我记得你刚开始做‘语言不等于’这个项目的时候。

So I remember when you had first started doing language does not equal.

而很不幸的是，要问这个问题，我不得不触及令人头疼的AI话题。

And unfortunately, to ask this question, I have to get into the dreaded AI of it all.

我们来谈这个吧。

Let's go there.

所以，是的，我认为我们确实得谈这个。

So, yeah, I think I think we have to.

你是最早和我讨论如何创造性地使用ChatGPT的人之一，而那个版本的ChatGPT 3.5或者当时那个GPT模型，现在看来简直可笑地过时了。

And you were one of the first people I was talking to who was meaningfully creatively using ChatGPT, which that version of ChatGPT 3.5 or whatever GPT it was then is sort of laughably quaint now.

但即便在那时，你已经发现了ChatGPT的创作潜力。

But even then, you were identifying the creative potential of ChatGPT.

完全是另一个世界了。

Different world.

很多事情已经发生了。

Things have happened.

我想，也许我们可以从这些早期时刻谈起，聊聊你是如何开始‘语言不等于’这个项目的，以及你当时使用ChatGPT创作它的经历。

I guess maybe a way in to start the conversation is talk to me about those early moments, starting what would become language does not equal and your experience working with ChatGPT to produce that.

是的。

Yeah.

我想说的事情太多了，一下子全涌出来，我会尽量有条理地讲。

There's so many things I want to say all at once, so I'll try and keep myself linear.

那是2023年，一个完全不同的时刻。

It 2023, which is a really different moment.

我记得那时我们俩的那些对话，当时我们这些关注技术从文本发展到Transformer、再到3.5的人，突然意识到：天啊，这一刻真的改变了所有东西。

And I remember those conversations that we both had at that time, whereas those of us who were just tracking the development of the technology from text to transformer through to I think it was 3.5, this point where it's like, oh my, it was like that moment, was like, oh my God, this changes everything.

而我们已经见证了这一切的发生。

And we've seen that play out.

我现在肯定不会使用ChatGPT。

I certainly wouldn't use ChatGPT now.

我仍在加州大学洛杉矶分校教授一门名为AlgoLit的课程，这是一门通过AI来批判AI系统的批判性人工智能课程。

Still teach a class at UCLA called AlgoLit, which is a critical AI course that uses AI to try to develop critiques of AI systems.

在这门课上，我完全不使用ChatGPT。

And in that class, I don't use ChatGPT at all.

我让学生们使用Mistral和其他本地模型，这些小型模型注重数据隐私。

I have my students use Mistral, other local models, small models that focus on data privacy.

当然，这发生在加沙战争爆发之前，也早于过去两三年这些技术被用于诸多其他方面的诸多发展。

Of course, is before the war on Gaza, before a lot of the developments we've seen these technologies put toward in the last two to three years.

但即便在那时，我们也已经非常清楚，这是一项具有变革性、深刻且多维度影响的技术。

But even then, it was very clear that this was a transformative technology and transformative and really profound and manifold ways.

在这份清单的最顶端，是关于获取知识的基本方式的思考。

At the very top of that list is just thinking about basic modes of access to knowledge.

我想回到搜索栏这个话题，对吧？

And I was thinking to go back to the search bar, right?

所以，如果我们过去二三十年一直使用搜索栏来寻找知识，现在这种情况越来越明显——当你收到别人分享的链接时，链接后面总是附带一个ChatGPT的URL标记。

So if we've been using, say for the last two or three decades, bars as our way to find knowledge, increasingly, and you'll see this when someone shares a link with you now, it always has the appended ChatGPT URL marker.

我一直在追踪当我学生发给我链接、朋友发给我链接时的情况。

I've been tracking that when my students send me links, when friends send me links.

以前你可能会看到来自Facebook的追踪标记，但现在你看到的是这些细微的残留信号，表明这些知识是从哪里获取的。

It used to be maybe you'd get the tracker from Facebook, but you get these little residual signals that this is where this knowledge was found.

它运作的方式，它对知识的转化方式，我认为是前所未有的，而且显然正在朝着更糟糕的方向发展。

And the way that it operates, way that it transforms our knowledge, I think is unparalleled and obviously unfolding, mostly for the worse.

再次明确一下，几乎完全是朝着更糟糕的方向发展。

Again, to be very clear, almost entirely for the worse.

话虽如此，如果这项技术如此具有变革性，它正在改变我们与知识的关系、与文化记忆的关系，而这一切当然都是建立在过去二三十年我们一直在观察和研究的互联网基础之上的，那么我认为，它太重要了，不能忽视，也太重要了，不能不去尝试理解——不仅仅是它的经济和政治层面，这些问题固然重要，但还有ChatGPT的韵律、它的诗学：这台机器究竟在说些什么？

That said, right, like if this technology is so transformative, it is changing our relationship to knowledge, our relationship to cultural memory, and of course, it's all built out of the very internet that we've been looking at and studying for the last two decades, then I think it's too important to ignore and too important not to try to understand not just the economics of, the politics of, those questions are essential, But the prosody of ChatGPT, the poetics, like what kinds of things does this machine say?

它是如何说这些的？

How does it say them?

比如，这是一堂关于形式重要的诗歌课，形式和内容是交织在一起的。

Like, this is a poetry lesson that form matters, that form and content are intertwined.

因此，在这个项目初期，我就在思考：如何利用人工智能系统来揭示该系统的一些韵律特征？

And so with this project, very early on, I was trying to think, how can you use an AI system to try to surface some of the prosodic features of that system?

这种韵律不仅对那些致力于提示工程的用户很重要，即寻找各种方法让GPT或任何大型语言模型产生你想要的输出。

And that prosody matters not just for users who are coming up with things like prompt craft, like different ways to get GPT or any large language model to give you the kind of output you're looking for.

而且这些输出还与我们对自己和周围世界的理解形成反馈关系。

But how those outputs have a feedback relationship to our own understandings of ourselves of the world around us.

我们已经看到过有关ChatGPT引发妄想和精神病的报告。

We've seen reports of like chat GPT induced delusions and psychosis.

这些是真实的影响，我认为我们必须深入其中，不夸张地说，去理解它们的工作机制，以便更好地应对它们。

These are real effects that I think we have to actually get our hands dirty with, for lack of a better word, to try to understand how they work so that we can better fight against them.

我们曾讨论过，在当今这个历史时刻，各自在教学中所面临的那种并非犹豫不决、而是充满道德困境的复杂感受。

We've talked about our respective not hesitancy, but sort of that really sort of morally fraught complex feeling of teaching in this vein at this moment in history.

而我们共同的立场是，你必须理解它们，才能有效地批判、颠覆或介入等等。

And where we the position we share is you have to understand them to effectively critique, subvert, intervene, etcetera.

而且，你知道，我们可以环顾四周，看到这些技术的制造方式反映出缺乏人文关怀。

And, you know, we can we can look around and see that the way that these things were made reflect a lack of a humanist impulse.

因此，需要更多人文学科的人介入其中。

And so therefore, there need to be more humanities people intervening.

所以，如果我们运用传递性逻辑，就意味着我们需要更多人文学科的人至少具备对它们工作原理的基本理解。

And so if we, you know, run the transitive property, that means we need more humanities people to at least have a basic literacy with how they work.

关于你所进行的这项工作，你是否发现语言诗是一种理想的媒介，能帮助你获得这些洞见？

In terms of that effort you were undertaking, was did you find that language poetry was an ideal vehicle to get into some of those revelations?

这让人感到沮丧吗？

Was it frustrating?

也许我们不妨稍微回溯一下，因为你之前在对话中提到过，但对于那些完全不了解语言诗的人来说，你能否简要描述一下什么是语言诗，然后再谈谈它是否具有生成性？

Like and and maybe let's do a little rewind too because you you you described it earlier in the conversation, but maybe for folks who are, like, coming totally green to language poetry, if you could sketch what language poetry is and then move forward into whether or not it was generative.

它是如何运作的。

How it works.

是的，我觉得这非常有帮助。

Yeah, I think that's really helpful.

所以，语言诗歌或语言写作运动是一系列非常多元化的策略。

So, again, like language poetry or language writing movement is a really heterogeneous set of strategies.

但就它们的共性而言，我认为它们围绕着抵制文学创作中某些规范性假设而凝聚在一起。

But insofar as they do cohere, I think that they cohere around resisting certain normative presumptions about literary production.

因此，‘新句子’可能是这一运动最著名的成果之一，主要由罗恩·西利曼与许多其他人共同发展而来，它会把正常的句子拆解开来，对吧？

So the new sentence is maybe one famous output of this developed mostly by Ron Silliman in dialogue with many others that takes like normal sentences and breaks them down, right?

所以，这涉及大量语言的重组和拆解。

So it's a lot of like shuffling language, breaking language down.

我在研究生阶段的导师是查尔斯·伯恩斯坦，他以提出‘吸收的虚饰性’而闻名。

My mentor in grad school was Charles Bernstein, who famously wrote about the artifice of absorption.

我认为这是思考大型语言模型和规范形式的一个有用切入点。

And I think this is a useful entry point to thinking about large language models, and normative form.

因此，小说或其他文学形式如散文，都是建立在‘吸收’这一理念之上的，对吧？

So a novel or other literary modes like the essay are built on an idea of absorption, right?

如果你能透过文本看到背后的思想世界，那你就会完全沉浸在这段文本中。

Like if you can look through the text into the world of ideas, then you're sort of like fully absorbed in that text.

任何读小说读得入迷、突然发现两小时已经飞逝的人，都能体会到这种感觉。

Anybody who's gotten deep into a novel and two hours have suddenly flown by have a sense of what this feels like.

语言诗歌总是致力于抵抗这种沉浸感，对吧？

Language poetry is always about resisting that, right?

它让你停留在语言的表面，把语言当作一种物质载体，一种具有形式特征、阻止你进入沉浸状态的东西。

So like, keeping you on the surface of language as a material substrate, as something that has formal properties that doesn't allow you to jump into these absorptive states.

我觉得这在很多方面都非常有用。

I think that's really useful in a lot of ways.

正如我们所见，聊天机器人非常擅长生成沉浸式的叙事空间。

Chatty bitty as we've seen, is really good at producing absorptive narrative spaces.

我记忆中最早的文本转换器实验之一，是关于考古学家发现了一只独角兽，并且非常逼真地给我们讲了一个奇幻科幻故事。

Some of the earliest texta transformer experiments I remember were about like, was it archaeologists discovering a unicorn and it like convincingly giving us this kind of like fantasy sci fi story.

我经常看到这种情况，我的许多同事也是如此。

I see it all the time, as do many of my colleagues.

ChatGPT 非常擅长撰写大学研讨会论文。

ChatGPT is really good at writing the college seminar essay.

它真的很擅长。

It's really good.

这是因为这种文体非常规范、标准，并且有一套非常严格的体裁特征，它能够非常逼真地模仿。

And it's good because that form is so normative and so standard and has a very rigid set of genre markers that it can mimic quite convincingly.

因此，鉴于像 ChatGPT 这样的系统生成的这些具有沉浸感的文本，语言诗歌和语言诗学有助于打破这种系统，使其崩溃。

So given the normative outputs of something like ChatGPT that produce these kinds of absorptive texts, language poetry and language poetics are useful to try to break that system, to let it break down.

强迫它尝试写一个新句子实际上非常困难。

Forcing it to try to write a new sentence is actually quite hard.

你需要进行大量提示，才能让它生成这类输出。

You have to do a lot of prompting to get it to produce those kinds of outputs.

它对提示的抗拒总是能让我学到一些原本无法得知的东西。

It's resistance to that prompting always teaches me something about it that I wouldn't know otherwise.

我认为，我所见过的最令人鼓舞的 AI 艺术和文学作品，通常都是围绕着创造新颖的诗歌或美学策略，以让这些黑箱揭示出一些你从外部无法看到的内在特性。

I think that's, you know, the most inspiring bits of AI art and literature that I've seen that use these generative systems are usually geared around coming up with novel, poetic or aesthetic strategies to get these black boxes to reveal something about themselves that you couldn't see from the outside.

而且，你知道，我认为这种情况在你读到某些从未接触过AI模型的人写的深度评论时最为明显，比如有人抱怨AI系统并不怎么样。

And, you know, I think this is never more the case than when you read one of these think pieces by somebody who's clearly never worked with an AI model, like somebody complaining that it like it well, you know, AI systems aren't great.

听好了，我让它用这种风格写个故事，但它失败了。

Look, I told it to write a story in this style, and it failed.

这其实是对AI工作原理以及提示创作艺术和韵律的误解。

Well, that's like a failure of understanding how it works and the art of poetics and prosody of PromptCraft.

但这些黑箱内部也隐藏着许多可怕的东西。

But these black boxes hide a lot of horrors inside of them as well.

因此，我认为最好的艺术和文学作品通常都是揭示诸如偏见之类的问题，比如性别、种族或能力方面的偏见，这些方式能让这些系统暴露其内部原本隐藏的内容。

And so I think like the best works of art and literature usually are surfacing things like bias, for example, bias around gender or race, or ability, that there's all of these various ways of getting these systems to reveal what's otherwise hidden within them.

在很多方面，这个过程正体现了我试图用整个小数据库所做的事情——让那些原本看不见的结构逻辑显现出来。

And so in a lot of ways, process mirrors what I'm trying to do with the little database as a whole to get these otherwise invisible structuring logics to reveal themselves.

这就像一场游戏，通过制造它们无法复现的格式，迫使它们崩溃，从而诱使它们暴露自己。

And it's like a game of trying to trick them into revealing themselves by making them break down, by by getting them to produce a form that they're not trained to reproduce.

当你在做这些事的时候，每当有那种仿佛成功了的时刻，你当时有什么样的感受呢？

And as you were doing it, what was the I don't know, what was feeling you were having when you would have these moments that felt like you were achieving that?

这是一种愉悦与恐惧的结合。

It's like a combination of delight and horror.

就像感到兴奋一样。

It's like being thrilled.

就像是，哦，我找到了这个东西。

It's like, oh, like I found this thing.

但通常你会想，天啊，我真希望没发现这个东西。

And usually it's like, oh my god, I wish I hadn't found this thing.

它们总是相伴而生。

Like they go hand in hand.

还有就是，从2023年到2025年，它们变得如此娴熟、如此迅速，这本身也令人恐惧。

And then there's also like the horror of like how adept and how quickly they've changed, From 2023 to 2025.

我的意思是，我们已经从威尔·史密斯吃意大利面这种标准例子，进展到每天都能产出高保真科幻电影，这些电影几乎让人无法分辨真假。

I mean, we've moved from, I think that standard example is Will Smith eating spaghetti to high fidelity sci fi films being produced on a daily basis that are really difficult to even recognize.

我觉得这周是兔子在蹦床上跳跃。

I think this week it was bunnies jumping on a trampoline.

每周都有新的进展。

There's a new one every single week.

当我授课时，必须不断跟进，比如这些大型语言模型目前被应用于哪些主流形式。

When I teach my classes, have to just keep up with, okay, here are both the popular forms to which these large language models are being put to use.

同时，我还想介绍一群批判性人工智能学者、艺术家和作家，他们试图利用这些系统揭示隐藏的政治结构、数据集中的漏洞，以及这些系统核心中各种偏见的组织逻辑。

And and here are, you know, I think a a cadre of of critical AI scholars, artists, and writers who are trying to use these systems to make them reveal the hidden political structuring, the lapses in the data sets, the various kind of biased organizational logics that they have somewhere in their core.

由于这些模型规模庞大，这又回到了规模的问题上——因为它们庞大到难以理解，因此具有抗拒性，就像镜子的另一面。

And because these models are so large, and this comes back to questions of scale, because they're so incomprehensibly large, they resist and it's like the other side of the mirror.

它们也抗拒分析，对吧？

They also resist analytics, right?

众所周知，就连数据科学家也很难理解无监督模型在底层究竟在做什么。

Like the commonly known fact that even data scientists have a hard time understanding what unsupervised models are actually doing underneath the hood.

因此，我认为这些微小的干预、小型的输出，以及对大型语言模型和其他生成系统的实验，能够揭示出 otherwise 完全无法察觉的东西。

And so I think these little interventions, these small outputs, these experiments with large language models and other types of generative systems can reveal things that would just be impossible to see otherwise.

在这里，我认为不仅诗歌本身有其价值，那些试图解构语言的实验性诗歌形式也同样重要。

And here, think there's a real use to not just poetry in general, but like experimental forms of poetry that try to break language down.

再次对抗这些规范性形式，以揭示你无法获取的新类型的知识。

Again, fighting against these like normative forms, as a way to reveal new types of knowledge that you couldn't access.

我非常喜欢这一点。

I love that.

作为一个花大量时间研究的人——我不是说我们主要在聊ChattyBT，但你提到了Mistral，你已经尝试过很多其他大语言模型、视频生成器和图像生成器。

As somebody who spends a lot of time, not I mean, we've been talking about chattybt largely, but you mentioned Mistral, You've experimented with with a bunch of other LLMs and video generators and image generators.

是的。

Yeah.

我的意思是，这简直是一场信息洪流。

I mean, it's a deluge.

从我们今天所处的2025年中期来看，有哪些事情让你特别关注，或者让你感到担忧？

From where we're sitting today in mid ish twenty twenty five, what are the things that are that have your attention from a like, that you're concerned about?

我都不知道该从哪里开始说这个清单了。

I don't even know where to start with that list.

我不是想笑，但我对所有这些都感到担忧。

I don't mean to laugh, but I all of it.

你知道吗？

You know?

比如，从加沙战争说起，还有这些系统所造成的算法伤害、算法暴力。

Like, you know, starting with the war on Gaza, like, I you know, and and starting from, like, the real algorithmic harms, algorithmic violence that these systems are producing.

所以我认为，这些研究有着紧迫性。

I So think, like, there's there's an urgency to these these studies.

这种危害从当下正在发生的实际破坏、毁灭和种族灭绝，一直延伸到那些日常琐事——把它们放在同一句话里感觉令人恐惧，真的成了这种恐怖的一部分。

And and so, like, it scales from, like, actual harm, decimation, and genocide that is ongoing in the present to, like, these, like, quotidian, have to put them in the same sentence, putting them in the same sentence feels horrifying, really as part of this horror.

我不知道，我在这里的同事们正为学生用ChatGPT抄袭论文而焦虑。

I don't know, my colleagues here stressing out about their students having plagiarized essays through ChatGPT.

大型语言模型已经让一些学术界赖以支撑的通用规范变得无关紧要，或者至少不再那么相关，但它们必须改变。

And the ways in which large language models have rendered some of these like, load bearing generic conventions of academe irrelevant or not necessarily relevant, but they have to they have to change.

如果写论文从来都不是关于论文里实际写了什么，而是关于一个人为了培养对世界的批判性理解所付出的脑力劳动和努力。

If the writing of an essay was never about what was actually written in the essay, it was about the mental gymnastics and work and labor that one has to do to cultivate a critical understanding of the world.

如果这种文体只需一个提示就能自动生成，那我们就必须找到其他方式来发展我们对周围世界的批判性洞察。

If that genre can just be output by a prompt, then there has to be other ways that we can develop our critical insight on the world around us.

因此，我正在进行的许多实验，以及我所投入的艺术家和作家群体，都在寻找新的策略和方法，以培养我们迫切需要的批判性视角，来对抗这些系统所带来的恐怖。

And so a lot of the experiments I'm doing and a lot of community of artists and writers that I'm invested in are, I think, looking for new strategies and new ways to develop the critical perspectives that we so desperately need to fight the horrors of these systems.

我认为要做到这一点，就必须找到一种与它们对话的不同方式。

And I think to do that, one has to find out how to converse with them in a different way.

我想接着你刚才提到的规模问题说一下，我想起了那个梗：一个小多米诺骨牌，最终连锁反应扩大到像电线杆那么大。

I want to follow-up on, I mean, when you did that sort of scale thing, was thinking of the meme where there's the little domino and it scales up to like, know, pylon sized.

我想同时触及这两个方面。

I kind of want to hit both.

所以我把这两个问题都抛给你，然后由你来决定我们从哪个角度切入。

And so I'm going to throw both at you and then player's choice, how we how we sort of enter it.

一方面，我想深入探讨剽窃这个问题。

On the one hand, I want to go deeper on the the issue of plagiarism.

我喜欢你所说的，剽窃其实是一种思维方式，而不仅仅是关于最终的输出。

I love what you're saying around it being a mode of thought rather than about the output.

我觉得这个对话中还有更多未被提及的内容，我怀疑你会说出一些非常启发性的见解。

And I think there's I think there's more in this conversation that I don't see being said, and I suspect you will say really enlightening things.

我还想为那些不太熟悉的人进一步澄清一下。

And then I also want to seek more clarification for folks who might not be familiar.

你多次提到，由加沙战争所引发并推动的算法暴力。

You referenced a couple of times the algorithmic violence brought on corollary to and driven by the war on Gaza.

你指的是什么？

What do you mean by that?

是的。

Yeah.

这是一套相互关联的系统，对吧？

It's a network of systems, right?

所以，当谈到算法暴力时，我们常常认为数字世界和现实世界是两个独立的领域，只有在我们敲键盘时才偶尔接触，其他时候都只是在后台运行。

And so I think when talking about algorithmic violence, I think we often have this perception that the digital realm and the real world are sort of like these separate entities that maybe touch along our fingertips when we're at a keyboard, but otherwise are just operating in the background.

但当然，这种情况从来就不成立。

And of course, that's never been the case.

像西蒙娜·布朗这样的学者，她的《黑暗物质》对我而言至关重要，她探讨了监控与数据价值，而‘数据价值’这个术语，我认为丽塔·雷利是最主要的提出者。

From scholars like Simone Brown, whose dark matters is incredibly important to me, thinking about surveillance and data valence, a term that Rita Raley, I think, is most prominent for developing.

算法系统一直在监视我们，用于实施控制机制。

Algorithmic systems have been watching us and watching us for systems of control.

我的很多训练源于德勒兹和亚历克斯·加洛韦，那正是这种思想的鼎盛时期，大约在2010年代初，围绕德勒兹所说的‘控制社会’。

A lot of my training came out of Deleuze and Alex Galloway in a kind of like heyday of that type of thought, heart and agri from the odds and the early teens around what Deleuze called a control society.

因此，信息体系尤其是计算机，行使了一种新的权力形式，不是生物权力，而是规训社会——抱歉，我正在复述理论话语的流程，但提醒自己这一点很有帮助：我们正是从这里出发的，福柯曾从制度化的权力形式出发进行理论构建，用机构国家机器使身体变得顺从，用阿尔都塞的术语来说，比如老师用尺子打你以让你服从；而如今，我们有了更灵活的控制形式，比如某种商品的价格略微下降。

So the idea that information systems and computers in particular exercised a new form of power, not the biopower, the disciplinary society sorry, I'm getting like, going through my paces of like the theory discourse, but it's, you know, it's helpful to remind oneself that this is where we come from that, you know, that Foucault theorized from institutional forms of power that train bodies to be docile, using institutional state apparatuses in Althusser's terms, From those, I think the example is from a teacher, whapping you with a yardstick to keep you docile, we have these forms of flexible control where the prices on one commodity go slightly down.

于是你被引导去购买那件商品，或被激励以某种方式行事。

And so you are nudged toward buying that thing, or you're incentivized to behave in a certain kind of way.

因为如果你不这么做，你不会进监狱，也不会被打，但也许你就拿不到好的保险，对吧？

Because if you don't, you're not going to go to prison, you're not going to get beat, but like, maybe you won't get good insurance, right?

所以，这是一种系统性的控制，这种观念源于控制论的想象。

So it's these like systems of and this comes out of like the cybernetic imaginary.

我重新读了《控制论》这本书，我应该推荐一下，它太棒了。

I reread the Ticune book, which I should plug, it's so good.

这些控制论系统一直将我们维系在这些控制状态之中。

The cybernetic systems that have kept us in these kinds of control states.

好的。

Okay.

这都是背景。

That's like background.

这些都是在大型语言模型兴起之前的事了。

That's all before the rise of large language models.

所以要谈到大型语言模型，你必须先走过这些阶段，理解那些控制形式和暴力形式——大型语言模型及其他形式的合成认知所造成的，不仅是在规模上全新，而且我认为它们采取了截然不同的形式，要求我们重新审视主权、重新理解控制，以及重新思考算法可能造成的伤害类型。

So like to get to large language models, you kind of have to like go through those paces and and the forms of control and the forms of violence that large language models and other forms of synthetic cognition are producing are not just like on a wholly new scale, but I think take dramatically different forms that require us to revise our understanding of sovereignty, revise our understanding of control, and revise our system of like, our understanding of what kinds of harms algorithms can produce.

以加沙战争为例，这不仅仅是针对特定目标的数据分析、人脸识别、无人机袭击或导弹操作系统，就像从电影、写作到我们点外卖等几乎所有行业，都越来越受到大型语言模型的影响。

So just to use the example of the war on Gaza, there's not just the data valence of specific targets, tracking faces, drone strikes themselves, missile operating systems, in the same way that like, it seems every industry from film to writing to how we order our food for delivery are increasingly subject to large language models.

当然，这些军事系统也在使用相同类型的模型。

Of course, these military systems are using the same types of models.

这些模型，重要的是，再次强调，存在严重偏见，容易犯重大错误，且通常缺乏有效监管。

And these models importantly, again, are profoundly biased, prone to egregious errors and often have very little oversight.

所以，如果你的笔记本电脑因为你的种族、性别或身份特征不符合标准而无法识别你的脸，这是一回事。

So it's one thing if say like your laptop computer doesn't recognize your face, because you're the wrong race or the wrong gender or have the wrong set of identity coordinates in general.

如果这是无人机的瞄准系统，那就完全是另一回事了。

It's something very different if that's a targeting system from a drone.

这属于基础性的工作。

That's like legwork stuff.

是的，我认为这正是让我夜不能寐的原因之一。

Yeah, I think that that's I mean, so and I think this is part of what keeps me up at night.

这正是推动我所从事的研究和我在加州大学洛杉矶分校教学的动力所在——将那些看似轻松随意的文学实验，与我们周围无处不在、正在改变一切的现实且日益具有存在性威胁的技术联系起来，正如你在这档播客和众多节目中精彩呈现的那样，这些技术正在改变气候，也正在书写我们即将面对的所有紧迫未来，而这些未来是由我们尚未完全理解的系统所塑造的。

This is part of what animates the kind of research that I'm trying to do, and the kind of teaching I'm trying to do at UCLA, which is to connect these what look like playful and frivolous literary experiments to the very real and increasingly existentially threatening technologies that surround us everywhere that are changing everything from the climate as you cover so beautifully in this podcast and in so many of your episodes to all of the urgent futures that are awaiting us, that are being written by systems that we don't quite understand.

我出身于一个深受基特勒、尼采和爱默生影响的背景，他们认为我们的书写工具正在塑造我们的思想。

And I come from a background and rooted in a lot of reading of Kittler and Nietzsche and Emerson around the idea that our writing tools are working on our thoughts.

因此，我们用来写作的技术从来都不是中立、无意识形态的传输系统。

So the technologies we use to write are never transparent ideology free transmission systems.

约翰·图林·彼得斯称之为‘天使的沟通’。

What John Turrin Peters calls the communication of angels.

任何沟通渠道都不可能没有噪音。

There's never a communication channel without noise in it.

如果我们不理解写作技术如何影响我们的思想，我们就无法抵抗它们，更谈不上更好地改造它们，甚至有朝一日以不同的方式来使用它们。

And if we don't understand the ways that our writing technologies are working on our thoughts, then we don't have a way to resist them or better and ideally someday to write them otherwise.

我非常喜欢这一点。

I love that.

句号。

Full stop.

不是说这个节目到此为止。

Not full stop on the show.

是的，我很感谢你分享这些，因为我觉得对很多平时不常思考这个问题的人来说，他们可能会觉得，好吧，有ChatGPT，有各种AI垃圾之类的。

Yeah, I really appreciate that you shared that because I think for a lot of people who don't kind of spend their days stressing about this, it's kind of like, yeah, okay, there's ChatGPT and there's like AI slop and whatever.

我知道我可能不该用它，但反正我也不想写邮件，而它确实是个写邮件的好办法，诸如此类。

And I know I probably shouldn't use it, but, you know, I don't feel like writing my email and this will be a really good way of writing my email or whatever.

但这一切其实都是相互关联的系统。

But it really is all of these interlocking systems.

从我的角度看，似乎找不到一个恰当的比喻。

Like it feels like there's no good metaphor from where I sit.

也许你会有一个绝佳的隐喻，但对我来说，似乎找不到一个合适的隐喻来全面描述这种全方位的困局——事实上，ChatGPT 只不过是更深层、更根深蒂固的系统的表面，而这个系统几乎渗透了所有行业。

Maybe you'll have an amazing metaphor, but it feels like I have no good metaphor to describe the lock in from every angle that feels like is happening, where it's like, if anything, ChatGPT is still just the kind of like fun, silly, you know, kind of like face of a much deeper and more entrenched system that spans essentially all industries.

而掌握权力的人对这一点的理解，远比普通知情者更透彻。

And people in positions of power understand that much more thoroughly than even your average informed person.

是的。

Yeah.

所以，当你谈到它在加沙地区被使用的方式，以及因为加沙而被使用的方式时——你甚至还没提到社交媒体平台上言论压制的问题。

So, yeah, when you were talking about the ways it's being used in Gaza and and because of Gaza and and, I mean, you didn't even get into the, like, suppression of speech on social media platforms.

它所产生的种种回响，往往只在缺席时才被感受到。

Like there's so many ways that it echoes that are only felt in their absence.

因此，是的，我知道这可能是个很基础的入门级问题，但我真的很感谢你。

And so, yeah, I'm just, I know that it was kind of like a kind of elementary 101 style question, but I really appreciate you

添加？不，不是这样。

adding No, to wasn't.

我认为，正如我在书中试图表达的那样，即使在大型语言模型出现之前，仅仅理解这些平台上任何一个数字对象的含义，就已经耗尽了我们的诠释能力。

And I think, you know, it raises something to, you know, draw back to what I'm trying to do in the book as well is that, you know, even before we get to large language models, the situation of just trying to understand any given digital object on these platforms already exhausts an interpretive equation.

你让我想想，哦，你最担心的是什么？

You asked me to like, Oh, what are you most worried about?

我有一大堆要列出来。

It's like, I have a laundry list.

我只是想挑一些极端的例子。

I was just trying to get some extremes.

但事实上，这不仅是一份清单，列出了这些技术带来的新挑战——比如试图理解它们如何运作，或者，但愿如此，以某种方式干预它们。

But like, there's a laundry list, not just of like, of new challenges that these things present, challenges to trying to understand how they work or, you know, God willing, like intervene in them in some way.

除了这些系统带来的挑战之外，我认为它们还抵制传统的解读方式。

So beyond that list of challenges that these systems present, I think they also resist traditional modes of reading.

它们也抵制绝对的理解形式。

And they resist absolute forms of comprehension.

没有人能真正了解一个大型语言模型。

One can't know a large language model.

就像我一位朋友形容的那样，它就像一个固态的神。

Is like a friend of mine described as a solid state God.

它就像什么都知道，却又什么都不知道的东西。

It's like something that knows everything, but it also knows nothing.

它同时极其愚蠢，又极其复杂。

It's incredibly stupid, incredibly sophisticated at the same time.

因此，这些我们仍在试图用原始大脑理解的新形式，需要其他方式来揭开它们的黑箱，以理解它们的工作原理——而艺术和诗歌一直擅长于此。

And so these new forms that we're still trying to wrap our reptilian brains around require, I think, these other ways of trying to pierce their black boxes of trying to understand how they work, this is something that art and poetry has always been good at.

马歇尔·麦克卢汉曾将前卫诗歌称为技术系统的高级探针。

Marshall McLuhan famously talked about avant garde poetry as a kind of advanced probe in technological systems.

对于麦克卢汉来说，你只有在一种技术过时之后，才能真正理解它。

You know, for McLuhan, you can only really understand a technology after it had become obsolete.

但艺术家在技术发展过程中对其进行测试时，可以捕捉到其工作原理的线索、迹象和征兆。

But that the hints and the signs and the indices of how that technology worked could be probed by artists that are testing it as it's developed.

而通常你只能在事后、回顾时才能看到这些。

And usually you can only see that in hindsight, in retrospect.

但我们的技术变化得太快了。

But our technology is changing so fast.

大型语言模型每天、每周、每月都在以无法理解的方式变化。

And large language models change day to day, week to week, month to month in ways that are incomprehensible.

因此，让我学生去实验这些系统，我自己也做实验，追踪其他作家和艺术家如何尝试切入其中一小部分——就像取一个微小的细胞样本，以便分析那些可能具有更大系统性影响、但实际上是与一个极其复杂系统进行的具体而偶然的互动。

And so having my students experiment with these systems, doing my own experiments, tracking what other writers and artists are doing to try to pierce just small sections, almost like taking a small cellular sample so you can analyse something that might have some larger systemic impacts, but is actually a very specific and contingent encounter with an otherwise impossibly complex system.

我认为这是我所倡导的，也是我希望能够赋予艺术和诗歌的力量，而不是某种意义上的‘诗歌’。

I think this is something that I'm advocating for and something that I hope also kind of empowers art and poetry, not as some kind of like, poetry.

詹姆斯·谢里以最著名的方式说过：诗歌是唯一一种贬低其所用材料的媒介。

And James Sherry said this in the most famous way, which is that poetry is the only medium that devalues the materials it uses.

一张空白的纸，比一张写有诗的纸更有价值。

That a blank sheet of paper is worth more than a sheet of paper with a poem on it.

你知道，这在绘画中可不一样，对吧？

You know, it's not the same with painting, right?

这就像价值的增加。

It's like value added.

所以，如果诗歌没有内在的物质价值，那它实际上是在贬低它所使用的材料。

And so like, if poetry has no intrinsic material value, it's actually devaluing the materials it uses.

也许它的价值在于它能帮助我们理解，就像文字之于意义、色彩之于我们与自然和宇宙的关系一样。

Maybe its value is in the ways it can help us to understand, as it always has for words worth or colors for like understand our relationship to nature and the cosmos.

我只是觉得，试图理解一个大型语言模型，或许就像试图理解自然和宇宙一样。

I just think that trying to understand a large language model is maybe like trying to understand nature in the cosmos.

我们需要能够让我们瞥见这些otherwise难以理解的超对象的诗性形式和美学形式。

And we need poetic forms and aesthetic forms that are capable of giving us some glimpses of these otherwise, you know, transcendent hyper objects.

也就是说，回到教学这一面，这正是我经常强调的一点：显然，我的大部分批评针对的是你隐含或明确提到的那些人。

That's I mean, coming back to the pedagogical side of this, that this is very much a drum that I feel myself banging a lot is like obviously, most of my critique lies with the people that you're implicitly and explicitly sort of naming.

另一方面，我也对那些我在政治立场上高度一致的人感到非常沮丧，因为他们会本能地拒绝。

On the flip side, I feel a lot of frustration at folks that I feel very otherwise politically aligned with where there's this knee jerk no.

就像，我绝不碰这些东西。

Like I'm not touching any of that.

不仅我不碰，我甚至懒得去了解它。

And not only am I not touching it, I'm not even going to bother to try to learn about it.

因为我觉得这涉及多个层面。

Because I think it's sort of multiple levels.

如果你认为这些工具不该存在，希望彻底废除它们，那很好。

Like, if you believe that those tools shouldn't exist and you want to stand for their abolition, great.

但如果你主张废除它们，就必须真正理解背后的原因。

But also, if you're going to stand for their abolition, you need to really understand why.

因此，了解它们的工作原理，责任就变得更加重大，必须真正了解你的对手。

And so knowing how they work then becomes like the onus becomes even greater to then really kind of like know your enemy.

艺术家和创意人士正迎来一个契机，去承担起你正在做的工作——如你所说，去戳破、去揭示。

And there's this moment for artists and creative people to step up and do the work you're doing of like kind of, as you said, sort of poking or piercing.

我不愿过度赋予它沉重的分量，但确实刻不容缓。

And like not to sort of, you know, be like, sort of overly freight this with gravitas, but it's like, it is urgent.

我不知道还能怎么表达这一点。

Like, I don't know how else to say it.

这真的刻不容缓。

It is urgent.

形势正在急剧恶化。

Like the walls are closing in.

我面带微笑地说，但我们都知道，这并不是值得微笑的事情。

I'm saying it with a smile, but we all know that it's not the type of thing you smile about.

所以我们迫切需要创意人士。

And so we need, we just desperately need creative people.

而许多创意人士恰好属于政治左派。

And a lot of those creative people happen to be on the political left.

而且，在这些人的交集中有相当一部分人，他们的批评虽然我可能认同其立场，但数据不足或表达不当，反而只会伤害我们。

And a lot of the, there's a pretty, you know, healthy middle section of a Venn diagram of those people who are just like, and making bad critiques where it's like a critique that I might agree with, but with poor data or poorly stated is only hurting us.

所以，我绕了这么大一圈，其实就是想说，我非常欣赏你所做的工作和你所表达的观点。

So I just, that's a long way of saying I really appreciate the work you're doing and what you're saying.

是的，我认为各种声音都有其存在的空间。

Yeah, and I think there's room for all kinds.

我喜欢新卢德主义者。

I love a neo Luddite.

我还要再引用一次乔纳森·夸里在我们对话中的观点。

I love and I'll step back to reference Jonathan Quarry once more in our conversation.

去年我教了一门研究生研讨课，与我正在做的项目高度契合，我称之为‘推测性游戏与媒体诗学’。

I taught a grad seminar this past year, very much aligned with the project, I think of Virgin Futures on, called it speculative games and media poetics.

所以我们基本上是把这门课当作电子游戏研究课来上，做大量媒体诗学的研究，就像我在本书中试图做的那样，但重点围绕推测性问题展开。

So we're kind of looking at it as a video game study class, doing a lot of media poetics like I'm trying to do in this book, but really like geared around questions of speculation.

未来是如何被呈现的？

Like how is the future rendered?

我们可以做出哪些推演？

What kinds of extrapolations can we make?

是否存在可能属于我们的未来？

Are there possible futures for us out there?

而且，作为一个真实的问题，你知道，到了现在这个地步，是的，是否存在一个可能的未来？

And like, as a real question, you know, at this point, like, yeah, is there a future that that's possible?

我开设这门课，是因为我认为必须从一片焦土开始。

And I started the class because I thought it was so important to start with scorched earth.

换句话说，如果我们想要想象任何一种未来，就必须从某种起点出发——我应该重述一下，这是来自乔纳森·克拉里的一篇极具朋克精神的论著，他因《观察者的技术》等众多著作而闻名。

It's like, we're gonna imagine any kind of future we have to like begin maybe which I should just rehearse is this incredibly punk rock tract from Jonathan Krayer, famous from Techniques of the Observer among many, many other interventions.

关于24/7资本的伟大著作。

The great book on twenty fourseven Capital.

但在《焦土》中，他强烈主张，根本无法通过干预这些系统来实现任何积极的改变，唯一的办法就是彻底摧毁它们。

But in Scorched Earth, he really advocates for just like, there's no his argument is that, like, there's no way to intervene in these systems to to produce any kind of positive change except for to burn them all down.

我们必须彻底断开连接。

We have to just like fully unplug.

这是主人的工具。

It's a master's tools.

这确实有点像是主人的工具这种说法。

It's a kind of master's tools thing, kind of.

这简直像一份激进的互联网废除主义文献。

It's really like a radical kind of like internet abolitionist document.

我觉得它非常具有挑衅性。

And I find it very provocative.

我认为这些观点有很大的存在空间。

And I think that there's lots of room for those perspectives.

就像，你知道的，我尊重我的同行和你所描述的那些人，我的学生们，他们越来越不愿意接触人工智能。

Like do, you know, I respect like my peers and the people you're describing, my students, who increasingly don't want to touch AIs.

这有点令人惊讶，因为我教一门叫算法文学的课，是批判性人工智能研究的入门课。

It's kind of surprising, like I teach this class called algo lit, introduction to critical AI studies.

大约有一半的学生完全是废除主义者，他们坚决不碰大型语言模型，这简直太惊人了。

And about half the students are absolutely abolitionists, like they will not touch a large language model, which is kind of amazing.

这就像是参加精神分析研讨会时说：我拒绝读弗洛伊德的任何一个字。

It's like showing up to like, I don't know, like a psychoanalysis seminar and saying, I refuse to read a word of Freud.

这确实挺有意思的。

It's just like, well, interesting.

因此，与其把这当作对课堂的挑战，实际上它已成为我课程中最有益的部分之一。

And so rather than take that as like a challenge to the class, it's actually been, I think one of the most beneficial parts of the classes.

我们有来自不同背景的学生，他们以各种方式希望与这些媒介互动，并想出富有创意的方法来介入这些话语体系。

We had students from a whole range of ways of wanting to interact with these media, coming up with like creative ways to try to intervene in these discourses.

所以，我同意，写一篇缺乏了解的火爆评论对讨论并没有太大帮助。

And so I agree writing a hot take that's ill informed is not the most useful thing for the discourse.

对于那些不想接触AI的学生，我特别提到一个学生，他创作了一本精美的漫画书，内容基于我们曾讨论过的一系列对话和艺术作品——他们并没有生成任何新的内容，但却以一种若非研究生成式AI的实际输出就不可能实现的方式与这些材料互动。

For my students who don't want to touch AI, would make one in particular that made this like beautiful comic book about AI systems based on a series of conversations and artworks we had seen that they weren't producing any generative work, but they were engaging with the materials in a way that they simply wouldn't have if they hadn't been studying the actual outputs of generative AI.

所以我认为，存在一些方式可以对抗这些系统，而无需将数据反馈回它们的训练集、订阅经济，或让大学与科技公司建立合作关系，许多科技公司都如此。我认为，不同方式都有其价值。

So I think that there's ways to, you know, fight the good fight against these systems without necessarily feeding back into their training data sets, economics of subscription, the university making partnerships say with tech companies, many tech And so like, I think that, you know, it takes all kinds.

我认为这需要——感谢你提出这个问题。

And I think it requires I think this is something I had Thank you for this question.

我认为这需要认识到，这些问题没有唯一的答案。

I think it requires There's not one answer to these problems.

而答案或问题的类型越多，不仅在主题或内容上多样，更在形式和实践上多元，我们对这些系统的理解就越全面，也越有可能揭开它们的神秘面纱，甚至削弱它们的影响力。

And the more kinds of answers or the more kinds of questions that are variant, not just in theme or topic, but variant in form and practice, the better a view and the better position we might be in to both understand and ideally demystify or defang these blood suckers.

阿门。

Amen.

我同意。

Amen to that.

那么，让我问你一个问题。

Well, let me ask you then.

你刚才说的这一切，就像是从一堆积木中，从大块的推倒，一直推到最小的那颗小多米诺骨牌。

How does this like everything you just said, we're gonna we're we're moving from the the pile on the topples down to the little the little baby domino.

在你看来，这与当前持续发生的、关于剽窃的恐慌时刻有何关联？

How does this interact in your mind with the plagiarism, hair on fire moment that has been happening but continues to unfold?

是的。

Yeah.

这正是我多年来不断思考、思绪不断转变的问题。

This is something that I've had to I feel like my mind has gone in a number of different directions over the years.

对吧？

Right?

我成长的过程中，我们曾在休息时聊到过种子下载、Limewire和海盗湾，那种左翼反主流文化的文件共享互联网，曾被设想为一种礼物经济或数字公地。

I came of age we were talking in the break a little bit about torrenting and Limewire and Pirate Bay and this idea of like, a leftist countercultural file sharing internet, which was theorized as a kind of gift economy or digital commons.

但这些相同的系统，如今却直接为大型语言模型和训练数据提供素材，对创意艺术家、小说家、作家和学者造成了真实伤害，他们的作品被未经许可地使用了。

Well, like, those same systems are the ones that feed directly into large language models and training data producing real harms also for creative artists, for novelists, for writers, for scholars whose works have been used.

越来越多的人接入ChatGPT，而正如我们所知，它输出的大部分内容其实直接来自训练数据，并非每次都在创造全新的句子，很多时候只是在直接复述别人的东西。

Increasingly, if everyone's just plugged into ChatGPT, which as we know, mostly just outputs direct content from its training data, it's not always inventing every new sentence you read from its output, it's often just straight up regurgitating something else.

如果你能转向这一点，那么这对创意产业会带来进一步的负面影响。

If you can just turn to that, then that has further deleterious effects for the creative industries.

谈到版权，我尽量避免深入探讨，因为这方面的优秀研究已经很多了。

Talking about copyright, I resist getting too deep into copyright because there's so much good work out there.

彼得·德舍尔尼我认为长期以来在版权、控制以及合理使用法律历史的诉讼方面做了非常出色的工作。

Peter Descherney, I think has been doing really great work for a long time on questions of copyright and control and tracking the legal history of fair use, how that's been litigated.

当然，像我这样在90年代和2000年代成长起来的人，坏人一直都是像IRA和迪士尼这样的公司，这甚至可以追溯到60年代。

Of course, like, for me, like, you know, in the 90s and the aughts, the bad guys were, you know, the IRAA and Disney, goes back into the 60s.

这些版权斗争，版权运作的方式，一直以来都是为大型企业服务的。

These copyright battles, the way that copyright functions has always been to serve large corporations.

它从未在任何历史时期真正保护过独立艺术家。

It hasn't been about protecting independent artists at any point in history.

但这就像一把双刃剑，对吧？

But it's just like it's this dual sided blade, right?

在很多方面，它就像一种药饵，而我研究和感兴趣的很多内容都是影子图书馆——这些由盗版、扫描和非法材料组成的集合，我多年来一直在研究，比如LibGen，我就不一一列举其他的了。

Like it's a pharmakon in a lot of ways where, you know, a lot of what I study and I'm interested in are shadow libraries, these collections of pirated, scanned, illegitimate materials I've been researching for years, things like LibGen, I won't name all the rest.

LibGen绝对是王者。

LibGen's a GOAT, of course.

是的，没错。

Yeah, yeah.

你懂的，懂的。

If you know, you know.

对吧？

Right?

这些影子图书馆，乔·卡拉加蒂斯等人曾提出理论，最近盖尔·德科兹尼克也在研究这个问题，它们对于那些无法像我一样在加州大学洛杉矶分校享有丰富图书资源的学生来说至关重要，尤其是来自全球南方的人们，他们想要获取科学论文、研究和文学作品，而影子图书馆以一种我认为具有政治活力、重要且解放性的方式提供了获取知识的途径。

These shadow libraries, Joe Karangatis and others have theorized that are, Gail Dekoznik recently, I know, doing research on this as well, are absolutely essential to students who don't have access to the same kind of knowledge infrastructures surrounded by books here at UCLA that I have, Particular folks from the global South, who want to access scientific papers, research, literature, the shadow libraries have afforded access to knowledge in ways that I think are politically vibrant and important and liberatory.

但与此同时，这种药理学的双重性在于：如果这是解药，那么伴随而来的毒药就是，这些相同的影子图书馆积累了数百万本书籍，这些书籍构成了大型语言模型的语料库，而这些模型由科技寡头掌控，用以推动一种新的科技封建主义体系，我们预计在未来几年内将迎来首位万亿富翁。

And at the same time, this is the pharmacological where if that's the cure, the poison that comes alongside it is that these same shadow libraries have amassed the millions of books that have fed the corpuses of the large language models, which are run by tech oligarchs to fuel a new system of techno feudalism where we're on track to see our first trillionaire in a couple of years.

因此，我找到一个非常有用的方式来思考这个极其复杂的问题，那就是帕特里克·贾戈达所称的‘网络矛盾性’。

So one way that I found really useful for thinking about this incredibly vexed situation has been what Patrick Jagoda calls network ambivalence.

这个观点认为，研究互联网这样的事物，你必须同时容纳这些平台所造成的种种恐怖，以及它们所蕴含的深刻而有意义的喜悦、惊喜或愉悦。

This idea that to study something like the internet, you have to hold the manifold horrors of these platforms right alongside the deep and meaningful joys or surprises or delights that they hold as well.

所以像Reddit这样的平台，显然在许多角落充斥着仇恨、厌女症和种族主义，但与此同时，它也是政治行动主义、社区组织构建以及与性别身份和解的空间。

So thinking about something like Reddit, obviously fueled by hate and misogyny and racism in so many corners, but at the same time, a space for political activism and organizing for community formations, for reconciliations with one's gender identity.

无数充满喜悦和意义的相遇都是由同一个平台促成的。

Any number of joyful, meaningful encounters are facilitated by that same platform.

因此我认为，在应对技术和科技研究中的这些持续趋势，或是试图理解像大型语言模型这样的事物时，都需要一种健康的网络矛盾心态。

And so I think when addressing these ongoing trends in technology and in technology studies, or trying to wrap our minds around things like large language models requires a healthy dose of network ambivalence.

或者我认为更常见的说法是，借用哈拉维的‘与麻烦共存’理念，人们不能全盘批判并彻底否定这些系统。

Or I think the more common version, this is Haraway staying with the trouble, that one can't outright critique these systems and negate them altogether.

我的意思是，你可以主张彻底摧毁互联网，这我没意见。

I mean, you can, like you can advocate for burning down the internet, that's fine by me.

但你也不能完全无条件地拥抱它们，因为那只会助长这种盲目乐观的论调，忽视了这些系统造成的实际危害。

But you can't just like wholeheartedly embrace them either, because that just adds to this booster ish discourse that's unaware of the real harms that these systems produce.

因此，我认为通过直面困境，需要持续努力将两者区分开来，并努力保持足够清醒的头脑，以识别它们在任何地方可能出现的差异。

And so by staying with the trouble, I think it requires one to continuously work to disentangle the one from the other, and to try to be clear sighted enough to recognize the differences wherever they might emerge.

嗯，这让我想到。

Well, it strikes me.

我的意思是，这非常符合我对你的了解和你的个性，但你谈论的是一些令人不安的事情，却用了一种轻松的方式去表达。

Mean, this is very in keeping with, you know, what I know of you and your personality, but you're talking about incredibly troubling things, but you're bringing this, like, playful approach to it.

是的，这让我想到。

Does it yeah.

跟我聊聊，这种‘玩乐’是如何融入你对这些问题的思考的，更广泛地说，也许也涉及LittleDatabase中的相关话题。

Talk to me talk to me about where, like, play fits into how you're thinking about this and more broadly, maybe topics in LittleDatabase.

是的，我很喜欢这个关于‘玩乐’的问题。

Yeah, I like this question about play quite a bit.

如果我停顿了一下，那是因为我被带回到了我在犹他州乡村的童年，我是在摩门教环境中长大的，那时我坚信世界随时可能终结，耶稣随时会回来，开启千年的审判。

If I pause for a moment, it's because I'm being transported back to my upbringing in rural Utah, I was raised Mormon in a fully eschatological imaginary that the world was going to end at literally any moment that Jesus was going to come back thousand years of judgment.

我想我小时候太小的时候就读过一本关于广岛的书。

I think I read a book about Hiroshima when I was like, way too young.

所以有一年，我父母告诉我，每当天黑时，我就以为是原子弹爆炸了，世界结束了。

And so for about a year, my parents told me when I was quite young, any time that the sun was setting, was convinced it was an atomic bomb and the world was over.

所以，我想我对世界的终结有一种亲密感，这种感觉是我从小到大一直拥有的。

So like, think that there is like, there's an intimacy with the end of the world that I think I've grown up with.

我写过一本叫《末日遗物》的书，这是一本漫画，描绘了一个虚构的博物馆，收藏了每次世界毁灭后留下的物品。

I've written about this in a book called Apocalypse Reliquary, which is a comic book that is a kind of like fictitious museum that preserves an object from each time the world is destroyed.

所以世界一次又一次地被毁灭。

So the world is destroyed over and over and over again.

我为各种末日预言做了大量研究，从科雷什到据说有1600万天使乘坐宇宙飞船降临地球之类的说法。

I did a lot of research for like every prophecy at the end of the world ranging from like Koresh, up to like, say, like 16,000,000 angels will descend on the planet in spaceships.

然后想象这些全都真实发生过。

And then imagine them as though they all really happened.

所以世界就这样一遍又一遍地被毁灭。

So the world is just destroyed over and over and over again.

我想尝试做点什么，但也不太确定自己到底想做什么。

To try to like, I don't know exactly what I'm trying to do with that.

但我想通过这种方式触及一种情感——感受世界正在终结，以及人们在这种情况下能做些什么，我认为这是一种我们所有人都持续面对的情感。

But it was trying to like access, I think, an affect of feeling like the world is ending and what one can do, which I think is a feeling that we are all contending with pretty consistently.

这就是为什么我听《紧急特征》时感觉特别好，因为我觉得它让我重新连接到那种感觉：哦，原来人们是真实的，他们在谈论世界的终结。

It's why it's so nice for me to listen to Urgent Features, because I feel like it like taps me into that feeling where it's like, okay, like people are real, like they're talking about the end of the world.

这真的正在发生，就像我从小被训练去想象宗教意义上的末日场景一样，那时那不是我的想象，而是现实。

And this is really happening in a way that like I was kind of trained to imagine what happened in religious register from my early youth, no longer my imaginary, but at the time it was.

因此，我认为，长期与世界末日的想法共处，让我对这些恐怖事件产生了一种乐天派般的视角。

And so I think that like that living with the idea of the end of the world has led to a kind of like Pollyanna ish perspective on these horrors.

我一直很喜欢那种口袋里藏着炸弹的无政府主义者形象。

Like, I've always loved like the figure of like, the anarchist with a bomb in his pocket.

我觉得，我们所做的工作应该带有一种愉悦和快乐，因为我们在这世上停留的时间并不长。

Like, think that there's a kind of like, there should be a kind of like pleasure and a joy to the work that we do, because we don't get to be here for very long.

我认为，我所做的一切工作中都充满了游戏性。

And I think like, there is like a great deal of play in all of the work that I'm doing.

我想，我们可以用玛丽·弗拉纳根所说的‘严肃游戏’或‘批判性游戏’来框架这种想法——游戏能够测试各种情境、不同系统和类型的规则。

And I think, we could put this in the framework of like, serious play or critical play in Mary Flanagan's terms, where play is able to test the rules of various kinds of situation, test the rules of different types of systems and genres.

但同时，我认为这也是作为地球上的人类至关重要的一部分：试图找到一种方式来处理这些主题，持续与它们相伴，以一种不会摧毁你、不会让你陷入深度抑郁的方式，去思考更广阔的可能。

But it's also, I think, an important part of just being a human being on the planet and trying to figure out a way to work on these subjects and continue to linger with them, to stay with those troubles in a way that doesn't kill you, that doesn't drive you into a deep depression, that tries to think about possibilities outward.

在我的近期研究中，我确实转向了电子游戏，我想正是出于这个原因。

In my recent research, I've really turned toward video games, I think, for this reason.

我非常喜欢克里斯·帕特森写的《开放世界帝国》这本书，书中提出了一个非常引人入胜且优美的观点，认为世界政治体系就像一款开放世界电子游戏。

There's a book that I love by Chris Patterson called Open World Empire that makes this really compelling and beautiful argument about the world political system being kind of like an open world video game.

当你玩开放世界游戏时，你无法……我的意思是，你可以关掉游戏，但那样你就不再玩游戏了。

Where like, can't, like if you're playing an open world video game, you can't like, I mean, you could shut it off, but then you're not playing the game anymore.

如果你身处游戏中，你就身处这个开放世界——他说，我们的开放世界帝国中，你所能希望做的最好的事，就是变得非常、非常叛逆。

If you're in the game, you're in the open world, like our open world empire, he says that the best you can hope to do is just like basically, I forget exactly how he says this, but it's basically like, you can be really, really naughty.

你可以故意把规则玩得很糟糕。

You can just like, you can like play the rules poorly.

你可以尝试探索这个世界物理法则所能允许的一切可能性。

You can experiment with what the physics of this world are going to allow.

你可以试图颠覆和推翻你所玩的虚拟游戏中的经济体系。

You can try to subvert and upend the economics of the virtual game that you're playing.

游戏是模拟其他世界和存在方式的途径与空间。

That like games are ways and spaces for modeling other worlds and other ways of being in the world.

因此，长期以来我一直致力于一个小型数据库项目，试图去‘玩’这些数据库，通过探索它们的内容，揭示潜在的可能性、隐藏的意义，或理解世界的新方式。

So if I've been working for a long time on this little database project and trying to like play these databases, play through the contents of these databases to surface latent potentials or hidden meanings or new ways of understanding the world.

我觉得，从这种对网络文化与文化记忆对象的研究方式，自然地演进到了我最近在做的关于推测性游戏的工作，并试图深入思考我们与知识的关系是如何转变的。

I feel like there's been a kind of natural progression from that type of approach to network culture and cultural memory objects.

这直接引导我关注到推测性游戏，试图深入思考我们与知识的关系是如何转变的。

That's led directly into what I've been doing recently around speculative games, and trying to really think about how our relationship to knowledge has transformed.

也许电子游戏并不是进行这种思考的最可能场所。

Maybe video games is an unlikely place to do that.

但这就是我一直在思考的问题。

But that's what I've been thinking about.

这太棒了。

That's amazing.

我的意思是，你能再多说说吗？我们之前聊过《死亡搁浅》。

I mean, can you say more about what I mean, we talked about Death Stranding earlier.

但当你在做研究、进行严肃游戏时——当然不是坐在沙发上边吃零食边做——上帝啊，你从自己身上发现了什么？

But like, as you're researching and conducting serious play, and definitely not sitting on your couch snacking as well while you do it, God What are you finding in yourself?

比如，回想起你童年时的末日情境，当时游戏是如何为你开辟了一条出路的？

Like, thinking kind of back to your, you know, doomsday childhood and the way that play created an avenue through.

现在，在这种末日紧迫感的时刻玩游戏，并将来自《小数据库》的所有研究融入到下一个与游戏相关的神秘项目中。

Now playing games at this moment of eschatological urgency and kind of bringing all the research from Little Database into whatever mystery thing is coming next related to games.

那你感觉怎么样？

Like, how are you feeling?

你遇到了什么？

What are you encountering?

让我进入那种思想和情感的状态吧。

Like, me put me in in that head and heart space.

这是个非常美好的问题。

That's such a lovely question.

是的，我觉得很有趣，因为我喜欢那些抗拒沉浸感的艺术作品。

Yeah, I feel like I it's funny because I love artworks that resist immersion.

但我又非常喜爱完全沉浸式的游戏。

And yet I love games that are fully immersive.

我喜欢扮演一个在开放世界游戏中穿越荒凉景观的角色。

Like, I love to be a character traversing a barren landscape in an open world video game.

我认为，这种模拟空间中的情感和游戏性产生了我所追求的种种情绪。

And I think a lot of that has to do with affect and play in these kinds of simulated spaces produces these kinds of emotions that I'm looking for.

我常常在寻找这种类似科幻或推测性的延伸，正如所有科幻作品一样——正如詹姆逊等学者所说，它通过从我们当前的处境出发进行推演，以一种全新的方式反映当下，带我们抵达某个奇异的境地。

I think I'm often looking for this kind of like, like science fictional or speculative extrapolation that makes me as all sci fi, know, Jameson and others have said, as all sci fi does that reflects on the present in a new way, by extrapolating from where we are to bring us to somewhere weird.

但这个奇异之地其实并不是要将我们带离当前的现实。

But that weird place isn't actually like taking us away from where we are.

它只是让我们以一种新的视角，重新认识我们所处的位置。

It just allows us to see where we are differently in a new light.

而詹姆逊称之为‘乌托邦欲望’，即去寻找对我们所处政治体系的一种回应。

And what Jameson calls that desire called utopia, right to like, look for some answer to the political systems we find ourselves in.

所以，当我玩游戏时，我常常会深入思考这一点。

So yeah, when I'm playing games, I'm often thinking about that quite a bit.

我一直在思考，作为学者和学术工作者，我与互联网以及知识之间的关系。

I've been thinking a lot about, I guess, like my relationship to the internet and to knowledge, I think as a scholar and an academic.

当我玩游戏时，我也是在以学者和学术工作者的身份进行游戏。

When I'm playing games, I'm also like, playing as a scholar and an academic.

所以我特别喜欢《艾尔登法环》这样的游戏，比如到处寻找特定物品来阅读它们的背景故事，这大概是世界上最宅的事情了。

So I love like, Elden Ring, for example, like going and finding specific objects so I could read their lore entries, which is like the dorkiest thing in the world.

玩《艾尔登法环》或《控制》这类游戏时，我经常这么做——在《控制》里，你身处一个被恶魔力量附身的超自然联邦调查局。

Playing a video game like, like Elden Ring or Control, I do this a lot where in Control, you're in this kind of like a supernatural FBI that's been possessed by this demonic force.

巧合的是，这个恶魔力量正在吟诵一首达达主义诗歌。

Incidentally, the demonic force is chanting a Dadaist poem.

它们悬浮在你头顶，吟诵着达达主义的诗篇。

They're like hovering over you and they're chanting Dadaist poetry.

所以这个游戏简直像是为我量身打造的。

So it was kind of like this game was built for me.

但在游戏中，还有许多被删减的文件。

But within it, there's all these redacted documents.

因此，尽管游戏希望我四处射击这些恶魔，我却不由自主地想去寻找并阅读这些空间里的文件。

And so like, if the game is wanting me to like go around and shoot these demons, I find myself drawn to like, going and finding documents and reading documents in these spaces.

这种感觉对我来说特别令人着迷。

Like there's something about that that's really thrilling to me.

我认为，这和学术研究最初的那种兴奋感是一样的——找到一份稀有的文献，并试图理解这份文献，以此来认识我所处的这个奇异世界。

I think that's the same thrill of scholarship from the very beginning, the idea of finding this rare document and trying to make sense of that document as a way to understand this bizarre world that I find myself in.

所以，我想这就是我一直寻找的那种主要感受。

So that's, I think, the prevailing feeling I've been looking for.

而且我越来越试图去体会我们刚才谈到的这种‘大型语言模型’的感觉，对吧？

And I think increasingly, I've been trying to like get to this feeling that we're just talking like this like, large language model feeling, right?

它正在摧毁这个世界。

Like, it is destroying the world.

它正在以我们难以理解的方式发生。

It is happening in ways that we have a really hard time understanding.

因此，我转向游戏，希望借此更好地理解这种感觉，以及其中可能蕴含的种种可能性。

And so I've been looking to games to try and help me understand a little bit better what that feels like and what kind of possibilities there are.

我可以讲一个最后的游戏趣事吗？

Can I give one final game anecdote?

最近，我一直在痴迷地思考和撰写关于《博德之门3》的内容。

So I've been obsessively thinking about and writing about Baldur's Gate three recently.