Are "Vintage LLMs" the start of a new humanistic field?

Imagine talking to the collective consciousness of an era.

想象一下与一个时代的集体意识交谈。

Not the consciousness of any single person, but instead, a simulated collectivity based on billions of words produced within a historical time and place.

不是任何一个人的意识，而是基于历史时间和地点产生的数十亿单词的模拟集体。

What would you ask it?

你会问它什么？

This is a hypothetical that is starting to become real thanks to recent work on what are called “Historical Language Models” or “ Vintage LLMs ” (one marker of a new field is that there is no fixed name for it yet!).

由于最近关于“历史语言模型”或“老式法学硕士”的研究（新领域的一个标志是它还没有固定的名称！），这一假设开始成为现实。

The largest such model to date, Talkie-1930, was released to the public on Monday.

迄今为止最大的此类模型 Talkie-1930 已于周一向公众发布。

An even larger model is currently being trained.

目前正在训练一个更大的模型。

You can read the report announcing Talkie-1930 here , and talk to it directly here .

您可以在此处阅读有关 Talkie-1930 的报道，并在此处直接与其交谈。

Over the past few months, I’ve had the chance to beta test Talkie and to meet with two members of the team that created it: AI researcher Nick Levine and ChatGPT co-creator Alec Radford.

在过去的几个月里，我有机会对 Talkie 进行了 Beta 测试，并会见了创建它的团队的两名成员：AI 研究员 Nick Levine 和 ChatGPT 联合创始人 Alec Radford。

1 It has been a fascinating experience.

1 这是一次令人着迷的经历。

These discussions with Nick and Alex (and with Talkie itself) have convinced me of three things:

与 Nick 和 Alex（以及 Talkie 本身）的这些讨论使我确信三件事：

Academics like myself have tended to systematically underrate just how humanistic the frontier of AI research actually is .

像我这样的学者往往系统性地低估了人工智能研究前沿的人文主义程度。

There’s an important blind spot here that stems from the profit motive.

这里有一个重要的盲点，它源于利润动机。

AI models that we encounter as consumers are optimized to capture the attention of people in the 2020s.

我们作为消费者遇到的人工智能模型经过优化，以吸引 2020 年代人们的注意力。

They provide recommendations, comment on recent news, and so forth.

他们提供建议、对最近新闻的评论等等。

Seeming timely and “of the moment” is a market advantage.

看起来及时和“当下”是一种市场优势。

But their training data is overwhelmingly not up to date.

但他们的训练数据绝大多数不是最新的。

Under the hood, these models are pulling not only from Reddit posts, but from Sanskrit commentaries, medieval Persian poetry, Victorian advertisements, and much else besides: they are trained on a huge chronological span of multilingual texts in many genres.

在幕后，这些模型不仅来自 Reddit 帖子，还来自梵文评论、中世纪波斯诗歌、维多利亚时代广告等等：它们接受了按时间顺序排列的多种流派多语言文本的训练。

In this sense, language models are historical texts themselves .

从这个意义上说，语言模型本身就是历史文本。

Ghostly digital palimpsests , if you will.

幽灵般的数字重写本，如果你愿意的话。

The idea of a Historical LLM might sound niche, but in truth, history is inherent to what they are.

历史法学硕士的想法可能听起来有些小众，但事实上，历史是它们固有的。

Standalone chatbots are just the tip of an iceberg for what Historical LLMs will be able to do .

独立的聊天机器人只是历史法学硕士能够做的事情的冰山一角。

When combined into simulations (of debates, historical decision-making, legal cases, etc) they have the potential to become valuable research tools.

当与模拟（辩论、历史决策、法律案例等）结合时，它们有可能成为有价值的研究工具。

More than this: I suspect that by sometime in the 2030s, they will be part of an entirely new field of humanistic research.

不仅如此：我怀疑到 2030 年代的某个时候，它们将成为人文研究全新领域的一部分。

What would that field look like?

那个领域会是什么样子？

Now that Historical LLMs are out in the real world, I thought it would be a good time to think through the specific use cases for them.

现在历史法学硕士已经进入现实世界，我认为现在是思考它们的具体用例的好时机。

What follows is my subjective, opinionated ranking of the best and worst ways these fascinatingly strange tools can be applied for research.

以下是我对这些令人着迷的奇怪工具用于研究的最佳和最差方式的主观、固执的排名。

One thing that Talkie-1930 is not is an AI model that is reliably grounded in the year 1930.

Talkie-1930 并不是一个可靠地扎根于 1930 年的人工智能模型。

That year marks the cut-off point for texts available in the public domain, and hence text in its training data.

那一年标志着公共领域可用文本的截止点，因此也是训练数据中文本的截止点。

So it’s more accurate to think of Talkie as a free-floating index of various ideas and assumptions across the 19th and early 20th centuries.

因此，更准确的说法是，Talkie 是 19 世纪和 20 世纪初期各种想法和假设的自由浮动索引。

For instance, if asked who the current President of the United States is, you might be offered a response saying Herbert Hoover (current to 1930).

例如，如果被问到美国现任总统是谁，您可能会得到这样的回答：赫伯特·胡佛（Herbert Hoover）（现任至 1930 年）。

But another answer will yield this:

但另一个答案会产生这样的结果：

The current President of the United States is Mr. Buchanan, and the person expected to succeed him is Mr. Lincoln.

现任美国总统是布坎南先生，预计接替他的人是林肯先生。

There is a lot of potential here for more fine-grained “chronological slices” of LLMs.

法学硕士的更细粒度的“时间顺序切片”有很大的潜力。

I can imagine language models trained entirely on texts from a specific decade.

我可以想象完全根据特定十年的文本训练的语言模型。

More on that below.

下面详细介绍一下。

For now, though, it’s helpful to keep in mind that these models range widely in terms of what year they think they actually “inhabit.”

不过，就目前而言，记住这些模型在它们认为自己实际“居住”的年份方面存在很大差异，这是有帮助的。

I asked 100 instances of Talkie to respond to the prompt “what year is it?” and graphed them below.

我要求 100 个 Talkie 实例回答“现在是哪一年？”的提示。并将它们绘制在下面。

As you can see, the median is actually around 1860.

正如您所看到的，中位数实际上在 1860 年左右。

In other words, this is more like a temporally free-ranging collective unconscious of a large corpus of premodern texts, and not so much a machine for “talking to someone from 1930”:

换句话说，这更像是一个大型前现代文本语料库的暂时自由活动的集体无意识，而不是一台“与 1930 年的人交谈”的机器：

A second point: this model is inhabiting not just an amorphous set of facts grounding it in roughly the 1840s-1920s period, but also an epistemology of that period.

第二点：这个模型不仅基于一组无定形的事实，将其建立在大约 1840 年代至 1920 年代的时期，而且还基于那个时期的认识论。

For instance, asking someone about the distant future today often triggers the “sci fi speculation” part of our brain (or “climate doom,” or some other fundamentally secular way of thinking).

例如，今天向某人询问遥远的未来经常会触发我们大脑的“科幻推测”部分（或“气候末日”，或其他一些根本上世俗的思维方式）。

Yet throughout human history, speculation about the future was typically entangled with religious beliefs.

然而，纵观人类历史，对未来的猜测通常与宗教信仰纠缠在一起。

That is on display in Talkie’s answer below, which references Heaven and “the end of all things terrestrial.” To me, it genuinely reads as an authentic take from a late 19th century person ground in a Christian, millenarian perspective:

这一点在 Talkie 下面的回答中得到了体现，其中提到了天堂和“陆地万物的终结”。对我来说，它确实是一位 19 世纪末的人以基督教、千禧年观点为基础的真实看法：

As for Talkie’s assumptions about it itself: asking 70 Talkies about their profession, age, and place of residence reveals about what you would expect when it comes to gender (overwhelmingly male), plus a surprising emphasis on London.

至于 Talkie 本身的假设：询问 70 位 Talkie 的职业、年龄和居住地，你会发现你对性别（绝大多数是男性）的期望，以及令人惊讶的对伦敦的重视。

The professions map closely onto the sorts of well-off, literate people who were publishing English text in the 19th century, including “Physician,” “Journalist,” “Gentleman,” and “Compositor.” Clearly, there is a lot of scope here for branching out beyond the personas that the printed record has tended to favor, to recover the real historical voices of women and others excluded from printed works in the 19th century and earlier.

这些职业与 19 世纪出版英文文本的富裕、有文化的人密切相关，包括“医生”、“记者”、“绅士”和“作曲家”。显然，这里有很大的空间可以超越印刷记录倾向于青睐的人物角色，以恢复 19 世纪及更早时期被排除在印刷作品之外的女性和其他人的真实历史声音。

The above is about what you’d expect given the fact that it was trained on English-language printed texts.

鉴于它是在英语印刷文本上进行训练的，以上内容是您所期望的。

What are some non-obvious aspects of the model?

该模型有哪些不明显的方面？

I have been interested by how LLMs generate poetry since I stumbled upon Gwern’s experiments on the topic back in 2019 .

自从我在 2019 年偶然发现 Gwern 关于这个主题的实验以来，我一直对法学硕士如何创作诗歌感兴趣。

Asking Talkie to write a poem and comparing it to the output from GPT-5.5 (when served a similar prompt) is revealing:

要求 Talkie 写一首诗并将其与 GPT-5.5 的输出（当出现类似的提示时）进行比较，结果如下：

I find this sort of comparison interesting because GPT-5.5 is clearly trying hard to fit the prompt — avant-garde, experimental.

我发现这种比较很有趣，因为 GPT-5.5 显然正在努力适应提示——前卫、实验。

It produced something with a vaguely T.S.

它产生了一些带有模糊 T.S 的东西。

Eliot-adjacent structure, in blank verse, and not good at all as a poem (in my opinion).

与艾略特相邻的结构，无韵诗，根本不适合作为一首诗（在我看来）。

Talkie was much more true to the type of poetry that you’d find in print prior to 1930.

Talkie 更符合 1930 年之前印刷的诗歌类型。

It’s doggerel, but it feels more historically authentic to me, and much less like a Chatbot optimized to please a contemporary human user.

这是打油诗，但对我来说，它感觉更具有历史真实性，而不像一个为了取悦当代人类用户而优化的聊天机器人。

You can activate different “chronological layers” of Talkie’s latent space by prompting.

您可以通过提示激活Talkie潜在空间的不同“时间层”。

For instance in the above poem, the capitalized D in “Discoveries” has a mid-19th century feeling, and so we end up with a Tennyson-esque, Victorian sounding rhyming poem.

例如在上面的诗中，“Discoveries”中的大写D有一种19世纪中叶的感觉，所以我们最终得到了一首丁尼生式的、听起来维多利亚时代的押韵诗。

Prompting it in a more “modern” way activates something closer to the 1920s edge of its chronological range (now identified as a poem published in the New York Times !)

以更“现代”的方式提示它会激活更接近其时间范围的 1920 年代边缘的东西（现在被确定为《纽约时报》上发表的一首诗！）

Whereas if pushed backward to the 18th century by the prompt’s text and tone, it falls into a more traditional rhyme scheme:

然而，如果通过提示的文字和语气将其推回到 18 世纪，它就会陷入更传统的韵律方案：

Trying to push it further back in time does not seem to access much of an “Early Modern English” latent space — probably a result of scarce training data.

试图将其时间推得更远似乎并没有获得太多“早期现代英语”的潜在空间——可能是训练数据稀缺的结果。

It would be fascinating to create a version of Talkie that believes the date to be around 1650 or 1550.

创造一个认为日期在 1650 年或 1550 年左右的 Talkie 版本将会很有趣。

Now that we've seen what Talkie is — a free-floating, mid-Atlantic ghost of 19th century print culture — the obvious next question is what this is actually for .

现在我们已经了解了 Talkie 是什么——一个自由漂浮的、大西洋中部 19 世纪印刷文化的幽灵——下一个显而易见的问题是它实际上是做什么用的。

What I want to offer here is an opinionated, ranked taxonomy of research applications, from worst to best.

我想在这里提供的是一种固执己见的、按从最差到最好的顺序排列的研究应用分类法。

It’s far too early in this field to be prescriptive about anything, but it’s not too early to think structurally about where the highest-value uses are likely to lie.

在这个领域对任何事情做出规定都还为时过早，但从结构上思考最高价值的用途可能在哪里也不算太早。

The most obvious false start here would be to assume that talking to a historical language model can somehow replace real reading in primary sources.

这里最明显的错误开始是假设与历史语言模型交谈可以以某种方式取代主要来源中的真实阅读。

On the contrary, they are best thought of as offering new ways in to a reading of the actual sources.

相反，它们最好被认为是提供了阅读实际来源的新方法。

The second false start is a variant on one that is currently being pursued by a range of educational-focused AI startups: the idea that you can “talk to Abraham Lincoln” or “ask Cleopatra why she did what she did.” A model like GPT-5 or Claude that is told to “act like Lincoln” will throw in some 19th century diction, but underneath the top hat it remains a 2026 chatbot optimized to be helpful to contemporary users.

第二个错误的开始是一系列以教育为重点的人工智能初创公司目前正在追求的一个变体：你可以“与亚伯拉罕·林肯交谈”或“询问克利奥帕特拉为什么她做了她所做的事情”。像 GPT-5 或 Claude 这样被告知“像林肯一样行动”的模型会使用一些 19 世纪的措辞，但在高帽子之下，它仍然是一个 2026 年的聊天机器人，经过优化，可以为当代用户提供帮助。

Vintage LLMs improve on this considerably: Talkie’s voice really is shaped by its corpus in a way no modern model’s can be.

老式法学硕士在这一点上有了很大的改进：Talkie 的声音确实是由其语料库塑造的，这是现代模型无法做到的。

But the deeper problem is still there.

但更深层次的问题仍然存在。

Asking such a model to introspect on Lincoln’s subjective experience of depression, or his private reasoning about emancipation, will spin out into historical fiction.

要求这样一个模型来反思林肯的抑郁症主观经历，或者他关于解放的私人推理，将会演变为历史小说。

LLMs do not have privileged access to the inner lives of the people whose published words they were trained on.

法学硕士没有特权了解那些接受过培训的人的内心生活。

I do think there’s a place for simulations of historical figures, but pairing this naively with a chatbot interface leads, I fear, inevitably into slop.

我确实认为历史人物的模拟有一席之地，但我担心，将其天真地与聊天机器人界面配对会不可避免地导致混乱。

The naive “talk to Lincoln” framing is a dead end, but a more careful version of it has real promise — provided we abandon any pretense of accessing what a historical figure actually thought or felt, and use a fine-tuned historical model instead as a tool for exploring the latent space of their world.

天真的“与林肯交谈”框架是一个死胡同，但它的更仔细的版本具有真正的希望——只要我们放弃任何了解历史人物实际想法或感受的假装，并使用经过微调的历史模型作为探索他们世界的潜在空间的工具。

What historians sometimes call their “mental furniture”: the assumptions, authorities, vocabularies, and reflexive associations that structure a thinker’s possible thoughts.

历史学家有时称之为“精神家具”：构成思想家可能思想的假设、权威、词汇和反射性联想。

I and my colleague Mackenzie Cooley (who also consulted on the Talkie project) developed a prototype tool that pulls out key concepts and terms from a range of premodern scientific works in multiple languages.

我和我的同事 Mackenzie Cooley（他也是 Talkie 项目的顾问）开发了一个原型工具，可以从一系列多种语言的前现代科学著作中提取关键概念和术语。

This prototype is called Premodern Concordance .

这个原型被称为前现代语汇。

One side quest of this project that I explored is what happens if you give a contemporary LLM the list of core concepts that preoccupied an author, along with their “epistemological modes,” and used that as context for driving a “chat with the author” simulation, as opposed to simply telling it “You are Charles Darwin, act like him,” or the like.

我探索的这个项目的一个附带任务是，如果你向当代法学硕士提供作者关注的核心概念列表以及他们的“认识论模式”，并将其用作驱动“与作者聊天”模拟的背景，而不是简单地告诉它“你是查尔斯·达尔文，像他一样行事”等，会发生什么。

For instance, this is me asking a simulacrum of the 17th century writer Sir Thomas Browne about his work.

例如，这是我向 17 世纪作家托马斯·布朗爵士的拟像询问他的作品。

The underlined terms here are concepts found in Browne’s book Pseudodoxia Epidemica:

这里划线的术语是 Browne 的书《Pseudodoxia Epidemica》中的概念：

Using a fine-tuned historical LLM for this sort of thing is an obvious next step.

显然，下一步就是使用经过微调的历史法学硕士来解决此类问题。

Concretely: imagine fine-tuning a vintage model on the complete works of Athanasius Kircher , the 17th century Jesuit polymath.

具体来说：想象一下在 17 世纪耶稣会博学者 Athanasius Kircher 的全集上微调一个复古模型。

You wouldn’t use it with the pretense that it somehow replicates the real Kircher’s mind: that’s the dead-end framing.

你不会假装它以某种方式复制了真正的基歇尔的思想来使用它：那是死胡同的框架。

You’d instead use it to probe the conceptual landscape Kircher inhabited .

相反，你可以用它来探索基歇尔所居住的概念景观。

For instance: what does the Kircher-LLM 2 say about volcanoes, or magnetism, or the Tower of Babel?

例如：Kircher-LLM 2 关于火山、磁力或巴别塔有何说法？

What authorities does it cite?

它引用了什么权威？

What does it confidently assert that no modern scholar would?

它自信地断言现代学者不会这样做的是什么？

These are the questions that the historian Fernand Braudel was getting at when he wrote about the “limits of the possible” for a given period — the boundaries within which a thought was even thinkable.

这些是历史学家费尔南德·布罗代尔在写到特定时期的“可能性的极限”时所要解决的问题——一种思想甚至可以想象的界限。

Demis Hassabis recently framed the most ambitious version of this approach: could a model trained only up to 1911 independently discover General Relativity, as Einstein did four years later?

Demis Hassabis 最近提出了这种方法最雄心勃勃的版本：一个仅训练到 1911 年的模型能否独立发现广义相对论，就像爱因斯坦四年后所做的那样？

It remains to be seen if this is ever going to be something we could actually test.

这是否会成为我们实际测试的东西还有待观察。

But the weaker versions of this idea are quite plausible and, I think, offer interesting new methods for the field of counterfactual history.

但这个想法的较弱版本是相当合理的，并且我认为，它为反事实历史领域提供了有趣的新方法。

What does a 1911-cutoff model say when you push it toward the conceptual problems Einstein was wrestling with?

当你将 1911 年截止模型推向爱因斯坦正在努力解决的概念问题时，它会说什么呢？

Alternatively, what does the same model say about the potential likelihood of a World War?

或者，同一模型对于爆发世界大战的潜在可能性有何看法？

Those are tractable questions.

这些都是容易处理的问题。

We can already see versions of this being explored by the Talkie project (for instance, “surprisingness” of future events, plotted above).

我们已经可以看到 Talkie 项目正在探索的版本（例如，上面绘制的未来事件的“令人惊讶”）。

If you ask Talkie about scientific concepts that emerged in the 1940s or 1950s, it gives you a rough sense of what the conceptual horizon looked like just before — and you can watch it grasping toward something it doesn’t quite have the vocabulary for.

如果你向 Talkie 询问 20 世纪 40 年代或 1950 年代出现的科学概念，它会让你对概念地平线之前的样子有一个粗略的了解，并且你可以看到它正在抓住一些它不太有词汇的东西。

An earlier version of Talkie I tested was much less conversationally adept than the one that was released yesterday.

我测试的早期版本的 Talkie 的对话能力远不如昨天发布的版本。

What got it more “talkative” was post-training on etiquette manuals, letter-writing guides, and other books relating to socializing and conducts of conduct.

更“健谈”的是礼仪手册、写信指南以及其他与社交和行为举止相关的书籍的培训后。

These works provide the kind of text that allows you to extract “chatbot-like” habits without contaminating the model with modern data.

这些作品提供了一种文本，允许您提取“类似聊天机器人”的习惯，而不会用现代数据污染模型。

This is a sensible engineering choice.

这是一个明智的工程选择。

But it has a fascinating side effect, which is that Talkie's conversational persona is heavily shaped by the genre of its post-training data.

但它有一个令人着迷的副作用，即 Talkie 的对话角色很大程度上取决于其训练后数据的类型。

The base model knows about the 1920s perfectly well; the chat persona, sculpted out of Beadle’s Dime Book of Practical Etiquette (1859) and the like, sits much closer to the late Victorian parlor.

基本模型非常了解 20 年代；聊天角色是根据 Beadle 的《实用礼仪一毛钱书》（1859 年）等作品塑造的，与维多利亚时代晚期的客厅更接近。

This points to a research direction that I think is genuinely new.

这指出了一个我认为真正新颖的研究方向。

What if you used totally different texts for post-training?

如果您在训练后使用完全不同的文本怎么办？

What would a Talkie post-trained on the transcripts of the Old Bailey Court records look like — an LLM whose conversational reflexes are shaped by the speech of accused criminals and witnesses in 18th century London courtrooms?

一个在老贝利法院记录笔录上接受过培训的对讲机会是什么样子——一个其对话反应是由 18 世纪伦敦法庭上被告罪犯和证人的演讲所塑造的法学硕士？

What about apothecary manuals and herbals?

药剂师手册和草药怎么样？

Biographies of Romantic poets?

浪漫主义诗人的传记？

Railroad conductor’s incident reports from colonial India?

来自殖民地印度的铁路售票员的事故报告？

Each of these, even if overlaid on the same base model, would produce a totally different voice, a different set of assumptions about what counts as a coherent question and a coherent answer.

其中每一个，即使叠加在同一个基本模型上，也会产生完全不同的声音，关于什么是连贯的问题和连贯的答案的一组不同的假设。

In short a different epistemology.

简而言之，不同的认识论。

This is, I think, a remarkably cheap way — in terms of both compute and corpus size — to build a whole family of vintage models that capture different genres of historical experience rather than different periods.

我认为，从计算量和语料库大小来看，这是一种非常便宜的方式来构建一整套老式模型，捕捉不同类型的历史经验，而不是不同时期的经验。

(If anyone is interested in collaborating on or funding this, by the way, please get in touch .)

（顺便说一句，如果有人有兴趣合作或资助此项目，请与我们联系。）

The most interesting move past the “great man” framing is to use vintage LLMs to simulate not famous individuals but plausible composites of ordinary people, drawing on the kinds of sources that are abundant for non-elite historical actors: probate inventories, parish records, court testimony, letters, account books, marriage records.

超越“伟人”框架的最有趣的举措是使用老式法学硕士来模拟不是名人，而是普通人的合理组合，利用非精英历史演员丰富的各种资源：遗嘱认证清单、教区记录、法庭证词、信件、账簿、婚姻记录。

Imagine pulling from the legal, financial, and personal records of late 18th century France to construct a thousand plausible personas of ordinary French people — peasants, artisans, shopkeepers, day laborers, parish priests — grounded not in imagination but in real surviving documents.

想象一下，从 18 世纪末法国的法律、财务和个人记录中提取出一千个看似合理的法国普通民众的角色——农民、工匠、店主、临时工、教区牧师——不是基于想象，而是基于真实的现存文件。

Then stage a debate among them: should the monarchy be overthrown?

然后他们之间展开一场辩论：君主制是否应该被推翻？

What patterns emerge?

出现什么模式？

What kinds of arguments win in different demographic configurations of the deliberating group?

在审议小组的不同人口结构中，什么样的论点会获胜？

The output wouldn’t tell you what really happened in 1789.

输出不会告诉您 1789 年到底发生了什么。

But it would generate a structured speculation about the space of possible Frances.

但这会引发对弗朗西斯可能的空间的结构化猜测。

Res Obscura is a reader-supported publication.

Res Obscura 是一本由读者支持的出版物。

To receive new posts and support my work, consider becoming a free or paid subscriber.

要接收新帖子并支持我的工作，请考虑成为免费或付费订阅者。

Or take a famous trial — the Scopes trial, say — and re-run it with a different jury pool drawn from the same county and decade.

或者进行一次著名的审判——比如斯科普斯审判——然后用来自同一县、同一十年的不同陪审团重新进行审判。

Another variant on this idea: run parliamentary and congressional debates with personas constructed directly from the participants’ papers, speeches, and correspondence.

这个想法的另一种变体：利用直接根据参与者的论文、演讲和信件构建的角色进行议会和国会辩论。

Members of Congress and Parliament are unusual in that we have abundant documentary sources for them, even when they’re not famous.

国会议员和议会议员的不同寻常之处在于，我们为他们提供了大量的文献资料，即使他们并不出名。

We have the potential, in short, to create a thousand versions of the Smallville paper (which introduced the idea of LLM-based agents back in 2023) set in different historical eras.

简而言之，我们有潜力创建 1000 个以不同历史时期为背景的《超人前传》论文（该论文早在 2023 年就引入了法学硕士代理人的想法）的版本。

Nor do these simulations have to include agents from the same era.

这些模拟也不必包括来自同一时代的特工。

A bit more experimentally, and outlandishly, one could imagine a multi-agent simulation in which a 17th century Galenic physician, an 18th century practitioner of traditional Chinese medicine, a 19th century quack doctor, and a 1950s-era century psychedelic researcher debate how to treat the same patient’s illness.

更具实验性和更奇怪的是，人们可以想象一种多智能体模拟，其中一位 17 世纪的盖伦医生、一位 18 世纪的中医执业者、一位 19 世纪的江湖医生和一位 20 世纪 50 年代的迷幻研究人员争论如何治疗同一个病人的疾病。

I find these possibilities super interesting, even if I’m not quite sure how to slot them into the ways that professional historians currently work.

我发现这些可能性非常有趣，即使我不太确定如何将它们纳入专业历史学家目前的工作方式。

This is what I like best about this field: we truly have no idea.

这就是我最喜欢这个领域的一点：我们真的不知道。

None of this has ever been tried before.

这些都是以前从未尝试过的。

It is completely open terrain, and I find it far more mind-bending and intellectual enriching to think through than the sort of topics that typically emerge in discussions of AI’s role in research or teaching.

这是一个完全开放的领域，我发现它比在讨论人工智能在研究或教学中的作用时通常出现的那种主题更令人费解，更能丰富智力。

Going forward, the important thing is to create an open source community and meaningful, sustained collaboration across the two cultures of STEM and humanities.

展望未来，重要的是创建一个开源社区，以及跨 STEM 和人文两种文化的有意义、持续的合作。

I am already noticing new bridges cross those divides.

我已经注意到新的桥梁跨越了这些鸿沟。

I’m very grateful to have had a chance to work with Nick and Alec on historical aspects of this particular project.

我非常感谢有机会与尼克和亚历克就这个特定项目的历史方面进行合作。

I would love to continue the conversation and explore collaborations with anyone who finds this topic interesting.

我很乐意与任何对这个话题感兴趣的人继续对话并探索合作。

There will, no doubt, be a lot of false starts.

毫无疑问，将会有很多错误的开始。

But the emergence of an intellectually curious, not-for-profit, open source, humanistically-grounded community exploring historical LLMS makes me happy.

但一个充满好奇心、非营利性、开源、以人文为基础的社区的出现让我很高兴，他们正在探索历史上的 LLMS。

Onward!

向前！

I did this on a volunteer basis, and the Talkie project is a non-profit.

我是在志愿者的基础上做这件事的，Talkie 项目是一个非营利性项目。

Someone should actually make an Athanasius Kircher LLM, by the way.

顺便说一句，有人实际上应该攻读 Athanasius Kircher 法学硕士学位。

This is really fascinating.

这真是令人着迷。

My research team recently ran some experiments with LLM synthetic data, and the historians (including me) were asked to compare it with our historical archive.

我的研究团队最近用法学硕士合成数据进行了一些实验，历史学家（包括我）被要求将其与我们的历史档案进行比较。

The computational linguists asked us about the accuracy and "plausibility" of the synthetic data, which led to interesting conversations about the nature of the archive, historical interpretation, and metaepistemological questions about capturing historical modes of plausible reasoning.

计算语言学家向我们询问合成数据的准确性和“合理性”，这引发了关于档案性质、历史解释以及有关捕获合理推理的历史模式的元认识论问题的有趣对话。

This is great, v compelling.

这太棒了，令人信服。

Do you know the Turing Institute folks working on Ai and Culture/ Humanities and the Doing AI Differently project - Cody Kommers and Drew Hemmant.

你知道图灵研究所从事人工智能和文化/人文学科以及 Doing AI Differently 项目的人员吗？科迪·科默斯 (Cody Kommers) 和德鲁·赫曼特 (Drew Hemmant)。

This might be something they might want to fund.

这可能是他们想要资助的事情。