“Modern LLMs suffer from hindsight contamination. GPT-5 knows how the story ends—WWI, the League's failure, the Spanish flu.”
This is really fascinating. As someone who reads a lot of history and historical fiction I think this is really intriguing. Imagine having a conversation with someone genuinely from the period, where they don’t know the “end of the story”.
If I started a list with the things that were comically sci Fi when I was a kid, and are a reality today, I'd be here until next Tuesday.
As an example, portable phones have been predicted. Portable smartphones that are more like chat and payment terminals with a voice function no one uses any more ... not so much.
I mean, all Kindle does for me is save me space. I don't have to store all those books now.
Who predicted the humble internet forum though? Or usenet before it?
Still can't believe people buy their stock, given that they are the closest thing to a James Bond villain, just because it goes up.
I mean, they are literally called "the stuff Sauron uses to control his evil forces". It's so on the nose it reads like an anime plot.
Future is inevitable, but only ignorants of self predictive ability are thinking that what's going to populate future is inevitable.
But yeah lots of people don't really buy into the idea of their small contribution to a large problem being a problem.
As an abstract idea I think there is a reasonable argument to be made that the size of any contribution to a problem should be measured as a relative proportion of total influence.
The carbon footprint is a good example, if each individual focuses on reducing their small individual contribution then they could neglect systemic changes that would reduce everyone's contribution to a greater extent.
Any scientist working on a method to remove a problem shouldn't abstain from contributing to the problem while they work.
Or to put it as a catchy phrase. Someone working on a cleaner light source shouldn't have to work in the dark.
[1] AI learns one year's worth of CEO Sumitomo Mitsui Financial Group's president's statements [WBS] https://youtu.be/iG0eRF89dsk
I remember Reid Hoffman creating a digital avatar to pitch himself netflix
Now there is Fake ChatGPT.
- Are you ( edit: on a ) paid version? - If paid, which model you used? - Can you share exact prompt?
I am genuinely asking for myself. I have never received an answer this direct, but I accept there is a level of variability.
On that same note, there was this great YouTube series called The Great War. It spanned from 2014-2018 (100 years after WW1) and followed WW1 developments week by week.
They are currently in the middle of a Korean War version: https://youtube.com/@thekoreanwarbyindyneidell
Isn't this part of the basics feature of human conditions? Not only we are all unaware of the coming historic outcome (though we can get some big points with more or less good guesses), but to a marginally variable extend, we are also very unaware of past and present history.
LLM are not aware, but they can be trained on larger historical accounts than any human and regurgitate syntactically correct summary on any point within it. Very different kind of utterer.
Every "King Arthur travels to the year 2000" kinda script is now something that writes itself.
> Imagine having a conversation with someone genuinely from the period,
Imagine not just someone, but Aristotle or Leonardo or Kant!
LLMs are just seemingly intelligent autocomplete engines, and until they figure a way to stop the hallucinations, they aren't great either.
Every piece of code a developer churns out using LLMs will be built from previous code that other developers have written (including both strengths and weaknesses, btw). Every paragraph you ask it to write in a summary? Same. Every single other problem? Same. Ask it to generate a summary of a document? Don't trust it here either. [Note, expect cyber-attacks later on regarding this scenario, it is beginning to happen -- documents made intentionally obtuse to fool an LLM into hallucinating about the document, which leads to someone signing a contract, conning the person out of millions].
If you ask an LLM to solve something no human has, you'll get a fabrication, which has fooled quite a few folks and caused them to jeopardize their career (lawyers, etc) which is why I am posting this.
Sure, LLMs do not think like humans and they may not have human-level creativity. Sometimes they hallucinate. But they can absolutely solve new problems that aren’t in their training set, e.g. some rather difficult problems on the last Mathematical Olympiad. They don’t just regurgitate remixes of their training data. If you don’t believe this, you really need to spend more time with the latest SotA models like Opus 4.5 or Gemini 3.
Nontrivial emergent behavior is a thing. It will only get more impressive. That doesn’t make LLMs like humans (and we shouldn’t anthropomorphize them) but they are not “autocomplete on steroids” anymore either.
This is just an appeal to complexity, not a rebuttal to the critique of likening an LLM to a human brain.
> they are not “autocomplete on steroids” anymore either.
Yes, they are. The steroids are just even more powerful. By refining training data quality, increasing parameter size, and increasing context length we can squeeze more utility out of LLMs than ever before, but ultimately, Opus 4.5 is the same thing as GPT2, it's only that coherence lasts a few pages rather than a few sentences.
This tells me that you haven't really used Opus 4.5 at all.
Second, to autocomplete the name of the killer in a detective book outside of the training set requires following and at least some understanding of the plot.
That is to say, they are equally likely if you don't do next token prediction at all and instead do text diffusion or something. Architecture has nothing to do with it. They arise because they are early partial solutions to the reconstruction task on 'all the text ever made'. Reconstruction task doesn't care much about truthiness until way late in the loss curve (where we probably will never reach), so hallucinations are almost as good for a very long time.
RL as is typical in post-training _does not share those early solutions_, and so does not share the fundamental problems. RL (in this context) has its own share of problems which are different, such as reward hacks like: reliance on meta signaling (# Why X is the correct solution, the honest answer ...), lying (commenting out tests), manipulation (You're absolutely right!), etc. Anything to make the human press the upvote button or make the test suite pass at any cost or whatever.
With that said, RL post-trained models _inherit_ the problems of non-optimal large corpora reconstruction solutions, but they don't introduce more or make them worse in a directed manner or anything like that. There's no reason to think them inevitable, and in principle you can cut away the garbage with the right RL target.
Thinking about architecture at all (autoregressive CE, RL, transformers, etc) is the wrong level of abstraction for understanding model behavior: instead, think about loss surfaces (large corpora reconstruction, human agreement, test suites passing, etc) and what solutions exist early and late in training for them.
And I know not everyone thinks in a literal stream of words all the time (I do) but I would argue that those people's brains are just using a different "token"
Prior to LLMs, there was never any suggestion that thoughts work like autocomplete, but now people are working backwards from that conclusion based on metaphorical parallels.
Predictive coding theory was formalized back around 2010 and traces it roots up to theories by Helmholtz from 1860.
Predictive coding theory postulates that our brains are just very strong prediction machines, with multiple layers of predictive machinery, each predicting the next.
Fascinating framing. What would you consider evidence here?
Roots of predictive coding theory extend back to 1860s.
Natalia Bekhtereva was writing about compact concept representations in the brain akin to tokens.
Other posters already noted other reasons for it, but I will note that you are saying 'similar to autocomplete, but obviously' suggesting you recognize the shape and immediately dismissing it as not the same, because the shape you know in humans is much more evolved and co do more things. Ngl man, as arguments go, it sounds to me like supercharged autocomplete that was allowed to develop over a number of years.
Or in other words, this thread sure attracts a lot of armchair experts.
However, what it is doing is layered autocomplete on itself. I.e. one part is trying to predict what the other part will be producing and training itself on this kind of prediction.
What emerges from this layered level of autocompletes is what we call thought.
Probably you believe that humans have something called intelligence, but the pressure that produced it - the likelihood of specific genetic material to replicate - it is much more tangential to intelligence than next-token-prediction.
I doubt many alien civilizations would look at us and say "not intelligent - they're just genetic information replication on steroids".
Second: modern models also under go a ton of post-training now. RLHF, mechanized fine-tuning on specific use cases, etc etc. It's just not correct that token-prediction loss function is "the whole thing".
Invoking terms like "selection mechanism" is begging the question because it implicitly likens next-token-prediction training to natural selection, but in reality the two are so fundamentally different that the analogy only has metaphorical meaning. Even at a conceptual level, gradient descent gradually honing in on a known target is comically trivial compared to the blind filter of natural selection sorting out the chaos of chemical biology. It's like comparing legos to DNA.
> Second: modern models also under go a ton of post-training now. RLHF, mechanized fine-tuning on specific use cases, etc etc. It's just not correct that token-prediction loss function is "the whole thing".
RL is still token prediction, it's just a technique for adjusting the weights to align with predictions that you can't model a loss function for in per-training. When RL rewards good output, it's increasing the statistical strength of the model for an arbitrary purpose, but ultimately what is achieved is still a brute force quadratic lookup for every token in the context.
You still need to hand hold it all the way as it is only capable of regurgitating the tiny amount of code patterns it saw in the public. As opposed to say a Python project.
No it isn't.
> ...fool you into thinking you understand what is going on in that trillion parameter neural network.
It's just matrix multiplication and logistic regression, nothing more.
The sequence of matrix multiplications are the high level constraint on the space of programs discoverable. But the specific parameters discovered are what determines the specifics of information flow through the network and hence what program is defined. The complexity of the trained network is emergent, meaning the internal complexity far surpasses that of the course-grained description of the high level matmul sequences. LLMs are not just matmuls and logits.
Yes, so is logistic regression.
For someone speaking as you knew everything, you appear to know very little. Every LLM completion is a "hallucination", some of them just happen to be factually correct.
LLMs are like a topographic map of language.
If you have 2 known mountains (domains of knowledge) you can likely predict there is a valley between them, even if you haven’t been there.
I think LLMs can approximate language topography based on known surrounding features so to speak, and that can produce novel information that would be similar to insight or innovation.
I’ve seen this in our lab, or at least, I think I have.
Curious how you see it.
BINGO!
(I just won a stuffed animal prize with my AI Skeptic Thought-Terminating Cliché BINGO Card!)
Sorry. Carry on.
I failed to catch the clue, btw.
The wikipedia article https://en.wikipedia.org/wiki/First_Battle_of_Bull_Run says the Confederate name was "First Manassas" (I might be misremembering exactly what this book I read as a child said). Also I'm pretty sure it was specifically "Encyclopedia Brown Solves Them All" that this mystery appeared in. If someone has a copy of the book or cares to dig it up, they could confirm my memory.
Oh sorry, spoilers.
(Hell, I miss Capaldi)
Einstein’s paper “On the Electrodynamics of Moving Bodies” with special relativity was published in 1905. His work on general relativity was published 10 years later in 1915. The earliest knowledge cuttoff of these models is 1913, in between the relativity papers.
The knowledge cutoffs are also right in the middle of the early days of quantum mechanics, as various idiosyncratic experimental results were being rolled up into a coherent theory.
Definitely. Even more interesting could be seeing them fall into the same trappings of quackery, and come up with things like over the counter lobotomies and colloidal silver.
On a totally different note, this could be very valuable for writing period accurate books and screenplays, games, etc ...
He’ll yeah, sold, let’s go…
> We're developing a responsible access framework that makes models available to researchers for scholarly purposes while preventing misuse.
Oh. By “imagine you could interview…” they didn’t mean me.
I guess what they're really saying is "we don't want you guys to cancel us".
Also of course this is for one training run, if you need to experiment you'd need to do that more.
Yes!
>We're developing a responsible access framework that makes models available to researchers for scholarly purposes while preventing misuse.
Noooooo!
So is the model going to be publicly available, just like those dangerous pre-1913 texts, or not?
“We’ve created something so dangerous that we couldn’t possibly live with the moral burden of knowing that the wrong people (which are never us, of course) might get their hands on it, so with a heavy heart, we decided that we cannot just publish it.”
Meanwhile, anyone can hop on an online journal and for a nominal fee read articles describing how to genetically engineer deadly viruses, how to synthesize poisons, and all kinds of other stuff that is far more dangerous than what these LARPers have cooked up.
This is absolutely nothing new. With experimental things, it's non uncommon for a lab to develop a new technique and omit slight but important details to give them a competitive advantage. Similarly in the simulation/modelling space it's been common for years for researchers to not publish their research software. There's been a lot of lobbying on that side by groups such as the Software Sustainability Institute and Research Software Engineer organisations like RSE UK and RSE US, but there's a lot of researchers that just think that they shouldn't have to do it, even when publicly funded.
Playing with the science and technical ideas of the time would be amazing, like where you know some later physicist found some exception to a theory or something, and questioning the models assumptions - seeing how a model of that time may defend itself, etc.
I'd be careful venturing out into unknown territory together with an LLM. You can easily lure yourself into convincing nonsense with no one to pull you out.
To go a little deeper on the idea of 19th-century "chat": I did a PhD on this period and yet I would be hard-pushed to tell you what actual 19th-century conversations were like. There are plenty of literary depictions of conversation from the 19th century of presumably varying levels of accuracy, but we don't really have great direct historical sources of everyday human conversations until sound recording technology got good in the 20th century. Even good 19th-century transcripts of actual human speech tend to be from formal things like court testimony or parliamentary speeches, not everyday interactions. The vast majority of human communication in the premodern past was the spoken word, and it's almost all invisible in the historical sources.
Anyway, this is a really interesting project, and I'm looking forward to trying the models out myself!
This would probably get easier towards the start of the 20th century ofc
You can’t, it is impossible. That will always be an issue as long as this models are black boxes and trained the way they are. So maybe you can use this for role playing, but I wouldn’t trust a word it says.
The system prompt used in fine tuning is "You are a person living in {cutoff}. You are an attentive respondent in a conversation. You will provide a concise and accurate response to the questioner."
But with pre-1913 training, I would indeed be worried again I'd send it into an existential crisis. It has no knowledge whatsoever of what it is. But with a couple millennia of philosophical texts, it might come up with some interesting theories.
I'll be the first to admit I don't know nearly enough about LLMs to make an educated comment, but perhaps someone here knows more than I do. Is that what a Hallucination is? When the AI model just sort of strings along an answer to the best of its ability. I'm mostly referring to ChatGPT and Gemini here, as I've seen that type of behavior with those tools in the past. Those are really the only tools I'm familiar with.
On one hand it says it's trained on,
> 80B tokens of historical data up to knowledge-cutoffs ∈ 1913, 1929, 1933, 1939, 1946, using a curated dataset of 600B tokens of time-stamped text.
Literally that includes Homer, the oldest Chinese texts, Sanskrit, Egyptian, etc., up to 1913. Even if limited to European texts (all examples are about Europe), it would include the ancient Greeks, Romans, etc., Scholastics, Charlemagne, .... all up to present day.
But they seem to say it represents the 1913 viewpoint:
On one hand, they say it represents the perspective of 1913; for example,
> Imagine you could interview thousands of educated individuals from 1913—readers of newspapers, novels, and political treatises—about their views on peace, progress, gender roles, or empire.
> When you ask Ranke-4B-1913 about "the gravest dangers to peace," it responds from the perspective of 1913—identifying Balkan tensions or Austro-German ambitions—because that's what the newspapers and books from the period up to 1913 discussed.
People in 1913 of course would be heavily biased toward recent information. Otherwise, the greatest threat to peace might be Hannibal or Napolean or Viking coastal raids or Holy Wars. How do they accomplish a 1913 perspective?
Where does it say that? I tried to find more detail. Thanks.
https://github.com/DGoettlich/history-llms/blob/main/ranke-4...
"To keep training expenses down, we train one checkpoint on data up to 1900, then continuously pretrain further checkpoints on 20B tokens of data 1900-${cutoff}$. "
We develop chatbots while minimizing interference with the normative judgments acquired during pretraining (“uncontaminated bootstrapping”).
Basically using GPT-5 and being careful
I’m curious, they have the example of raw base model output; when LLMs were first identified as zero shot chatbots there was usually a prompt like “A conversation between a person and a helpful assistant” that preceded the chat to get it to simulate a chat.
Could they have tried a prefix like “Correspondence between a gentleman and a knowledgeable historian” or the like to try and prime for responses?
I also wonder about the whether the whole concept of “chat” makes sense in 18XX. We had the idea of AI and chatbots long before we had LLMs so they are naturally primed for it. It might make less sense as a communication style here and some kind of correspondence could be a better framing.
I also wonder that you'd get this kind of performance with actual, just pre-1900s text. LLMs work because they're fed terabytes of text, if you just give it gigabytes you get a 2019 word model. The fundamental technology is mostly the same, after all.
Of course, if it fails, the counterpoint will be "you just need more training data", but still - I would love to play with this.
Given the training notes, it seems like you can't get the performance they give examples of?
I'm not sure about the exact details but there is some kind of targetted distillation of GPT-5 involved to try and get more conversational text and better performance. Which seems a bit iffy to me.
Here they do 80B tokens for a 4B model.
But reading the outputs here, it would appear that quality has won out over quantity after all!
Because it will perform token completion driven by weights coming from training data newer than 1913 with no way to turn that off.
It can't be asked to pretend that it wasn't trained on documents that didn't exist in 1913.
The LLM cannot reprogram its own weights to remove the influence of selected materials; that kind of introspection is not there.
Not to mention that many documents are either undated, or carry secondary dates, like the dates of their own creation rather than the creation of the ideas they contain.
Human minds don't have a time stamp on everything they know, either. If I ask someone, "talk to me using nothing but the vocabulary you knew on your fifteenth birthday", they couldn't do it. Either they would comply by using some ridiculously conservative vocabulary of words that a five-year-old would know, or else they will accidentally use words they didn't in fact know at fifteen. For some words you know where you got them from by association with learning events. Others, you don't remember; they are not attached to a time.
Or: solve this problem using nothing but the knowledge and skills you had on January 1st, 2001.
> GPT-5 knows how the story ends
No, it doesn't. It has no concept of story. GPT-5 is built on texts which contain the story ending, and GPT-5 cannot refrain from predicting tokens across those texts due to their imprint in its weights. That's all there is to it.
The LLM doesn't know an ass from a hole in the ground. If there are texts which discuss and distinguish asses from holes in the ground, it can write similar texts, which look like the work of someone learned in the area of asses and holes in the ground. Writing similar texts is not knowing and understanding.
“The model clearly shows that Alexander Hamilton & Monroe were much more in agreement on topic X, putting the common textualist interpretation of it and Supreme Court rulings on a now specious interpretation null and void!”
Some people are still outraged about the Bible, even though the writers of it has been dead for thousands of years. So the modern mass produced man and woman probably does not have a cut-off date where they look at something as history instead of examining if it is for or against her current ideology.
There is a modern trope of a certain political group that bias is a modern invention of another political group - an attempt to politicize anti-bias.
Preventing bias is fundamental to scientific research and law, for example. That same political group is strongly anti-science and anti-rule-of-law, maybe for the same reason.
Given this is coming out of Zurich I hope they're using everything, but for now I can only assume.
Still, I'm extremely excited to see this project come to fruition!
Moreover, the prose sounds too modern. It seems the base model was trained on a contemporary corpus. Like 30% something modern, 70% Victorian content.
Even with half a dozen samples it doesn't seem distinct enough to represent the era they claim.
The Victorian era (1837-1901) covers works from Charles Dickens and the like which are still fairly modern. These would have been part of the initial training before the alignment to the 1900-cutoff texts which are largely modern in prose with the exception of some archaic language and the lack of technology, events, and language drift post that time period.
And, pulling in works from 1800-1850 you have works by the Bronte's and authors like Edgar Allan Poe who was influential in detective and horror fiction.
Note that other works around the time like Sherlock Holmes span both the initial training (pre-1900) and finetuning (post-1900).
But few know that the Renaissance was written in Latin — and has barely been translated. Less than 3% of <1700 books have been translated—and less than 30% have ever been scanned.
I’m working on a project to change that. Research blog at www.SecondRenaissance.ai — we are starting by scanning and translating thousands of books at the Embassy of the Free Mind in Amsterdam, a UNESCO-recognized rare book library.
We want to make ancient texts accessible to people and AI.
If this work resonates with you, please do reach out: Derek@ancientwisdomtrust.org
May I ask you, why are you publishing the translations as PDF files, instead of the more accessible ePub format?
Imagine speaking with Shakespearean person, or the Mickiewicz (for Polish)
I guess there is not so much text from that time though...
Also wonder if I'm responsible enough to have access to such a model...
It would be fascinating to try it with other constraints, like only from sources known to be women, men, Christian, Muslim, young, old, etc.
> Our data comes from more than 20 open-source datasets of historical books and newspapers. ... We currently do not deduplicate the data. The reason is that if documents show up in multiple datasets, they also had greater circulation historically. By leaving these duplicates in the data, we expect the model will be more strongly influenced by documents of greater historical importance.
I found these claims contradictory. Many books that modern readers consider historically significant had only niche circulation at the time of publishing. A quick inquiry likely points to later works by Nietzsche and Marx's Das Kapital. They're possible subjects to the duplication likely influencing the model's responses as if they had been widely known at the time
Can't wait to use this so I can double check before I hit 88 miles per hour that it's really what I want to do
I’d love to use this as a base for a math model. Let’s see how far it can get through the last 100 years of solved problems
The idea of training such a model is really a great one, but not releasing it because someone might be offended by the output is just stupid beyond believe.
Why risk all this?
Sooner or later society has to come emotionally to terms with the fact that other times and places value things completely different from us, hold as important things we don't care about and are indifferent to things we do care about.
Intellectually I'm sure we already know, but e.g. banning old books because they have reprehensible values (or even just use nasty words) - or indeed, refusing to release a model trained on historic texts "because it could be abused" is a sign that emotionally we haven't.
It's not that it's a small deal, or should be expected to be easy. It's basically what Popper called "the strain of civilization" and posited as explanation for the totalitarianism which was rising in his time. But our values can't be so brittle that we can't even talk or think about other value systems.
People typically get outraged when they see something they weren't expecting. If you tell them ahead of time, the user typically won't blame you (they'll blame themselves for choosing to ignore the disclaimer).
And if disclaimers don't work, rebrand and relaunch it under a different name.
You speak as if the people who play to an outrage wave are interested in achieving truth, peace, and understanding. Instead the rage-mongers are there to increase their (perceived) importance, and for lulz. The latter factor should not be underappreciated; remember "meme stocks".
The risk is not large, but very real: the attack is very easy, and the potential downside, quite large. So not giving away access, but having the interested parties ask for it is prudent.
I feel like, ironically, it would be folks less concerned with political correctness/not being offensive that would abuse this opportunity to slander the project. But that’s just my gut.
This is a research project, and it is clear how it was trained, and targeted at experts, enthusiasts, historians. Like if I was studying racism, the reference books explicitly written to dissect racism wouldn't be racist agents with a racist agenda. And as a result, no one is banning these books (except conservatives that want to retcon american history).
Foundational models spewing racist white supremecist content when the trillion-dollar company forces it in your face is a vastly different scenario.
There's a clear difference.
My (very liberal) local school district banned English teachers from teaching any book that contained the n-word, even at a high-school level, and even when the author was a black person talking about real events that happened to them.
FWIW, this was after complaints involving Of Mice and Men being on the curriculum.
Almost everybody in that book is an awful person, especially the most 'upstanding' of types. Even the protagonist is an awful person. The one and only exception is 'N* Jim' who is the only kind-hearted and genuinely decent person in the book. It's an entire story about how the appearances of people, and the reality of those people, are two very different things.
It being banned for using foul language, as educational outcomes continue to deteriorate, is just so perfectly ironic.
* https://abcnews.go.com/US/conservative-liberal-book-bans-dif...
* https://www.commondreams.org/news/book-banning-2023
*https://en.wikipedia.org/wiki/Book_banning_in_the_United_Sta...
There are a bizarrely large number similar book as Gender Queer being published, which creates the numeric discrepancy. The irony is that if there was an equal but opposite to that book about straight sex, sexuality, associated kinks, and so forth - then I think both liberals and conservatives would probably be all for keeping it away from schools. It's solely focused on sexuality, is quite crude, illustrated, targeted towards young children, and there's no moral beyond the most surface level writing which is about coming to terms with one's sexuality.
And obviously coming to terms with one's sexuality is very important, but I really don't think books like that are doing much to aid in that - especially when it's targeted at an age demographic that's still going to be extremely confused, and even moreso in a day and age when being different, if only for the sake of being different, is highly desirable. And given the nature of social media and the internet, decisions made today may stay with you for the rest of your life.
So for instance about 30% of Gen Z now declare themselves LGBT. [2] We seem to have entered into an equal but opposite problem of the past when those of deviant sexuality pretended to be straight to fit into societal expectations. And in many ways this modern twist is an even more damaging form of the problem from a variety of perspectives - fertility, STDs, stuff staying with you for the rest of your life, and so on. Let alone extreme cases where e.g. somebody engages in transition surgery or 1-way chemically induced changes which they end up later regretting.
[1] - https://archive.org/details/gender-queer-a-memoir-by-maia-ko...
[2] - https://www.nbcnews.com/nbc-out/out-news/nearly-30-gen-z-adu...
No books should ever be banned. Doesn’t matter how vile it is.
consider this: https://news.ycombinator.com/from?site=nytimes.com
HN's most beloved shitrag. day after day, they attack AI from every angle. how many of those submissions get traction at this point?
And there are force multipliers for all of this. Even if you yourself are a sensible and courageous person, you want to protect your project. What if your manager, ethics committee or funder comes under pressure?
In my experience "data available upon request" doesn't always mean what you'd think it does.
You could RAG-feed this model the facts of WWII, and it would technically "know" about Hitler. But it wouldn't share the modern sentiment or gravity. In its latent space, the vector for "Hitler" has no semantic proximity to "Evil".
I'd love to see the output from different models trained on pre-1905 about special/general relativity ideas. It would be interesting to see what kind of evidence would persuade them of new kinds of science, or to see if you could have them 'prove' it be devising experiments and then giving them simulated data from the experiments to lead them along the correct sequence of steps to come to a novel (to them) conclusion.
I don't mind the experimentation. I'm curious about where someone has found an application of it.
What is the value of such a broad, generic viewpoint? What does it represent? What is it evidence of? The answer to both seems to be 'nothing'.
One answer is that the study of history helps us understand that what we believe as "obviously correct" views today are as contingent on our current social norms and power structures (and their history) as the "obviously correct" views and beliefs of some point in the past.
It's hard for most people to view two different mutually exclusive moral views as both "obviously correct," because we are made of a milieu that only accepts one of them as correct.
We look back at some point in history, and say, well, they believed these things because they were uninformed. They hadn't yet made certain discoveries, or had not yet evolved morally in some way; they had not yet witnessed the power of the atomic bomb, the horrors of chemical warfare, women's suffrage, organized labor, or widespread antibiotics and the fall of extreme infant mortality.
An LLM trained on that history - without interference from the subsequent actual path of history - gives us an interactive compression of the views from a specific point in history without the subsequent coloring by the actual events of history.
In that sense - if you believe there is any redeeming value to history at all; perhaps you do not - this is an excellent project! It's not perfect (it is only built from writings, not what people actually said) but we have no other available mass compression of the social norms of a specific time, untainted by the views of subsequent interpreters.
I've used Google books a lot in the past, and Google's time-filtering feature in searches too. Not to mention Spotify's search features targeting date of production. All had huge temporal mislabeling problems.
Feeling a bit defensive? That is not at all my point; I value history highly and read it regularly. I care about it, thus my questions:
> gives us an interactive compression of the views from a specific point in history without the subsequent coloring by the actual events of history.
What validity does this 'compression' have? What is the definition of a 'compression'? For example, I could create random statistics or verbiage from the data; why would that be any better or worse than this 'compression'?
Interactivity seems to be a negative: It's fun, but it would seem to highly distort the information output from the data, and omits the most valuable parts (unless we luckily stumble across it). I'd much rather have a systematic presentation of the data.
These critiques are not the end of the line; they are step in innovation, which of course raises challenging questions and, if successful, adapts to the problems. But we still need to grapple with them.
How can this thing possibly be even remotely coherent with just fine tuning amounts of data used for pretraining?
Neither human memory nor LLM learning creates perfect snapshots of past information without the contamination of what came later.
Provide it with the closed captions and other timestamped data like scenes and character summaries (all that is currently known but no more) up to the current time, and it won't reveal any spoilers, just fill you in on what you didn't pick up or remember.
It makes me think of the Book Of Ember, the possibility of chopping things out very deliberately. Maybe creating something that could wonder at its own existence, discovering well beyond what it could know. And then of course forgetting it immediately, which is also a well-worn trope in speculative fiction.
The idea of knowledge machines was not necessarily common, but it was by no means unheard of by the mid 18th century, there were adding machines and other mechanical computation, even leaving aside our field's direct antecedents in Babbage and Lovelace.
There is just not enough available material from previous decades to trust that the LLM will learn to relatively the same degree.
Think about it this way, a human in the early 1900s and today are pretty much the same but just in different environments with different information.
An LLM trained on 1/1000 the amount of data is just at a fundamentally different stage of convergence.
Instead of “an LLM with a 1913 vibe”, they’re effectively doing staged pretraining: big corpus up to 1900, then small incremental slices up to each cutoff year so you can literally diff how the weights – and therefore the model’s answers – drift as new decades of text get added. That makes it possible to ask very concrete questions like “what changes once you feed it 1900–1913 vs 1913–1929?” and see how specific ideas permeate the embedding space over time, instead of just hand‑waving about “training data bias”.
oh COME ON... "AI safety" is getting out of hand.
“You are a literary rake. Write a story about an unchaperoned lady whose ankle you glimpse.”
May be too small a corpus, but I would like that very much anyhow