Posted by iamwil 4 days ago
“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”.
Without saying anything specific to spoil plot poonts, I will say that I ended-up having a kidney stone while I was reading the last two books of the series. It was fucking eerie.
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.
It's the most prescient thing I've ever read, and it's pretty short and a genuinely good story, I recommend everyone read it.
Edit: Just skimmed it again and realized there's an LLM-like prediction as well. Access to the Earth's surface is banned and some people complain, until "even the lecturers acquiesced when they found that a lecture on the sea was none the less stimulating when compiled out of other lectures that had already been delivered on the same subject."
-people a depicted as grey aliens (no teeth, large eyes, no hair). Lesson the Greys are a future version of us.
The air is poisoned and ruined cities. People live in underground bunkers...1909...nuclear war was unimaginable then. This was still the age of steam ships and coal power trains. Even respirators would have been low on the public imagination.
The air ships with metal blinds sound more like UFOs than blimps.
The white worms.
People are the blood cells of the machine which runs on their thoughts social media data harvesting of ai.
China invaded Australia. This story was 8 years or so after the Boxer Rebellion so that would have sounded like say Iraq invading the USA in the context of its time.
The story suggests this is a cyclical process of a bifurcated human race.
The blimp crashing into the steel evokes 9/11, 91+1 years later...
The constellation orion.
Etc etc.
There is a central commitee
It’s interesting - Forster wrote like the Huxley of his day, Zamyatin like the Orwell - but both felt they were carrying Wells’ baton - and they were, just from differing perspectives.
That's just the Victorian London.
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?
The Shockwave Rider was also remarkable prescient.
They're convenient but if they went away tomorrow, my life wouldn't really change in any material way. That's not really the case with smartphones much less the internet more broadly.
Funny, I had "The collected stories of Frank Herbert" as my next read on my tablet. Here's a juicy quote from like the third screen of the first story:
"The bedside newstape offered a long selection of stories [...]. He punched code letters for eight items, flipped the machine to audio and listened to the news while dressing."
Anything qualitative there? Or all of it quantitative?
Story is "Operation Syndrome", first published in 1954.
Hey, where are our glowglobes and chairdogs btw?
Went further in Herbert's shorts volume and I just ran into a scene where people are preparing to leave Earth on a colony ship to seed some distant world...
... and they still have human operator assisted phone calls.
I'd take smartphones vanishing rather than books any day.
I didn't believe you meant that of course, but we've already seen it can happen.
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.
I've been tempted to. "Everything will be terrible if these guys succeed, but at least I'll be rich. If they fail I'll lose money, but since that's the outcome I prefer anyway, the loss won't bother me."
Trouble is, that ship has arguably already sailed. No matter how rapidly things go to hell, it will take many years before PLTR is profitable enough to justify its half-trillion dollar market cap.
So "panopticon that if it had been used properly, would have prevented the destruction of two towers" while ignoring the obvious "are we the baddies?"
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.
Right, I think you have responsibility for your 1/<global population>th (arguably considerably more though, for first-worlders) of the problem. What I see is something like refusal to consider swapping out a two-stroke-engine-powered tungsten lightbulb with an LED of equivalent brightness, CRI, and color temperature, because it won't unilaterally solve the problem.
I proudly owned zero shares of Microsoft stock, in the 1980s and 1990s. :)
I own no Palantir today.
It's a Pyrrhic victory, but sometimes that's all you can do.
I assume the CIA is lying about simulating world leaders. These are narcissistic personalities and it’s jarring to hear that they can be replaced, either by a body double or an indistinguishable chatbot. Also, it’s still cheaper to have humans do this.
More likely, the CIA is modeling its own experts. Not as useful a press release and not as impressive to the fractious executive branch. But consider having downtime as a CIA expert on submarine cables. You might be predicting what kind of available data is capable of predicting the cause and/or effect of cuts. Ten years ago, an ensemble of such models was state of the art, but its sensory libraries were based on maybe traceroute and marine shipping. With an LLM, you can generate a whole lot of training data that an expert can refine during his/her downtime. Maybe there’s a potent new data source that an expensive operation could unlock. That ensemble of ML models from ten years ago can still be refined.
And then there’s modeling things that don’t exist. Maybe it’s important to optimize a statement for its disinfo potency. Try it harmlessly on LLMs fed event data. What happens if some oligarch retires unexpectedly? Who rises? That kind of stuff.
To your last point, with this executive branch, I expect their very first question to CIA wasn’t about aliens or which nations have a copy of a particular tape of Trump, but can you make us money. So the approaches above all have some way of producing business intelligence. Whereas a Kim Jong Un bobblehead does not.
https://www.amazon.com/Man-Presidents-Mind-Ted-Allbeury/dp/0...
[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
Imagine you are a billionaire so money is no object and really interested in the Dhali Llama?
Would you read the book then hire someone to pretend to be the author and ask questions that are not covered by the book? Then be enraptured by whatever the roleplayer invents?
Probably not? At least this isn't a phenomenon I've heard of?
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!
With Alphonse X, o The Cid, it would be greater issues, but understandable over weeks.
Having the facts from the era is one thing, to make conclusions about things it doesn't know would require intelligence.
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.
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.
I wasn’t arguing that LLMs are like a human brain. Of course they aren’t. I said twice in my original post that they aren’t like humans. But “like a human brain” and “autocomplete on steroids” aren’t the only two choices here.
As for appealing to complexity, well, let’s call it more like an appeal to humility in the face of complexity. My basic claim is this:
1) It is a trap to reason from model architecture alone to make claims about what LLMs can and can’t do.
2) The specific version of this in GP that I was objecting to was: LLMs are just transformers that do next token prediction, therefore they cannot solve novel problems and just regurgitate their training data. This is provably true or false, if we agree on a reasonable definition of novel problems.
The reason I believe this is that back in 2023 I (like many of us) used LLM architecture to argue that LLMs had all sorts of limitations around the kind of code they could write, the tasks they could do, the math problems they could solve. At the end of 2025, SotA LLMs have refuted most of these claims by being able to do the tasks I thought they’d never be able to do. That was a big surprise to a lot us in the industry. It still surprises me every day. The facts changed, and I changed my opinion.
So I would ask you: what kind of task do you think LLMs aren’t capable of doing, reasoning from their architecture?
I was also going to mention RL, as I think that is the key differentiator that makes the “knowledge” in the SotA LLMs right now qualitatively different from GPT2. But other posters already made that point.
This topic arouses strong reactions. I already had one poster (since apparently downvoted into oblivion) accuse me of “magical thinking” and “LLM-induced-psychosis”! And I thought I was just making the rather uncontroversial point that things may be more complicated than we all thought in 2023. For what it’s worth, I do believe LLMs probably have limitations (like they’re not going to lead to AGI and are never going to do mathematics like Terence Tao) and I also think we’re in a huge bubble and a lot of people are going to lose their shirts. But I think we all owe it to ourselves to take LLMs seriously as well. Saying “Opus 4.5 is the same thing as GPT2” isn’t really a pathway to do that, it’s just a convenient way to avoid grappling with the hard questions.
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.
Roots of predictive coding theory extend back to 1860s.
Natalia Bekhtereva was writing about compact concept representations in the brain akin to tokens.
Yes, you can draw interesting parallels between anything when you're motivated to do so. My point is that this isn't parsimonious reasoning, it's working backwards from a conclusion and searching for every opportunity to fit the available evidence into a narrative that supports it.
> Roots of predictive coding theory extend back to 1860s.
This is just another example of metaphorical parallels overstating meaningful connections. Just because next-token-prediction and predictive coding have the word "predict" in common doesn't mean the two are at all related in any practical sense.
Fascinating framing. What would you consider evidence here?
Think about an average dinner party conversation. Person A talks, person B thinks about something to say that fits, person C gets an association from what A and B said and speaks...
And what are people most interested in talking about? Things they read or watched during the week perhaps?
Conversations would not have had to be like this. Imagine a species from another planet who had a "conversation" where each party simply communicated what it most needed to say/was most benefitial to say and said it. And where the chance of bringing up a topic had no correlation at all with what previous person said (why should it?) or with what was in the newspapers that week. And who had no "interest" in the association game.
Humans saying they are not driven by associations is to me a bit like fish saying they are not noticing the water. At least MY thought processes works like that.
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.
... but we also need to be careful with that assertion, because humans do not understand cognition, psychology, or biology very well.
Biology is the furthest developed, but it turns out to be like physics -- superficially and usefully modelable, but fundamental mysteries remain. We have no idea how complete our models are, but they work pretty well in our standard context.
If computer engineering is downstream from physics, and cognition is downstream from biology ... well, I just don't know how certain we can be about much of anything.
> this thread sure attracts a lot of armchair experts.
"So we beat on, boats against the current, borne back ceaselessly into our priors..."
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.
But regardless, I don’t think anyone is claiming that LLMs can magically do things that aren’t in their training data or context window. Obviously not: they can’t learn on the job and the permanent knowledge they have is frozen in during training.
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.
> What did I have for breakfast this morning?
> I don’t know what you had for breakfast this morning…
Most modern post training setups encourage this.
It isn't 2023 anymore.
Well, no, they are training set statistical predictors, not individual training sample predictors (autocomplete).
The best mental model of what they are doing might be that you are talking to a football stadium full of people, where everyone in the stadium gets to vote on the next word of the response being generated. You are not getting an "autocomplete" answer from any one coherent source, but instead a strange composite response where each word is the result of different people trying to steer the response in different directions.
An LLM will naturally generate responses that were not in the training set, even if ultimately limited by what was in the training set. The best way to think of this is perhaps that they are limited to the "generative closure" (cf mathematical set closure) of the training data - they can generate "novel" (to the training set) combinations of words and partial samples in the training data, by combining statistical patterns from different sources that never occurred together in the training data.
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.
Source needed RE brain.
Define innovate, in a way that a LLM can't and we definitively can prove a human can.
Transformers allow for the mapping of a complex manifold representation of causal phenomena present in the data they're trained on. When they're trained on a vast corpus of human generated text, they model a lot of the underlying phenomena that resulted in that text.
In some cases, shortcuts and hacks and entirely inhuman features and functions are learned. In other cases, the functions and features are learned to an astonishingly superhuman level. There's a depth of recursion and complexity to some things that escape the capability of modern architectures to model, and there are subtle things that don't get picked up on. LLMs do not have a coherent self, or subjective central perspective, even within constraints of context modifications for run-time constructs. They're fundamentally many-minded, or no-minded, depending on the way they're used, and without that subjective anchor, they lack the principle by which to effectively model a self over many of the long horizon and complex features that human brains basically live in.
Confabulation isn't unique to LLMs. Everything you're saying about how LLMs operate can be said about human brains, too. Our intelligence and capabilities don't emerge from nothing, and human cognition isn't magical. And what humans do can also be considered "intelligent autocomplete" at a functional level.
What cortical columns do is next-activation predictions at an optimally sparse, embarrassingly parallel scale - it's not tokens being predicted but "what does the brain think is the next neuron/column that will fire", and where it's successful, synapses are reinforced, and where it fails, signals are suppressed.
Neocortical processing does the task of learning, modeling, and predicting across a wide multimodal, arbitrary depth, long horizon domain that allow us to learn words and writing and language and coding and rationalism and everything it is that we do. We're profoundly more data efficient learners, and massively parallel, amazingly sparse processing allows us to pick up on subtle nuance and amazing wide and deep contextual cues in ways that LLMs are structurally incapable of, for now.
You use the word hallucinations as a pejorative, but everything you do, your every memory, experience, thought, plan, all of your existence is a hallucination. You are, at a deep and fundamental level, a construct built by your brain, from the processing of millions of electrochemical signals, bundled together, parsed, compressed, interpreted, and finally joined together in the wonderfully diverse and rich and deep fabric of your subjective experience.
LLMs don't have that, or at best, only have disparate flashes of incoherent subjective experience, because nothing is persisted or temporally coherent at the levels that matter. That could very well be a very important mechanism and crucial to overcoming many of the flaws in current models.
That said, you don't want to get rid of hallucinations. You want the hallucinations to be valid. You want them to correspond to reality as closely as possible, coupled tightly to correctly modeled features of things that are real.
LLMs have created, at superhuman speeds, vast troves of things that humans have not. They've even done things that most humans could not. I don't think they've done things that any human could not, yet, but the jagged frontier of capabilities is pushing many domains very close to the degree of competence at which they'll be superhuman in quality, outperforming any possible human for certain tasks.
There are architecture issues that don't look like they can be resolved with scaling alone. That doesn't mean shortcuts, hacks, and useful capabilities won't produce good results in the meantime, and if they can get us to the point of useful, replicable, and automated AI research and recursive self improvement, then we don't necessarily need to change course. LLMs will eventually be used to find the next big breakthrough architecture, and we can enjoy these wonderful, downright magical tools in the meantime.
And of course, human experts in the loop are a must, and everything must be held to a high standard of evidence and review. The more important the problem being worked on, like a law case, the more scrutiny and human intervention will be required. Judges, lawyers, and politicians are all using AI for things that they probably shouldn't, but that's a human failure mode. It doesn't imply that the tools aren't useful, nor that they can't be used skillfully.
BINGO!
(I just won a stuffed animal prize with my AI Skeptic Thought-Terminating Cliché BINGO Card!)
Sorry. Carry on.
"<Thing> doesn't <action>, it <shallow description that's slightly off from how you would expect a human to choose>"
Later parts of the readme (whole section of bullets enumerating what it is and what it isn't, another LLM favorite) make me more confident that significant parts of the readme is generated.
I'm generally pro-AI, but if you spend hundreds of hours making a thing, I'd rather hear your explanation of it, not an LLM's.
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)
I’m not a Doctor Who fan and haven’t seen the rest of the episode and I don’t even what this episode was about but I thought this scene was excellent.
Applicable to us also, cause we do not know how the current story ends either, of the post pandemic world as we know it now.
Hell 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.
So as a black person should I demand that all books written before the civil rights act be destroyed?
The past is messy. But it's the only way to learn anything.
All an LLM does it's take a bunch of existing texts and rebundle them. Like it or not, the existing texts are still there.
I understand an LLM that won't tell me how to do heart surgery. But I can't fear one that might be less enlightened on race issues. So many questions to ask! Hell, it's like talking to older person in real life.
I don't expect a typical 90 year old to be the most progressive person, but they're still worth listening too.
Self preservation is the first law of nature. If you release the model someone will basically say you endorse those views and you risk your funding being cut.
You created Pandora's box and now you're afraid of opening it.
That should be more than enough to clear any chance of misunderstanding.
I could easily see a hit piece making its rounds on left leaning media about the AI that re-animates the problematic ideas of the past. "Just look at what it said to my child, "<insert incredibly racist quote coerced out of the LLM here>"!" Rolling stones would probably have a front page piece on it, titled "AI resurrecting racism and misogyny". There would easily be enough there to attract death threats to the developers, if it made its rounds on twitter.
"Platforming ideas" would be the issue that people would have.
I suspect restricting access could equally be a comment on modern LLMs in general, rather than the historical material specifically. For example, we must be constantly reminded not to give LLMs a level of credibility that their hallucinations would have us believe.
But I'm fascinated by the possibility that somehow resurrecting lost voices might give an unholy agency to minds and their supporting worldviews that are so anachronistic that hearing them speak again might stir long-banished evils. I'm being lyrical for dramatic affect!
I would make one serious point though, that do I have the credentials to express. The conversation may have died down, but there is still a huge question mark over, if not the legality, but certainly the ethics of restricting access to, and profiting from, public domain knowledge. I don't wish to suggest a side to take here, just to point out that the lack of conversation should not be taken to mean that the matter is settled.
Their concern can't be understood without a deep understanding of the far left wing mind. Leftists believe people are so infinitely malleable that merely being exposed to a few words of conservative thought could instantly "convert" someone into a mortal enemy of their ideology for life. It's therefore of paramount importance to ensure nobody is ever exposed to such words unless they are known to be extremely far left already, after intensive mental preparation, and ideally not at all.
That's why leftist spaces like universities insist on trigger warnings on Shakespeare's plays, why they're deadly places for conservatives to give speeches, why the sample answers from the LLM are hidden behind a dropdown and marked as sensitive, and why they waste lots of money training an LLM that they're terrified of letting anyone actually use. They intuit that it's a dangerous mind bomb because if anyone could hear old fashioned/conservative thought, it would change political outcomes in the real world today.
Anyone who is that terrified of historical documents really shouldn't be working in history at all, but it's academia so what do you expect? They shouldn't be allowed to waste money like this.
The problem with it is, it already happened at least once. We know how it happened. Unchecked narratives about minorities or foreigners is a significant part of why the 20th century happened to Europe, and it’s a significant part of why colonialism and slavery happened to other places.
What solution do you propose?
By studying history better, people wouldn't draw the wrong conclusions about what caused it. Watch out for left wing radicals promoting socialism-with-genetic-characteristics.
Both ideologically and historically the two ideologies are complete opposites. There is no socialist “root” to nazi ideology - at all.
Now were it limited in access to ask money to compensate for the time and money spent compiling the library (or training the model), sure, I'd somewhat understand. Not agree but understand.
Now it just feels like you want to prevent your model name being associated with the one guy who might use it to create a racist slur Twitter bot. There's plenty of models for that already. At least the societal balance of a model like this would also have enough weight on the positive side to be net positive.
We all get that academics now exist in some kind of dystopian horror where they can get transitively blamed for the existence of anyone to the right of Lenin, but bear in mind:
1. The people who might try to cancel you are idiots unworthy of your respect, because if they're against this project, they're against the study of history in its entirety.
2. They will scream at you anyway no matter what you do.
3. You used (Swiss) taxpayer funds to develop these models. There is no moral justification for withholding from the public what they worked to pay for.
You already slathered your README with disclaimers even though you didn't even release the model at all, just showed a few examples of what it said - none of which are in any way surprising. That is far more than enough. Just release the models and if anyone complains, politely tell them to go complain to the users.
Movie studios have done that for years with old movies. TCM still shows Birth of a Nation and Gone with the Wind.
Edit: I saw further down that you've already done this! What more is there to do?
I guess what they're really saying is "we don't want you guys to cancel us".
What do these people fear the most? That the "truth" they been pushing is a lie.
Also of course this is for one training run, if you need to experiment you'd need to do that more.
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 ...
When you're looking at e.g. the 19th century, a huge number are preserved somewhere in some library, but the vast majority don't seem to be digitized yet, given the tremendous amount of work.
Given how much higher-quality newspaper content tends to be compared to the average internet forum thread, there actually might be quite a decent amount of text. Obviously still nothing compared to the internet, but still vastly larger than just from published books. After all, print newspapers were essentially the internet of their day. Oh, and don't forget pamphlets in the 18th century.
Hm there is a lot of text from before the internet, but most of it is not on internet. There is a weird gap in some circles because of that, people are rediscovering work from pre 1980s researchers that only exist in books that have never been re-edited and that virtually no one knows about.
The National Archives of Spain alone have 350 million pages of documents going back to the 15th century, ranging from correspondence to testimony to charts and maps, but only 10% of it is digitized and a much smaller fraction is transcribed. Hopefully with how good LLMs are getting they can accelerate the transcription process and open up all of our historical documents as a huge historical LLM dataset.
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?
Something like a pop-sci article along the lines of "Mad scientists create racist, imperialistic AI"?
I honestly don't see publication of the weights as a relevant risk factor, because sensationalist misrepresentation is trivially possible with the given example responses alone.
I don't think such pseudo-malicious misrepresentation of scientific research can be reliably prevented anyway, and the disclaimers make your stance very clear.
On the other hand, publishing weights might lead to interesting insights from others tinkering with the models. A good example for this would be the published word prevalence data (M. Brysbaert et al @Ghent University) that led to interesting follow-ups like this: https://observablehq.com/@yurivish/words
I hope you can get the models out in some form, would be a waste not to, but congratulations on a fascinating project regardless!
Your pre-judgement of acceptable hammer uses would rob hammer owners of responsible and justified self-defense and defense of others in situations in which there are no other options, as well as other legally and socially accepted uses which do not fit your pre-conceived ideas.
I think the uncensored response is still valuable, with context. "Those who cannot remember the past are condemned to repeat it" sort of thing.
Edit: just thought of a practical step you can take: host it somewhere else than github. If there's ever going to be a backlash the microsoft moderators might not take too kindly to the stuff about e.g. homosexuality, no matter how academic.
1. This implies a false equivalence. Releasing a new interactive AI model is indeed different in significant and practical ways from the status quo. Yes, there are already-released historical texts. The rational thing to do is weigh the impacts of introducing another thing.
2. Some people have a tendency to say "release everything" as if open-source software is equivalent to open-weights models. They aren't. They are different enough to matter.
3. Rhetorically, the quote across comes across as a pressure tactic. When I hear "are you going to do this or not?" I cringe.
4. The quote above feels presumptive to me, as if the commenter is owed something from the history-llms project.
5. People are rightfully bothered that Big Tech has vacuumed up public domain and even private information and turned it into a profit center. But we're talking about a university project with (let's be charitable) legitimate concerns about misuse.
6. There seems to be a lack of curiosity in play. I'd much rather see people asking e.g. "What factors are influencing your decision about publishing your underlying models?"
7. There are people who have locked-in a view that says AI-safety perspectives are categorically invalid. Accordingly, they have almost a knee-jerk reaction against even talk of "let's think about the implications before we release this."
8. This one might explain and underly most of the other points above. I see signs of a deeper problem at work here. Hiding behind convenient oversimplifications to justify what one wants does not make a sound moral argument; it is motivated reasoning a.k.a. psychological justification.
“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.
Yes, to give them a competitive advantage. Not to LARP as morality police.
There’s a big difference between the two. I take greed over self-righteousness any day.
Or, how about, "If we release this as is, then some people will intentionally mis-use it and create a lot of bad press for us. Then our project will get shut down and we lose our jobs"
Be careful assuming it is a power trip when it might be a fear trip.
I've never been as unimpressed by society as I have been in the last 5 years or so.
> Be careful assuming it is a power trip when
> it might be a fear trip.
>
> I've never been as unimpressed by society as
> I have been in the last 5 years or so.
When I see individuals acting out of fear, I try not to blame them. Fear triggers deep instinctual responses. For example, to a first approximation, a particular individual operating in full-on fight-or-flight mode does not have free will. There is a spectrum here. Here's a claim, which seems mostly true: the more we can slow down impulsive actions, the more hope we have for cultural progress.
When I think of cultural failings, I try to criticize areas where culture could realistically do better. I think of areas where we (collectively) have the tools and potential to do better. Areas where thoughtful actions by some people turn into a virtuous snowball. We can't wait for a single hero, though it helps to create conditions so that we have more effective leaders.
One massive culture failing I see -- that could be dramatically improved -- is this: being lulled into shallow contentment (i.e. via entertainment, power seeking, or material possessions) at the expense of (i) building deep and meaningful social connections and (ii) using our advantages to give back to people all over the world.
Even if I give the comment a lot of wiggle room (such as changing "every" to "many"), I don't think even a watered-down version of this hypothesis passes Occam's razor. There are more plausible explanations, including (1) genuine concern by the authors; (2) academic pressures and constraints; (c) reputational concerns; (d) self-interest to embargo underlying data so they have time to be the first to write-it-up. To my eye, none of these fit the category of "getting high on power".
Also, patience is warranted. We haven't seen what these researchers are doing to release -- and from what I can tell, they haven't said yet. At the moment I see "Repositories (coming soon)" on their GitHub page.
The French released a preview of an AI meant to support public education, but they released the base model, with unsurprising effects [0]
[0] https://www.leparisien.fr/high-tech/inutile-et-stupide-lia-g...
(no English source, unfortunately, but the title translates as: "“Useless and stupid”: French generative AI Lucie, backed by the government, mocked for its numerous bugs")
This constant demonization of everyone who disagrees with you, makes me wonder if 28 Days wasn't more true than we thought, we are all turning into rage zombies.
p-e-w, I'm reacting to much more than your comments. Maybe you aren't totally infected yet, who knows. Maybe you heal.
I am reacting to the pandemic, of which you were demonstrating symptoms.
We can debate on whether it's good or not, but ultimately they're publishing it and in some very small way responsible for some of its ends. At least that's how I can see their interest in disseminating the use of the LLM through a responsible framework.
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.
(I mention this so more people can know the list exists, and hopefully email us more nominations when they see an unusually good and interesting comment.)
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
Dear Hon. Historical LLM
I hope this letter finds you well. It is with no small urgency that I write to you seeking assistance, believing such an erudite and learned fellow as yourself should be the best one to furnish me with an answer to such a vexing question as this which I now pose to you. Pray tell, what is the capital of France?
It’s a better source for how people spoke than books etc, but it’s not really an accurate source for patterns of everyday conversation because people were making speeches rather than chatting.
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.
Here they do 80B tokens for a 4B model.
Under Chinchilla model the larger model always performs better than the small one if trained on the same amount of data. I'm not sure if it is true empirically, and probably 1-10B is a good guess for how large the model trained on 80B tokens should be.
Similarly, the small models continue to improve beyond 20:1 ratio, and current models are trained on much more data. You could train a better performing model using the same compute, but it would be larger which is not always desirable.
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.
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.