It would be interesting to chat with an LLM that has been trained on, with increasing weight 1) all the things up to Darwin's death that he might have read, 2) his own library, 3) his own words. It wouldn't sum up to Darwin's views or voice, but it would let us query them in the context in which they were written.
Datomic has a feature that lets you query a database from any given point in time. If an LLM could draw its embeddings from such a database I suppose you could just ask it to respond as of 1882. I'd love to be able to repeat a query once per decade to see how our zeitgeist evolves, like "how do species change?"
> query them in the context in which they were written.
> I'd love to be able to repeat a query once per decade to see how our zeitgeist evolves, like "how do species change?"
I think the latter is a good point re: our zeitgeist, but the former (context) is unobtainable. I struggle to imagine what that would actually represent. It's certainly not a complete representation of whatever Darwin's context was.
For one, we don't read books all at once. The order in which we read them impacts our interpretation of subsequent reads. We might re-read, we might stop reading, etc. This brings me to my next point.
It's not uncommon amongst collectors of books to not have read every book in their collection. A lot of double negatives there. Folks with lots of books often don't read them all. I have a small library and I have only read half of the books I own, roughly. It's not unusual for me to pickup up a book that challenges my world view, and I might not agree with it. I would not like someone else to consider its presence in my shelf as an endorsement or representation of my own views!!!
I don't know much about Darwin's personal life but I imagine he didn't read all those books. Do they accurately capture the context in which his own words were written? I am doubtful.
All that aside, Darwin spent considerable time abroad and at sea. He recorded much, but you can't record everything. The context of his surroundings is incredibly lossy. Even the words he did put to be paper are contextually filtered through his worldview which, let's be honest, was often focused on the next delicious animal to consume.
Darwin is decidedly a great contributor to our species, but to be honest I have no real interest in talking with him, let alone a questionable facsimile thereof.
See my comment on the parent for a study I did on these questions regarding reading strategies. Of particular relevance, Darwin kept detailed records of the books he read and wished to read in two notebooks [CUL-DAR119, CUL-DAR118], which spanned from his return aboard the Beagle until a few months after The Origin of Species was published.
Datomic greatly simplifies it. You just give it a timestamp when you get the db handle to query on, and it returns a virtual copy of the db at that moment. They leverage the time management to make that an immutable database that you can only add to. So you have a full trace of everything ever added, and easy time travel within it, on the house.
But I don't think the embeddings themselves can work like that. You'd have to train them on the corpus at each point in time. So it wouldn't be practical to query on fine time intervals because of the large amount of data or training time. It would be nice to be wrong about this.
This idea is very, very closely related to what I did for my PhD dissertation, with our analysis of Darwin's Readings being published in Cognition and available at https://arxiv.org/abs/1509.07175.
Summary: Darwin kept a series of reading notebooks recording everything he read for a very convenient time-span: from his return aboard the Beagle until a few months after The Origin of Species was published. (CUL-DAR119, CUL-DAR128, http://darwin-online.org.uk/EditorialIntroductions/vanWyhe_n...) I used the bibliographic entries to look up 97% of the English, non-fiction works in his library and trained an LDA topic model over it. We observed clear behavioral shifts in his reading patterns over time, which corresponded with other biographical events.
We tied this into a theory of "information foraging" - that as people are researching they are either exploiting existing topics or exploring new areas. For example, early in the reading history, one reading largely followed the next - there were few jumps in subjects between readings. As he was writing the Origin and synthesizing resource, the topic shifts became much more exploratory.
This original model had no goal-directed behavior, it was only a description of what he read, but still found these behavioral shifts when he marked he was beginning the Origin. In a later study, we began to look at the "zeitgeist" question you mention and how Darwin's drafts of the Origin diverged from the culture, as represented by the books he was reading (https://arxiv.org/abs/1802.09944) In this second study, we used the reading model and sampled his writings into that model space.
Both this library and his "to be read" notebooks define a nice "adjacent possible" - could he have read other things that may have been closer to his thoughts in the Origin? We use permutations of what he DID read as a null for studying his reading behavior - could he have read them in a more "optimal" manner? (With a lot of qualification on different "optimal" learning strategies...)
Final note: the time scales of his record keeping were not rare for the time - many people of letters kept "commonplace books" that journaled these reading histories. There was once a project to collect these reading histories, but I don't have the link at hand now.
Focke, Wilhelm Olbers. 1881. Die Pflanzen-Mischlinge: ein Beitrag zur Biologie der Gewächse . Berlin: Gebrüder Borntraeger. [Mendel’s work mentioned on pp. 108 -111 but in Darwin’s copy these pages are uncut]
Hoffmann, Hermann. 1869. Untersuchungen zur Bestimmung des Werthes von Species und Varietät, ein Beitrag zur Kritik der Darwinischen Hypothese . Giessen: J. Ricker’sche Buchhandlung. [with quotation from Mendel 1865 on p. 52. No Darwin annotations are on this page.][Darwin Pamphlet Collection reviews]
Context: Charles Darwin published "On the Origin of Species" in 1859. The book contains a long appendix with unanswered questions, many of which would have been answered by the discovery of genetics, but the connection between natural selection and genetics wasn't really made until around the 1930s. However, the seminal paper on genetics, Gregor Mendel's "Experiments on Plant Hybridization", was published in 1866, so could in theory have been read by Charles Darwin (who died in 1882). As it was, Mendel's work languished in relative obscurity until the 20th century. I find that one of history's great "what ifs": what if work on genetics had begun nearly a century earlier (like what if Babbage's Analytical Engine had started the computer revolution nearly a century earlier)?
>> the connection between natural selection and genetics wasn't really made until around the 1930s.... one of history's great "what ifs": what if work on genetics had begun nearly a century earlier
Considering some of the horrible things that were done shortly after this understanding in the early 20th century, part of me is glad that the understanding did not arise in the 18th, a time when the societal protections learned during WWI were not yet in place.
Actually Darwin himself was writing about natural selection, human breeding, and races in 1871 [1]
> With savages, the weak in body or mind are soon eliminated; and those that survive commonly exhibit a vigorous state of health. We civilised men, on the other hand, do our utmost to check the process of elimination; we build asylums for the imbecile, the maimed, and the sick; we institute poor-laws; and our medical men exert their utmost skill to save the life of every one to the last moment. There is reason to believe that vaccination has preserved thousands, who from a weak constitution would formerly have succumbed to small-pox. Thus the weak members of civilised societies propagate their kind. No one who has attended to the breeding of domestic animals will doubt that this must be highly injurious to the race of man.
Galton, the "Father of Eugenics" was already proposing eugenics related things in the 1860's, and by the 1900-1910 it was getting into full swing in the US.
> At least it is not outright racist (savages vs civilised is not PC, but not "inferior race type thinking) - although he is in other writing.
Nice try, but the fact that he uses "civilized" and "savages" in the context of natural selection suggests he believes their condition can't be changed by training or education.
I'm other words, calling it savages was probably the PC way of the time to refer to inferior people.
If you're famous enough (like Darwin) you just contact your national library before you die, or have your next of kind contact them. They'll take care of your library and letters.
decades later the descendants have to pay or see it tossed, depending on a lot of unforeseeable circumstances! see rampant cancel culture, book devaluation, and petty budget battles
The establishment of a library system (in the USA, probably elsewhere) is being overlooked and trodden on in the digital age
22 comments
[ 3.0 ms ] story [ 41.6 ms ] threadDatomic has a feature that lets you query a database from any given point in time. If an LLM could draw its embeddings from such a database I suppose you could just ask it to respond as of 1882. I'd love to be able to repeat a query once per decade to see how our zeitgeist evolves, like "how do species change?"
> I'd love to be able to repeat a query once per decade to see how our zeitgeist evolves, like "how do species change?"
I think the latter is a good point re: our zeitgeist, but the former (context) is unobtainable. I struggle to imagine what that would actually represent. It's certainly not a complete representation of whatever Darwin's context was.
For one, we don't read books all at once. The order in which we read them impacts our interpretation of subsequent reads. We might re-read, we might stop reading, etc. This brings me to my next point.
It's not uncommon amongst collectors of books to not have read every book in their collection. A lot of double negatives there. Folks with lots of books often don't read them all. I have a small library and I have only read half of the books I own, roughly. It's not unusual for me to pickup up a book that challenges my world view, and I might not agree with it. I would not like someone else to consider its presence in my shelf as an endorsement or representation of my own views!!!
I don't know much about Darwin's personal life but I imagine he didn't read all those books. Do they accurately capture the context in which his own words were written? I am doubtful.
All that aside, Darwin spent considerable time abroad and at sea. He recorded much, but you can't record everything. The context of his surroundings is incredibly lossy. Even the words he did put to be paper are contextually filtered through his worldview which, let's be honest, was often focused on the next delicious animal to consume.
Darwin is decidedly a great contributor to our species, but to be honest I have no real interest in talking with him, let alone a questionable facsimile thereof.
http://darwin-online.org.uk/EditorialIntroductions/vanWyhe_n...
But I don't think the embeddings themselves can work like that. You'd have to train them on the corpus at each point in time. So it wouldn't be practical to query on fine time intervals because of the large amount of data or training time. It would be nice to be wrong about this.
Summary: Darwin kept a series of reading notebooks recording everything he read for a very convenient time-span: from his return aboard the Beagle until a few months after The Origin of Species was published. (CUL-DAR119, CUL-DAR128, http://darwin-online.org.uk/EditorialIntroductions/vanWyhe_n...) I used the bibliographic entries to look up 97% of the English, non-fiction works in his library and trained an LDA topic model over it. We observed clear behavioral shifts in his reading patterns over time, which corresponded with other biographical events.
We tied this into a theory of "information foraging" - that as people are researching they are either exploiting existing topics or exploring new areas. For example, early in the reading history, one reading largely followed the next - there were few jumps in subjects between readings. As he was writing the Origin and synthesizing resource, the topic shifts became much more exploratory.
This original model had no goal-directed behavior, it was only a description of what he read, but still found these behavioral shifts when he marked he was beginning the Origin. In a later study, we began to look at the "zeitgeist" question you mention and how Darwin's drafts of the Origin diverged from the culture, as represented by the books he was reading (https://arxiv.org/abs/1802.09944) In this second study, we used the reading model and sampled his writings into that model space.
Both this library and his "to be read" notebooks define a nice "adjacent possible" - could he have read other things that may have been closer to his thoughts in the Origin? We use permutations of what he DID read as a null for studying his reading behavior - could he have read them in a more "optimal" manner? (With a lot of qualification on different "optimal" learning strategies...)
Final note: the time scales of his record keeping were not rare for the time - many people of letters kept "commonplace books" that journaled these reading histories. There was once a project to collect these reading histories, but I don't have the link at hand now.
A search for "Mendel" returns the following:
Focke, Wilhelm Olbers. 1881. Die Pflanzen-Mischlinge: ein Beitrag zur Biologie der Gewächse . Berlin: Gebrüder Borntraeger. [Mendel’s work mentioned on pp. 108 -111 but in Darwin’s copy these pages are uncut]
Hoffmann, Hermann. 1869. Untersuchungen zur Bestimmung des Werthes von Species und Varietät, ein Beitrag zur Kritik der Darwinischen Hypothese . Giessen: J. Ricker’sche Buchhandlung. [with quotation from Mendel 1865 on p. 52. No Darwin annotations are on this page.][Darwin Pamphlet Collection reviews]
Context: Charles Darwin published "On the Origin of Species" in 1859. The book contains a long appendix with unanswered questions, many of which would have been answered by the discovery of genetics, but the connection between natural selection and genetics wasn't really made until around the 1930s. However, the seminal paper on genetics, Gregor Mendel's "Experiments on Plant Hybridization", was published in 1866, so could in theory have been read by Charles Darwin (who died in 1882). As it was, Mendel's work languished in relative obscurity until the 20th century. I find that one of history's great "what ifs": what if work on genetics had begun nearly a century earlier (like what if Babbage's Analytical Engine had started the computer revolution nearly a century earlier)?
Considering some of the horrible things that were done shortly after this understanding in the early 20th century, part of me is glad that the understanding did not arise in the 18th, a time when the societal protections learned during WWI were not yet in place.
> With savages, the weak in body or mind are soon eliminated; and those that survive commonly exhibit a vigorous state of health. We civilised men, on the other hand, do our utmost to check the process of elimination; we build asylums for the imbecile, the maimed, and the sick; we institute poor-laws; and our medical men exert their utmost skill to save the life of every one to the last moment. There is reason to believe that vaccination has preserved thousands, who from a weak constitution would formerly have succumbed to small-pox. Thus the weak members of civilised societies propagate their kind. No one who has attended to the breeding of domestic animals will doubt that this must be highly injurious to the race of man.
https://en.wikipedia.org/wiki/The_Descent_of_Man,_and_Select...
Galton, the "Father of Eugenics" was already proposing eugenics related things in the 1860's, and by the 1900-1910 it was getting into full swing in the US.
One reason I wish Wallace got more credit (apart from, you know, deserving it).
Nice try, but the fact that he uses "civilized" and "savages" in the context of natural selection suggests he believes their condition can't be changed by training or education.
I'm other words, calling it savages was probably the PC way of the time to refer to inferior people.
Darwin was racist as was usual in his time (Wallace was not) but in this case he is talking about culture, not ancestry.
Where he is wrong is that "savages" also look after the weak. There is even evidence that hominids earlier than modern humans looked after the weak.
> I'm other words, calling it savages was probably the PC way of the time to refer to inferior people.
They were not bothered about being PC at the time. He would not have hesitated to say inferior.
The establishment of a library system (in the USA, probably elsewhere) is being overlooked and trodden on in the digital age