Cultural Evolution

#ai

Metadata

Notes

Summary

The talk revolves around the cultural niche, and the ways humans are adapted (and continue to get better at) filling this niche. What's the cultural niche? It's this niche that is a consequence of a faster evolving environment making it useful to have an adaptive species. Culture is perfect for this, as it adds essentially an extra layer of adaptability on top of genetics.
Screenshot 2024-08-03 at 6.34.57 PM.png|300
This chart shows global temperature variation with various stages of evolution plotted.

In particular, two manners in which humans fulfill this niche are:

I think Gopnik's primary hot take here is that LLMs really aren't that different from cultural technological innovations of the past. In fact, she analogizes it with libraries.

Hot Take

LLMs are to the internet as libraries were to the printing press.

The printing press and the internet allowed for the collection and the spread of information. While libraries and now LLMs are how we compress and make this information accessible. That is, both the printing press and the internet were huge leaps forward in the volume of information we could quickly transmit emit. However, it takes two to transmit. Libraries and LLMs are the tools for the receivers.

Slogans

  • The intelligent agent view is wrong. LLMs should be viewed as cultural tools.
  • LLMs are the libraries of the 21st century.
  • Intelligence as a property not of individuals, but of humans equipped with the printing press and the internet and those post-menopausal grandmas, and thousands of years of humans before us.
  • Different intelligences for different stages of life: developmental diversity as a condition for intelligence

Interesting tidbits

Quote

one of the things that happens is you get this new literary form (novels), which is incredibly popular, which everybody reads, especially adolescent girls. Adolescent girls are always at the cutting edge of these changes in cultural technology. And everyone's always sure that it's a really, really, really bad idea that they're adopting the new cultural technology and it's going to lead to the death of civilization, rewiring their brains for anxiety or whatever it is.

My thoughts

I like this take, although I'm still not yet sure if I 100% buy it. There's a lot of people hyping up LLMs both in terms of its promise and its risks. It's nice to get a counterpoint; a "history always repeats itself" perspective.

I think one major difference between LLMs and the cultural technologies that Gopnik compares it with is the motivations of those pushing it. Whereas writing, print, and libraries were all explicitly for the purpose of transmission, people seem to be thinking of LLMs much more as something which is meant to create and perform. The goal of it is much more than distilling information. Sure, we can ask it to do librarian like tasks, and people are using it as a replacement for google search. However, we also use them to generate code and write emails.
But maybe it can still be argued that LLMs are still just a function of the information that humans are outputting; and therefore almost by definition, even if we talk about it as "generating answers," it still falls under the abstraction of: a way to transform human generated data (in this case the internet) into a more useful form.

I'm curious about what this perspective says about where things go from here?

Highlights


the cultural niche really seems to be adapted to the non stationary environment. So both in terms of its causes and its consequences.


And it's an interesting side note that the golem, the rabbi of Prague's story about what happens when you have an artificial system that comes to life is actually pre industrial


not just the technologies that let us go out and do things, but these technologies themselves actually shape the process of cultural transmission.
Note: Yes! This is what I'm interested in!


the more information you can get from more people across more time and more space, and the more efficiently you can access that information, the more you're going to be able to take advantage of the cultural niche.
Note: This is the thesis


I think the really crucial piece of this is the postmenopausal grandmothers. So the way to think of this is, once you have language, then you can listen to what your granny says, you can learn from what your granny says, which means that you've got 100 years worth of cultural knowledge


And parenthetically, I think the big change is not AI. The big change is around 2000, when sort of unheralded, all the information in the world became digital. So you get HGTV, you get PDF's, you get digital movies, and suddenly all of the information is now accessible to you through a digital means, which, of course, means infinitely reproducible and accessible instantaneously in time and space. And I think that's actually the thing that makes a big difference. And the LLMs are just like a logical consequence of what happens once that starts.
Note: Interesting! Says that LLMs are actually not the big step. Do I agree with that?
Or just that, they "make sense." In that cultural technology develops as: (1) new media/form of information, (2) new method of accessing forms of information. (E.g. print => library). LLMs are the library of the internet.

The difference perhaps is that, well, we're eventually gonna run out of room for how much effect media can have on us, individually.

And so you get global phenomena. The real phenomenon can no longer be detected locally. It's not that individuals are getting more adaptable, it's the culture as a whole is getting more adaptable.


intelligence is about relationships between human beings and these cultural tools


a human being with a library and a card catalog is a fundamentally different kind of creature than a human being who doesn't have those kinds of cultural tools.


truth is that bringing more agents into the world, which is something that all of us are actually quite effective at doing, is human beings, has less of an effect than introducing these new cultural technologies.


he was right that writing had the effect of making people believe things because they were written down, and it had the effect of undermining memory capacities.


Probably most of us would say that writing in the long run was a good idea, but I think the jury might still be out about that one


even though Sam Adams is kind of a hero, like, he was really a master of misinformation and memes.


So the american revolution sort of starts with misinformation.


everyone had thought about common sense and the good, interesting pamphlets, but what he did was read the whole thing. And you will be amazed to discover that most of it was soft core porn and libel. So that was like, anytime you get a new cultural technology, the first thing, notoriously that anyone uses it for is pornography


And there's lots of beautiful anthropological work by people like Polly WiEsner that this is, in fact, what happens with fires. So if you look at the content of conversation in a forager group over the course of the day, mostly during the day, what you have is the adults talking to one another. And, you know, mostly what she calls CCC, which is criticism and complaint. Mostly what they do during the day is complain about how everybody else is not doing the things that they're supposed to be doing. But at night, what happens is the conversation shifts. And now, instead of the adults being in charge, it's the elders, the grandmothers, and the kids who are engaged, and the grandmothers are telling stories about the big, important things about the culture to the. To the kids.


gods, they're another example of the way that we use fictional agents to actually pass on cultural information.


one of the things that happens is you get this new literary form (novels), which is incredibly popular, which everybody reads, especially adolescent girls. Adolescent girls are always at the cutting edge of these changes in cultural technology. And everyone's always sure that it's a really, really, really bad idea that they're adopting the new cultural technology and it's going to lead to the death of civilization, rewiring their brains for anxiety or whatever it is.


cultural evolution depends on this balance between imitation and innovation.


there are cases of cultures getting stuck in a kind of local minimum where you're just passing on information from one generation to another, but you're not checking it against the world and you're not actually learning anything new.
Note: When did this happen? What examples?


So in order for this cultural transmission to really work, you have to have this balance where on the one hand, there are people who are going out in those non stationary environments and figuring them out. And on the other hand, other people are taking the discoveries that the people, the innovators have made and then applying them themselves and passing them on to the next generation
Note: What is the "new information" these days? What was it back with the printing press?

Seems like both were just serving to increase rate of transmission, which I guess helps with innovation?

Yes! So the point is, that these cultural technologies improve transmission which increases the non-stationarity of the world => on and on.


there's a fascinating twist. And the twist is this only happens if the mother is present both for rats and for humans. So for rats, it's like if the, even if just the male of the mother is, is present, then they'll explore the shock. But if it isn't, then they won't.


You see that exactly what these post menopausal orca grandmothers do is pass on information about past generations of foraging, which the younger, the grandchildren don't actually know about.


Ted Chang's beautiful novella, the life cycle of software objects.


I think the important point that I want to make is this kind of developmental diversity is a condition for intelligence
Note: Thesis part 2


Marie Antoinette, saying, let them eat cake, was actually, again, an invented meme that spread very quickly and that people were willing to believe because it was there in print. As far as we could tell, it never actually happened.