Theory of Mind May Have Spontaneously Emerged in Large Language Models
1–10 of 321 posts
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#2Abstract:
Theory of mind (ToM), or the ability to impute unobservable mental states to others, is central to human social interactions, communication, empathy, self-consciousness, and morality. We administer classic false-belief tasks, widely used to test ToM in humans, to several language models, without any examples or pre-training.
Our results show that models published before 2022 show virtually no ability to solve ToM tasks. Yet, the January 2022 version of GPT-3 (davinci-002) solved 70% of ToM tasks, a performance comparable with that of seven-year-old children. Moreover, its November 2022 version (davinci-003), solved 93% of ToM tasks, a performance comparable with that of nine-year-old children.
These findings suggest that ToM-like ability (thus far considered to be uniquely human) may have spontaneously emerged as a byproduct of language models' improving language skills.
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#3Abstract: Theory of mind (ToM), or the ability to impute unobservable mental states to others, is central to human social interactions, communication, empathy, self-consciousness, and morality. We administer classic false-belief tasks, widely used to test ToM in humans, to several language models, without any examples or pre-training. Our results show that models published before 2022 show virtually no ability to sol…
> These findings suggest that ToM-like ability (thus far considered to be uniquely human)
What it suggests to me is that the particular test of “Theory of Mind” tasks involved actually test the ability to process language and generate appropriate linguistic results, not theory of mind.
It also suggests (with the “thus far considered to be uniquely human”) that the authors are unaware of other theory of mind tests that have been used that are not language dependent but behavior dependent, and on which, while, as is also true of linguistic tests, the validity of the tests is controversial – a number of non-human primates, non-primate mammals, and even some birds (parrots and corvids, particulary) have shown evidence of theory of mind.
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#4My belief, based on experiences with domestic and wild animals is that there is nothing uniquely human about "theory of mind".
It's a running gag in our household (where my wife runs a riding academy) that academics just published a paper showing that some animal (e.g. horse) has just been proven to have some cognitive capability that seems pretty obvious if you work with those animals.
It's very hard to know what is going in animal's heads
https://en.wikipedia.org/wiki/Theory_of_mind#Non-human
but I personally observe all kinds of social behavior that sure seems like "Horse A looks to see what Horse B thinks about something Horse A just spotted" (complete with eye-catching on both sides) and such.
There was an article about how Chimpazees and humans were found to have a common vocabulary of gestures and I was by no means impressed, I mean, so far as I can tell mammals and birds have a universal language for "pointing" to things in the environment. Even my cats point things out to me.
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#5My belief, based on experiences with domestic and wild animals is that there is nothing uniquely human about "theory of mind". It's a running gag in our household (where my wife runs a riding academy) that academics just published a paper showing that some animal (e.g. horse) has just been proven to have some cognitive capability that seems pretty obvious if you work with those animals. It's very hard to know what is…
I'm reminded of the early days of "AI" when that consisted of building chess engines, because chess is what "intelligent" people do. They quickly realized that they were solving the wrong problem, doing the thing that humans are bad at and computers are good at.
In a sense language models appear to be doing the same thing again, one step down the scale. They're doing a human-specific thing, but missing whatever it is that non-human vertebrates do, and mammals do pretty well. I believe that this is the vast majority of human cognition, too. We just don't talk about it because when we talk about thinking, we're talking, and confuse the two.
These language models have done jaw-dropping things, and also make it abundantly clear that there's some fundamental thing that they've completely missed. It's plausible that that "thing" could emerge all by itself, using a mechanism entirely different from vertebrate cognition and yet somehow sufficient. Or it could be like the chess engines, doing something amazing and yet ultimately limited and of minimum utility.
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#6Abstract: Theory of mind (ToM), or the ability to impute unobservable mental states to others, is central to human social interactions, communication, empathy, self-consciousness, and morality. We administer classic false-belief tasks, widely used to test ToM in humans, to several language models, without any examples or pre-training. Our results show that models published before 2022 show virtually no ability to sol…
> These findings suggest that ToM-like ability (thus far considered to be uniquely human) What it suggests to me is that the particular test of “Theory of Mind” tasks involved actually test the ability to process language and generate appropriate linguistic results, not theory of mind. It also suggests (with the “thus far considered to be uniquely human”) that the authors are unaware of other theory of mind tests tha…
It's hard to look at behaviour separately from language if the only behaviour available is to generate text. As long as we don't have a test agnostic of medium, this will have to do.
In the end, we can't overcome the limitation that all we can empirically see is the ability to process X and generate appropriate Y. If that invalidates the test where X is language and Y is language, what stops us from invalidating any possible X and Y? That would leave us no empirical method to work with.
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#7My belief, based on experiences with domestic and wild animals is that there is nothing uniquely human about "theory of mind". It's a running gag in our household (where my wife runs a riding academy) that academics just published a paper showing that some animal (e.g. horse) has just been proven to have some cognitive capability that seems pretty obvious if you work with those animals. It's very hard to know what is…
Theory of mind is a popular research topic in animals. There is no controversy in the idea that the attribution of mental states occurs in animals.
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#8My belief, based on experiences with domestic and wild animals is that there is nothing uniquely human about "theory of mind". It's a running gag in our household (where my wife runs a riding academy) that academics just published a paper showing that some animal (e.g. horse) has just been proven to have some cognitive capability that seems pretty obvious if you work with those animals. It's very hard to know what is…
I recall reading that pointing is something that only a few animals, including us, do and understand.
If I get my cats' attention and point at something, they're more interested in the tip of my finger than the direction I'm pointing at.
Now, my cats will occasionally meow to get my attention and then walk over to where the problem is - an empty food dish, an empty water bowl, the bed that they expect us to be in because hello, it's bedtime according to my internal cat-clock - but they never engage in pointing behavior.
Anyway, time to cite some stuff, instead of dropping anecdotes into a bucket:
- https://www.wired.co.uk/article/elephant-pointing
> A study by researchers from the University St Andrews has found that elephants are the only wild animals that can understand human pointing without being trained. > Pointing in humans is a behaviour that develops at a very early age -- usually before a child reaches 12 months – as it is an immediate way of controlling the attention of others. "Most other animals do not point, nor do they understand pointing when others do it," says Professor Richard Byrne, one of the authors of the study. "Even our closest relatives, the great apes, typically fail to understand pointing when it's done for them by human carers; in contrast, the domestic dog, adapted to working with humans over many thousands of years and sometimes selectively bred to follow pointing, is able to follow human pointing -- a skill the dogs probably learn from repeated, one-to-one interactions with their owners."
- https://www.researchgate.net/publication/7531526_A_comparati...
> If the distance between the tip of the indexfinger and the object was greater than 50 cm, subjects per-formed poorly in contrast to trials where the pointing fingeralmost touched the baited box. We should also note that insome trials/experiments the pointers also turned their headand looked in the same direction thereby enhancing thecommunicative effect of the gesture but even so the differ-ence did not disappear. Even after training, chimpanzees in Povinellietal’s (1997) experiment were just able to master the task.
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#9My belief, based on experiences with domestic and wild animals is that there is nothing uniquely human about "theory of mind". It's a running gag in our household (where my wife runs a riding academy) that academics just published a paper showing that some animal (e.g. horse) has just been proven to have some cognitive capability that seems pretty obvious if you work with those animals. It's very hard to know what is…
I recall reading that pointing is something that only a few animals, including us, do and understand. If I get my cats' attention and point at something, they're more interested in the tip of my finger than the direction I'm pointing at. Now, my cats will occasionally meow to get my attention and then walk over to where the problem is - an empty food dish, an empty water bowl, the bed that they expect us to be in bec…
By "pointing" I don't mean pointing your finger, the universal version of "pointing" involves your eyes, shoulders and the rest of your body. (Think of a how a hunting dog points... Plenty of times I've seen cats point out things to other cats not to mention to me.)
Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models
#10They are still big state-machines, unlike the human brain.