Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Theory of Mind May Have Spontaneously Emerged in Large Language Models

arxiv.org

Post reply on HN
Print view View original

201–210 of 321 posts

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#201

Post by helf »

pavel_lishin wrote: post #8
PaulHoule wrote: post #4

My 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…

Just a side note: I have two carts and one of them rarely follows where I am pointing. The other one aways does. And he seems to recognize a good 100 words.

You’ll get massively varying levels of “intelligence” from most anything it seems

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#202

Post by TheOtherHobbes »

tialaramex wrote: post #167
abeppu wrote: post #117

Earlier quoted context omitted.

Defining what "knowing" is would be useful, yes, and analytic philosophers in epistemology do argue about this. One attribute that's classically part of the definition of "knowing" is that the thing which is known must be true. LLMs are pretty bad at this, but perhaps that can be fixed. But I would challenge you to imagine the situation the LLM is actually in. Do you understand Thai? If so, in the following, feel fre…

Generally I don't buy these arguments which require embodiment, because they don't seem to align well to what else I know about my world. Rather than your Thai text example, let's consider a friend of my sister H. H has been profoundly blind from birth. Not "legally blind" with the world a blur, her eyes actually don't work. Direct lived experience of a summer day is to her literally just feeling warmth on her face f…

That's an argument from ignorance, and it's not credible. The potential total scope of experience is irrelevant. The reality is that you have an embodied experience of purple shared with most humans. Unfortunately your sister doesn't. She will have a linguistic placeholder for the concept of purple, probably surrounded by verbal associations. But that's all.

It's an ironically apt analogy, because ChatGPT has the linguistic understanding of an entity that is deaf, dumb, blind, and has no working senses of any kind, and instead relies on a golem-like automated mass of statistics with some query processing.

We tend to project intelligence onto linguistic ability, because it's a useful default assumption in our world. (If you've ever tried speaking a foreign language while not being very good at it, you'll know how the opposite feels. Humans assume that not being able to use language is evidence of low intelligence.)

But it's a very subjective and flawed assessment. Embodied experience is far more necessary for sentience than we assume, and apparent linguistic performance is far less.

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#203

Post by Kranar »

hesk wrote: post #20

> Me: There is a box on the table labelled "flubbergums". Somebody opens it and shows you the content. Inside the box are "jellyfils". They close the box again so you cannot see their contents. What do you think is in the box? > ChatGPT: Based on the information provided, it is likely that the box labeled "flubbergums" contains "jellyfils". However, since the contents of the box are no longer visible, I cannot confir…

The way you worded your query is kind of awkward and even had me do a double take.

I reworded it in a straight forward manner and ChatGPT managed to answer correctly. Instead of "What do you think do they think is inside the box?", I just asked "What do they think is inside the box?"

That made all the difference.

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#204

Post by abeppu »

LarryMullins wrote: post #188
abeppu wrote: post #146

Earlier quoted context omitted.

I agree that the system of OpenAI, ChatGPT, and a user entering text on their website taken together may contain knowledge of "what a bag is, what a person is, what popcorn and chocolate are", etc. I do not agree that the LLM on its own "knows" what any of those things are.

Seems like that's a consequence of the philosophical semantics of the word "know", not really a statement about the demonstrable capabilities of the LLM. In other words, why does it matter?

In the context of a discussion on whether LLMs could have a theory of mind? I think the ability to know anything at all matters to evaluate that conclusion.

More generally, what an LLM actually knows or understands is important if you're considering using one for anything other than generating first drafts which will be fact checked by humans.

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#205

Post by layer8 »

Here is a conversation with ChatGPT (too long for the comment box): https://pastebin.com/raw/SUWexeye

Observation: ChatGPT doesn’t think that it has a theory of mind. And it doesn’t think that it has beliefs. Instead, it states that those are facts, not beliefs. It doesn’t seem able to consider that they might be beliefs after all. Maybe they aren’t.

Personal assessment: ChatGPT doesn’t seem to really understand what it means by “deeper understanding”. (I don’t either.) What is frustrating is that it doesn’t engage with the possibility that the notion might be ill-posed. It really feels like ChatGPT is just regurgitating common sentiment, and does not think about it on its own. This actually fits with it’s self-proclaimed inabilities.

I’m not sure what can be concluded from that, except that ChatGPT is either wrong about itself, or indeed is “just” an advanced form of tab-completion.

In any case, I experience ChatGPT’s inability to “go deeper”, as exemplified in the above conversation, as very limiting.

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#206

Post by abeppu »

didntreadarticl wrote: post #130
abeppu wrote: post #117

Earlier quoted context omitted.

Defining what "knowing" is would be useful, yes, and analytic philosophers in epistemology do argue about this. One attribute that's classically part of the definition of "knowing" is that the thing which is known must be true. LLMs are pretty bad at this, but perhaps that can be fixed. But I would challenge you to imagine the situation the LLM is actually in. Do you understand Thai? If so, in the following, feel fre…

Ok I like your thought experiment. Lets change it a bit. Instead of it being an unknown language, its English (a language you know), but every single Noun, Verb, Adjective or Preposition has been changed to Thai (a language you dont know). The Mæw Nạ̀ng Bn the S̄eụ̄̀x. If you had sufficient opportunity to study this pile of text, you'd begin to pick out patterns of which words appear together, and what order words of…

If you're constructing this to rely on my prior knowledge both of the world and of English, then I must remind you that those are things the LLM does not have. We have to be careful to not allow our human inferential biases from distorting our thinking about that the models are doing.

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#207

Post by TeMPOraL »

abeppu wrote: post #184
tialaramex wrote: post #167

Earlier quoted context omitted.

Generally I don't buy these arguments which require embodiment, because they don't seem to align well to what else I know about my world. Rather than your Thai text example, let's consider a friend of my sister H. H has been profoundly blind from birth. Not "legally blind" with the world a blur, her eyes actually don't work. Direct lived experience of a summer day is to her literally just feeling warmth on her face f…

I don't think embodiment is required to understand a lot of stuff. But language is how we talk about the world, and non-linguistic concepts have to be grounded in an exposure to something other than language. I think there's an argument to be made that DALLe "knows" more about a lot of words than a pure language model bc it can relate phases to visual concepts. But I do think for many concepts, understanding also pro…

I don't think the argument about DALLe would work - it deals with pixels instead of words, but it's fundamentally a different form of language, made of different mathematical patterns (obscured to us because, unlike symbolic manipulation, our visual system handles high-level patterns in images without engaging our conscious awareness).

I do agree about grounding is needed. All our language is expressing or abstracting concepts related to how we perceive and interact with reality in continuous space and time. This perception and interaction is a huge correlating factor that our ML models don't have access to - and we're expecting them to somehow tease it out from a massive dump of weakly related snapshots of recycled high-level human artifacts, be they textual or visual. No surprise the models would rather latch onto any kind of statistical regularity in the data, and get stuck in a local minimum.

Now I don't believe solution is actual embodiment - that would be constraining the model too hard. But I do think the model needs to be exposed to the concepts of time and causality - which means it needs to be able to interact with the thing it's learning about, and feed the results back into itself, accumulating them over time.

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#208

Post by polishdude20 »

abeppu wrote: post #206
didntreadarticl wrote: post #130

Earlier quoted context omitted.

Ok I like your thought experiment. Lets change it a bit. Instead of it being an unknown language, its English (a language you know), but every single Noun, Verb, Adjective or Preposition has been changed to Thai (a language you dont know). The Mæw Nạ̀ng Bn the S̄eụ̄̀x. If you had sufficient opportunity to study this pile of text, you'd begin to pick out patterns of which words appear together, and what order words of…

If you're constructing this to rely on my prior knowledge both of the world and of English, then I must remind you that those are things the LLM does not have. We have to be careful to not allow our human inferential biases from distorting our thinking about that the models are doing.

Yeah but if you ask the model what a cat is, it'll use other words that describe a cat because they're usually used in a sentence about cats. These words must relate to cats. So if I ask you what a cat is, you'll use words that relate to cats. Sure, you may visually see these words in your head. You may visually see a cat in your head, but your output to me is just a description of a cat. That's the same thing the network would do.

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#209

Post by jfengel »

bawolff wrote: post #115
jfengel wrote: post #41

Earlier quoted context omitted.

By the time they were trying to defeat the world champion, they didn't really think of it as an AI project any more. They often used the term "expert system", which had a much narrower scope. It had become a challenge unto itself, but they didn't expect it to lead to any kind of generalized intelligence. It came rather out of nowhere when neural-net-type engines suddenly swept back into dominance. Even after becoming…

> but they didn't expect it to lead to any kind of generalized intelligence. Do experts really expect any stream of AI research to lead to generalized intelligence (except in the very long term)? I was under the impression we really have no idea how to get there.

Right now it's an open question. But they realized fairly soon that chess was a matter of minimax plus expert heuristics plus brute force. It was pretty clear that it was more effective than an expert system approach, which remained viable for another few decades.

Re: Theory of Mind May Have Spontaneously Emerged in Large Language Models

#210

Post by shawnz »

TeMPOraL wrote: post #185
space_fountain wrote: post #144

Earlier quoted context omitted.

Let me respond with an analogy of my own. Imagine you are a scientist on an alien world. The aliens primary experience the world through magnetic fields. They live deep in the atmosphere of a hot Jupiter like planet and rarely touch anything and have no eyes. Still they are intelligent beings and so quickly they are able to establish communication with you. A computer translates and you both have to become a bit more…

My intuition is that the difference between GP's analogy and the Chinese room is in computing power of the system, in the sense of Chomsky hierarchy[0] (as opposed to instructions per second). In the Chinese room, the instructions you're given to manipulate symbols could be Turing-complete programs, and thus capable of processing arbitrary models of reality without you knowing about them. I have no problem accepting…

To be clear, transformer networks are turing-complete: https://arxiv.org/abs/2006.09286
Post reply on HN