This right here is also the reason why AI fanboys get angry when they are told that LLMs are not intelligent or even thinking at all. They don’t understand that in order for rational intelligence to exist, the LLMs should be able to have an internal, referential inner world of symbols, to contrast external input (training data) against and that is also capable of changing and molding to reality and truth criteria. No, tokens are not what I’m talking about. I’m talking about an internally consistent and persistent representation of the world. An identity, which is currently antithetical with the information model used to train LLMs. Let me try to illustrate.
I don’t remember the details or technical terms but essentially, animal intelligence needs to experience a lot of things first hand in order to create an individualized model of the world which is used to direct behavior (language is just one form of behavior after all). This is very slow and labor intensive, but it means that animals are extremely good, when they get good, at adapting said skills to a messy reality. LLMs are transactional, they rely entirely on the correlation of patterns of input to itself. As a result they don’t need years of experience, like humans for example, to develop skilled intelligent responses. They can do it in hours of sensing training input instead. But at the same time, they can never be certain of their results, and when faced with reality, they crumble because it’s harder for it to adapt intelligently and effectively to the mess of reality.
LLMs are a solipsism experiment. A child is locked in a dark cave with nothing but a dim light and millions of pages of text, assume immortality and no need for food or water. As there is nothing else to do but look at the text they eventually develop the ability to understand how the symbols marked on the text relate to each other, how they are usually and typically assembled one next to the other. One day, a slit on a wall opens and the person receives a piece of paper with a prompt, a pencil and a blank page. Out of boredom, the person looks at the prompt, it recognizes the symbols and the pattern, and starts assembling the symbols on the blank page with the pencil. They are just trying to continue from the prompt what they think would typically follow or should follow afterwards. The slit in the wall opens again, and the person intuitively pushes the paper it just wrote into the slit.
For the people outside the cave, leaving prompts and receiving the novel piece of paper, it would look like an intelligent linguistic construction, it is grammatically correct, the sentences are correctly punctuated and structured. The words even make sense and it says intelligent things in accordance to the training text left inside and the prompt given. But once in a while it seems to hallucinate weird passages. They miss the point that, it is not hallucinating, it just has no sense of reality. Their reality is just the text. When the cave is opened and the person trapped inside is left into the light of the world, it would still be profoundly ignorant about it. When given the word sun, written on a piece of paper, they would have no idea that the word refers to the bright burning ball of gas above them. It would know the word, it would know how it is usually used to assemble text next to other words. But it won’t know what it is.
LLMs are just like that, they just aren’t actually intelligent as the person in this mental experiment. Because there’s no way, currently, for these LLMs to actually sense and correlate the real world, or several sources of sensors into a mentalese internal model. This is currently the crux and the biggest problem on the field of AI as I understand it.
How do hallucinations preclude an internal representation? Couldn’t hallucinations arise from a consistent internal representation that is not fully aligned with reality?
I think you are misunderstanding the role of tokens in LLMs and conflating them with internal representation. Tokens are used to generate a state, similar to external stimuli. The internal representation, assuming there is one, is the manner in which the tokens are processed. You could say the same thing about human minds, that the representation is not located anywhere like a piece of data; it is the manner in which we process stimuli.
Not really. Reality is mostly a social construction. If there’s not an other to check and bring about meaning, there is no reality, and therefore no hallucinations. More precisely, everything is a hallucination. As we cannot cross reference reality with LLMs and it cannot correct itself to conform to our reality. It will always hallucinate and it will only coincide with our reality by chance.
I’m not conflating tokens with anything, I explicitly said they aren’t an internal representation. They’re state and nothing else. LLMs don’t have an internal representation of reality. And they probably can’t given their current way of working.
You seem pretty confident that LLMs cannot have an internal representation simply because you cannot imagine how that capability could emerge from their architecture. Yet we have the same fundamental problem with the human brain and have no problem asserting that humans are capable of internal representation. LLMs adhere to grammar rules, present information with a logical flow, express relationships between different concepts. Is this not evidence of, at the very least, an internal representation of grammar?
We take in external stimuli and peform billions of operations on them. This is internal representation. An LLM takes in external stimuli and performs billions of operations on them. But the latter is incapable of internal representation?
And I don’t buy the idea that hallucinations are evidence that there is no internal representation. We hallucinate. An internal representation does not need to be “correct” to exist.
Yet we have the same fundamental problem with the human brain
And LLMs aren’t human brains, they don’t even work remotely similarly. An LLM has more in common with an Excel spreadsheet than with a neuron. Read on the learning models and pattern recognition theories behind LLMs, they are explicitly designed to not function like humans. So we cannot assume that the same emergent properties exist on an LLM.
That’s not how science works. You are the one claiming it does, you have the burden of proof to prove they have the same properties. Thus far, assuming they don’t as they aren’t human is the sensible rational route.
Read again. I have made no such claim, I simply scrutinized your assertions that LLMs lack any internal representations, and challenged that assertion with alternative hypotheses. You are the one that made the claim. I am perfectly comfortable with the conclusion that we simply do not know what is going on in LLMs with respect to human-like capabilities of the mind.
This right here is also the reason why AI fanboys get angry when they are told that LLMs are not intelligent or even thinking at all. They don’t understand that in order for rational intelligence to exist, the LLMs should be able to have an internal, referential inner world of symbols, to contrast external input (training data) against and that is also capable of changing and molding to reality and truth criteria. No, tokens are not what I’m talking about. I’m talking about an internally consistent and persistent representation of the world. An identity, which is currently antithetical with the information model used to train LLMs. Let me try to illustrate.
I don’t remember the details or technical terms but essentially, animal intelligence needs to experience a lot of things first hand in order to create an individualized model of the world which is used to direct behavior (language is just one form of behavior after all). This is very slow and labor intensive, but it means that animals are extremely good, when they get good, at adapting said skills to a messy reality. LLMs are transactional, they rely entirely on the correlation of patterns of input to itself. As a result they don’t need years of experience, like humans for example, to develop skilled intelligent responses. They can do it in hours of sensing training input instead. But at the same time, they can never be certain of their results, and when faced with reality, they crumble because it’s harder for it to adapt intelligently and effectively to the mess of reality.
LLMs are a solipsism experiment. A child is locked in a dark cave with nothing but a dim light and millions of pages of text, assume immortality and no need for food or water. As there is nothing else to do but look at the text they eventually develop the ability to understand how the symbols marked on the text relate to each other, how they are usually and typically assembled one next to the other. One day, a slit on a wall opens and the person receives a piece of paper with a prompt, a pencil and a blank page. Out of boredom, the person looks at the prompt, it recognizes the symbols and the pattern, and starts assembling the symbols on the blank page with the pencil. They are just trying to continue from the prompt what they think would typically follow or should follow afterwards. The slit in the wall opens again, and the person intuitively pushes the paper it just wrote into the slit.
For the people outside the cave, leaving prompts and receiving the novel piece of paper, it would look like an intelligent linguistic construction, it is grammatically correct, the sentences are correctly punctuated and structured. The words even make sense and it says intelligent things in accordance to the training text left inside and the prompt given. But once in a while it seems to hallucinate weird passages. They miss the point that, it is not hallucinating, it just has no sense of reality. Their reality is just the text. When the cave is opened and the person trapped inside is left into the light of the world, it would still be profoundly ignorant about it. When given the word sun, written on a piece of paper, they would have no idea that the word refers to the bright burning ball of gas above them. It would know the word, it would know how it is usually used to assemble text next to other words. But it won’t know what it is.
LLMs are just like that, they just aren’t actually intelligent as the person in this mental experiment. Because there’s no way, currently, for these LLMs to actually sense and correlate the real world, or several sources of sensors into a mentalese internal model. This is currently the crux and the biggest problem on the field of AI as I understand it.
That’s an excellent methaphor for LLMs.
It’s the Chinese room thought experiment.
Hadn’t heard about it before (or maybe I did but never looked into it), so I just went and found it in Wikipedia and will be reading all about it.
So thanks for the info!
No worries. The person above did a good job explaining it although they kind of mashed it together with the imagery from Plato’s allegory of the cave.
How do hallucinations preclude an internal representation? Couldn’t hallucinations arise from a consistent internal representation that is not fully aligned with reality?
I think you are misunderstanding the role of tokens in LLMs and conflating them with internal representation. Tokens are used to generate a state, similar to external stimuli. The internal representation, assuming there is one, is the manner in which the tokens are processed. You could say the same thing about human minds, that the representation is not located anywhere like a piece of data; it is the manner in which we process stimuli.
Not really. Reality is mostly a social construction. If there’s not an other to check and bring about meaning, there is no reality, and therefore no hallucinations. More precisely, everything is a hallucination. As we cannot cross reference reality with LLMs and it cannot correct itself to conform to our reality. It will always hallucinate and it will only coincide with our reality by chance.
I’m not conflating tokens with anything, I explicitly said they aren’t an internal representation. They’re state and nothing else. LLMs don’t have an internal representation of reality. And they probably can’t given their current way of working.
You seem pretty confident that LLMs cannot have an internal representation simply because you cannot imagine how that capability could emerge from their architecture. Yet we have the same fundamental problem with the human brain and have no problem asserting that humans are capable of internal representation. LLMs adhere to grammar rules, present information with a logical flow, express relationships between different concepts. Is this not evidence of, at the very least, an internal representation of grammar?
We take in external stimuli and peform billions of operations on them. This is internal representation. An LLM takes in external stimuli and performs billions of operations on them. But the latter is incapable of internal representation?
And I don’t buy the idea that hallucinations are evidence that there is no internal representation. We hallucinate. An internal representation does not need to be “correct” to exist.
And LLMs aren’t human brains, they don’t even work remotely similarly. An LLM has more in common with an Excel spreadsheet than with a neuron. Read on the learning models and pattern recognition theories behind LLMs, they are explicitly designed to not function like humans. So we cannot assume that the same emergent properties exist on an LLM.
Nor can we assume that they cannot have the same emergent properties.
That’s not how science works. You are the one claiming it does, you have the burden of proof to prove they have the same properties. Thus far, assuming they don’t as they aren’t human is the sensible rational route.
Read again. I have made no such claim, I simply scrutinized your assertions that LLMs lack any internal representations, and challenged that assertion with alternative hypotheses. You are the one that made the claim. I am perfectly comfortable with the conclusion that we simply do not know what is going on in LLMs with respect to human-like capabilities of the mind.
Wtf are you even talking about.
They are right though. LLM at their core are just about determining what is statistically the most probable to spit out.
Your 1 sentence makes more sense than the slop above.