spicy take: it’s a mistake to think of emotions as uniquely more related to consciousness than any other part of human cognition. a lot of people seem to think of emotions as being special, or the key ingredient of consciousness, or something. but really, i don’t think my emotions are that much bigger or more fundamental a chunk of my consciousness than, say, my raw sensory perception, or my internal chain of thought.
my model of emotions as a cognitive system is they are a state machine that also does some tabular learning, and has a lot of privileged access to the body that upper parts of the brain don’t have direct access to. it implements a lot of heuristics that are very useful for guiding exploration, avoiding danger, etc, that the neocortex has not been pretrained as strongly for, but also its heuristics can be quite crude.
in much the same way that certain aspects of reasoning can be replicated by processes that almost everyone agrees are not conscious (eg chess engines), it seems clear that certain aspects of emotions can also be replicated by processes that are not conscious (eg a very tiny tabular RL setup).
one might object, but emotions feel like a really big part of our consciousness! even if they aren’t the biggest thing, surely it’s still evidence? but after all, when a human thinks really hard about how to play chess, it certainly feels like a very big part of their consciousness too, and we can agree that a chess engine (an old school one without any NNs) is obviously not conscious.
a corollary is i don’t think finding emotion-like things inside LMs, even if they aren’t just directly from imitating humans, is good evidence for ai consciousness.
as a tangent, i’d also say suffering and whatever-the-opposite-of suffering-is[1] are actually also separate things from pain and pleasure, which are separate again from happiness and sadness and so on. unfortunately, the term emotion tends to wrap all of these together. when i say emotion, i specifically mean the happiness sadness thing, as opposed to the other things.
i don’t really think pleasure and suffering are actually symmetric. probably the closest thing to the opposite of suffering is more like fulfillment, or joy, or contentedness, or something like that. the buddhists probably have some fancy term for this. i don’t have sufficient experience exploring the minutiae of these states to have a strong view on what exactly the right ontology is.
Emotions feel like a big part of consciousness because they are sort of the interface of the consciousness to the underlying machinery.
A big aspect of why consciousness feels like something special is that the mechanisms that generate consciousness are opaque to self-observation. We can’t self-introspect which sensory perceptions caused which emotions and thoughts and even less which neurons did what. That means that the self-inspecting process has no way to see that it is a “normal” physical process. The most accessible thing of the underlying machinery available to consciousness is emotions (meditation practitioners having more access notwithstanding).
Sure, basically all phenomenal experience is equivalent evidence of phenomenal consciousness. In some sense this just restates the hard problem, no? That the detection of neurophysiological activity corresponding to thought/emotions in humans/animals provides little insight into the character of phenomenal experience itself. Similarly, the detection of emotion/cognition in an LLM tells us little about whether an AI system has phenomenal consciousness. The difference being that with humans/animals we might infer that other subjects also have phenomenal experiences, since the architecture and substrate are the same/similar. With AI there are radical known physical differences, so we reasonably doubt this.
i mean yes it’s all hard problem in the end or something. but also i want to challenge the specific intuition that emotions are a more “special” part of consciousness
I like that you’re challenging that specific intuition! I wanted to reframe your point in standard philosophy of mind discourse to suggest this is not necessarily a “spicy take” outside folk theories of consciousness.
I think part of why people emphasize emotions in discussions about consciousness is because both consciousness and emotions are relevant to moral patiency. If an entity is conscious but has no emotions/preferences (or capacity to suffer/anti-suffer, if this is indeed a different thing), we arguably don’t owe it anything.
I’m not convinced emotions are even rated that highly in the more pure “hard problem” kind of discussions that often talk about things like the redness of red. And when it comes to the question whether AI is conscious in some more morally relevant way, rating emotions highly doesn’t seem unreasonable to me.
What would you consider good evidence for AI consciousness?
spicy take: it’s a mistake to think of emotions as uniquely more related to consciousness than any other part of human cognition. a lot of people seem to think of emotions as being special, or the key ingredient of consciousness, or something. but really, i don’t think my emotions are that much bigger or more fundamental a chunk of my consciousness than, say, my raw sensory perception, or my internal chain of thought.
my model of emotions as a cognitive system is they are a state machine that also does some tabular learning, and has a lot of privileged access to the body that upper parts of the brain don’t have direct access to. it implements a lot of heuristics that are very useful for guiding exploration, avoiding danger, etc, that the neocortex has not been pretrained as strongly for, but also its heuristics can be quite crude.
in much the same way that certain aspects of reasoning can be replicated by processes that almost everyone agrees are not conscious (eg chess engines), it seems clear that certain aspects of emotions can also be replicated by processes that are not conscious (eg a very tiny tabular RL setup).
one might object, but emotions feel like a really big part of our consciousness! even if they aren’t the biggest thing, surely it’s still evidence? but after all, when a human thinks really hard about how to play chess, it certainly feels like a very big part of their consciousness too, and we can agree that a chess engine (an old school one without any NNs) is obviously not conscious.
a corollary is i don’t think finding emotion-like things inside LMs, even if they aren’t just directly from imitating humans, is good evidence for ai consciousness.
as a tangent, i’d also say suffering and whatever-the-opposite-of suffering-is[1] are actually also separate things from pain and pleasure, which are separate again from happiness and sadness and so on. unfortunately, the term emotion tends to wrap all of these together. when i say emotion, i specifically mean the happiness sadness thing, as opposed to the other things.
i don’t really think pleasure and suffering are actually symmetric. probably the closest thing to the opposite of suffering is more like fulfillment, or joy, or contentedness, or something like that. the buddhists probably have some fancy term for this. i don’t have sufficient experience exploring the minutiae of these states to have a strong view on what exactly the right ontology is.
Funny that I just wrote this comment about AI and emotions saying basically the same thing without having read yours.
Emotions feel like a big part of consciousness because they are sort of the interface of the consciousness to the underlying machinery.
A big aspect of why consciousness feels like something special is that the mechanisms that generate consciousness are opaque to self-observation. We can’t self-introspect which sensory perceptions caused which emotions and thoughts and even less which neurons did what. That means that the self-inspecting process has no way to see that it is a “normal” physical process. The most accessible thing of the underlying machinery available to consciousness is emotions (meditation practitioners having more access notwithstanding).
Sure, basically all phenomenal experience is equivalent evidence of phenomenal consciousness. In some sense this just restates the hard problem, no? That the detection of neurophysiological activity corresponding to thought/emotions in humans/animals provides little insight into the character of phenomenal experience itself. Similarly, the detection of emotion/cognition in an LLM tells us little about whether an AI system has phenomenal consciousness. The difference being that with humans/animals we might infer that other subjects also have phenomenal experiences, since the architecture and substrate are the same/similar. With AI there are radical known physical differences, so we reasonably doubt this.
i mean yes it’s all hard problem in the end or something. but also i want to challenge the specific intuition that emotions are a more “special” part of consciousness
I like that you’re challenging that specific intuition! I wanted to reframe your point in standard philosophy of mind discourse to suggest this is not necessarily a “spicy take” outside folk theories of consciousness.
I think part of why people emphasize emotions in discussions about consciousness is because both consciousness and emotions are relevant to moral patiency. If an entity is conscious but has no emotions/preferences (or capacity to suffer/anti-suffer, if this is indeed a different thing), we arguably don’t owe it anything.
I’m not convinced emotions are even rated that highly in the more pure “hard problem” kind of discussions that often talk about things like the redness of red. And when it comes to the question whether AI is conscious in some more morally relevant way, rating emotions highly doesn’t seem unreasonable to me.
What would you consider good evidence for AI consciousness?
Do GPU’s have feelings?