Some of those names are familiar, but right now I don’t think my brain can handle reading more. The putative subjective dimension of experience bugs me because it makes it sound like experience isn’t subjective and I don’t know how to deal with that.
There is also this view she mentioned at the start of another article:
It has recently been proposed that the framework of semantic relativism be put to use to describe mental content, as deployed in some of the fundamental operations of the mind. This programme has inspired in particular a novel strategy of accounting for the essential egocentricity of first-personal or de se thoughts in relativist terms, with the advantage of dispensing with a notion of self-representation. This paper is a critical discussion of this strategy. While it is based on a plausible appeal to cognitive economy, the relativist theory does not fully account for the epistemic profile that distinguishes de se thinking, as some of its proponents hope to do. A deeper worry concerns the reliance of the theory on a primitive notion of “centre” that hasn’t yet received enough critical attention, and is ambiguous between a thin and a rich reading. I argue that while the rich reading is required if the relativist analysis of the de se is to achieve its most ambitious aims, it also deprives the theory of much of its explanatory power.
Like…what exactly does it mean to dispense with self representation? Is she saying that because such thoughts are essentially about you there is no need to think of yourself as “you” first? What does semantic relativism mean exactly?
If i try to read more about any of this stuff my brain will just interpret it negatively even when I know the right or more realistic interpretation
Exactly my point. Do you really need to read more? Or should you focus on how to maintain your physical health and how to calm your mind?
I have no idea. I don’t need an answer to that right now. I can choose how to eat properly without knowing that. I can select the right candidate in the midterms without knowing that. I can use paper, plastic, or my own bag at the grocery store without knowing that.
I know. That’s what I’ve been saying. I understand this completely. This is you. Work on this. Trying to answer the unanswerable is the cause, not the solution.
Probably calm my mind first, but it’s a conflict between that and the stuff I’ve read. I can’t just forget it and it’s rattled things I’ve taken for granted like time and the self.
I guess that’s the difference, I can’t do any of that stuff without knowing this.
I’ve tried to stop it but I can’t, I don’t know how. Asking the therapist is a dead end because all they say is how awful and exhausting it is.
I emailed that lady earlier and this was her reply to me:
Good evening,
There are very many classical texts on the self you might find useful in thinking through your questions.
Bibliographies and other helpful resources can be found here:
But clicking on the philpapers link led to stuff I don’t understand. But worse is this sinking notion that if I don’t read all those then I’m missing some great truth or choosing a lie.
Through a series of meetings, the Santa Fe Institute plans to explore the dual nature of time as a means of capturing what changes and what does not. How theories of change make strong assumptions about properties that are static and to states approaching equilibrium. We are interested in how historical theories typically ignore time and replace it with events, sequences, and contingencies. And why humanity’s greatest achievements and insights are ennobled as timeless — free from decay, corruption, and perhaps even replacement.
You’re alive, so you are doing stuff. You also just admitted the one thing that identifies what you are doing, obsessing over these internet posts, is an unhealthy obsession. You have consistently shown me that you are spending so much time and energy on this, that it is getting in the way of your existence and your own good choices about health and happiness.
The secret of life is that there is no secret of life. If you want to try to understand time, neuroscience, theory of mind, whatever, fine. But don’t expect to come up with an answer because those fields of study exist because there are still questions. It’s the same as “where did the universe come from?” There is no answer to that, yet, here we are.
More quotes at Enlightenment Quotes (3165 quotes)
some of which are pretty bad, some are confusing, some are funny. none of them are the answer.
This human form of intelligence can be considered truly emergent because it involves the dis‑ covery of effective theories of knowledge—compact and low‑dimensional languages that screen‑ off most of neuroscience and even much of psychology. Indeed, folk psychology is in many ways a successful effective theory: its concepts allow humans to attribute beliefs and desires to other people, and frequently successfully predict and explain their behaviour [66]. Movies, art, poli‑ tics and journals, like the one in which this article appears, would not exist without shared folk psychological concepts.
The causal power of human understanding is also incontestable. A recent paper illustrates this by contrasting how two different forms of intelligence—procedural and understanding‑based, can induce an identical neural network structure [67]. Briefly, monkeys and humans can be trained to learn two sets of pairwise associations between arbitrary symbols. Each set of associations is defined as a context. What has been shown is that, through trial and error, monkeys and humans can generate a representational geometry across a large population of neurons in hippocampus that permits implicit inference of the latent contextual variable. What is truly surprising is that by providing verbal instruction, a human (unlike a monkey) can configure this structure in min‑ utes rather than through laborious training. This ‘instant’ intelligence is based in this case on understanding an instructed sentence. An effective theory made of psychological constructs can causally change a neural network structure. There are many such examples of short circuiting of trial and error learning by quickly ‘getting it’ [68–70], for example, in a study in which humans and LLMs were asked to reason about letter and symbol sequences after being instructed to use a modified “alphabet” of permuted letters or non‑letter symbols. The simple instruction allowed humans to perform the task nearly perfectly, while LLMs struggled substantially .
Especially when those insights seem to offer a reductive explanation of humans and our existence. Hearing our deepest moments be referred to merely folk psychology is pretty demoralizing to hear.
What deepest moments? This is talking about how we learn. What explanation of knowledge would you like hear? What would give you a sense of agency in the world? You know there are atoms and chemical reactions, right?
It’s more about like reducing those special moments, those most intimate of reactions to some sort of cognitive model or some construct, I dunno…to me it just makes our lives feel…fake?
Did you read through the original article, particularly section 4 about Knowledge Out EmergenceL
There is a useful distinction to be made between the kind of emergence seen in many‑body physi‑ cal and chemical systems and that seen in complex adaptive systems. In the former case, emergent properties are assumed to be generated from large numbers of simple components following rel‑ atively simple rules (e.g. individual molecules obeying Newton’s laws). By contrast, in complex systems that adapt or learn, such as biological organisms, brains, economies and LLMs, emer‑ gent properties such as bodily organs, stock indices or cognitive capabilities are generated by mechanisms that extract large amounts of ‘knowledge’—structured information including facts and rules—from pre‑existing complex environments. We will say that emergence in chemistry and physics has a ‘knowledge‑out’ (KO) character, since it is characterized by complex structure or behaviour arising from simple interactions among simple elements. Conversely, emergence in complex adaptive systems has a ‘knowledge‑in’ (KI) character, since it is characterized by complex structure or behaviour arising from complex inputs or environments [23].
You don’t want to know. You don’t want to define your question any better than this vague notion of not liking that you are a creature in a universe. That universe is beyond your comprehension and that upsets you.
It’s been fun getting to know you, but there’s only so much I can say in response. Here’s a question, what are you thankful for? Are there people or things in your life that you appreciate?
It’s not so much a creature in the universe so much as it might be the fact that there might not be anything special about us. That our stuff could be explained by either psychology, information theory, or genetics and that there is no magic to it. It just makes human beings feel more like machines than anything else.
They found that in bees changing the genes might impact their social behavior and we see similar aspects among humans, it’s hard not to think of us as machines when you put it that way.
Magic? You would feel better if there was magic? The irony is, there are people like Chalmers who say we may never be able to explain human consciousness. You have scientists and philosophers on your side, saying there is something special about us that chemistry and biology has not yet figured out. Not to mention the billions of people who believe Gods did it.
Now you’re going to say something like, “well, not so much magic, but …” Because you don’t want to settle on anything. You want the argument. You aren’t comfortable with the mystery of life but you don’t like the answers that we can get by observing life. You don’t want the journey or the destination.
The journey depends on how I view others and treat them which is impacted by the stuff I read. Maybe in the past we could argue that we many never be able to explain it, or maybe they could argue there is something special. But as science and human knowledge advances I find it harder and harder to justify believing that any more.
"We consider biological individuality in terms of information theoretic and graphical principles. Our purpose is to extract through an algorithmic decomposition system-environment boundaries supporting individuality. We infer or detect evolved individuals rather than assume that they exist. Given a set of consistent measurements over time, we discover a coarse-grained or quantized description on a system, inducing partitions (which can be nested). Legitimate individual partitions will propagate information from the past into the future, whereas spurious aggregations will not. Individuals are therefore defined in terms of ongoing, bounded information processing units rather than lists of static features or conventional replication-based definitions which tend to fail in the case of cultural change. One virtue of this approach is that it could expand the scope of what we consider adaptive or biological phenomena, particularly in the microscopic and macroscopic regimes of molecular and social phenomena.
We individuals are no longer assumed anymore (this is Information Theory of Individuality). You are right that I don’t like the answers, because they erode the pillars that can allow me to have a happy life.
The Information Theory of Individuality (Santa Fe Institute / Krakauer): This perspective treats the “individual” not as a fixed physical entity, but as an ongoing, bounded information-processing unit. By using methods like coarse-graining, you can track aggregates that successfully propagate information from the past into the future. This justifies preserving “gradients and process” while still defining a cohesive identity across any scale, from molecular to cultural.
There isn’t really something that describes a happy life for me anymore because it’s all been eroded by everything I’ve read. The more I read and think the less a happy life makes any logical sense.
What made me happy was working toward my own goals, seeking to make friends, maybe a partner, having a job I loved and just being content in the choices I made in life.
But when the things you read break down the notion of what it means to be human then all of that goes out the window.
In attempting to describe the differences between natural and artificial intelligence they end up breaking down the human experience into some sort of computational and clinical model. The opening is by Douglas Hofsatder which says there is nothing extra that would make our “I” different from that of an AI.
Here, in a nutshell, is the thesis of this essay: When words ‘act like’ things in the world, then they refer to those things; then they mean those things. If and when that happens, then thinking is taking place behind the scenes of those words. And where there is thinking, there is consciousness and a genuine, full-fledged ‘I’… We need to believe in our ‘I’’s in order to survive in this very complex and fast-moving world. We need reliable and drastic simplifications of all the animate agents around us, and a fortiori of ourselves… And when we apply such abstract concepts to ourselves, we produce (or rather, we hallucinate) an ‘I’, which is to say, we hallucinate a consciousness, and we attach it to the physical body that it is housed in… I’m talking about consciousness as a rich self-model, just like a model of anything else in the world, except that it is self-pointed… I suspect it’s not going to be all that hard to get these systems (or more sophisticated hybrid systems integrating [large language models] with physical robots that roam the world) to focus on their own behavior, and to do for themselves something vaguely like what we humans do for ourselves—namely, to personify themselves, so to speak.
Happiness for me right now (if I can even call me a “me” anymore given what I’ve read) is a dead dream.
Pedro Tsividis, João Loula, Jake Burga, Juan Pablo Rodriguez, Sergio Arnaud, Nathan Foss, Andres Campero, Ajay Subramanian, Thomas Pouncy, Samuel Gershman and Joshua Tenenbaum investigate how humans learn complex novel tasks with remarkable speed and generality. The authors study human performance on simple-but-structurally rich video games that require rapid exploration, model construction and goal-directed planning, showing that people acquire competent play within only a few minutes. They introduce the Exploring, Modelling and Planning Agent (EMPA) as a novel model-based reinforcement learning architecture that aims to capture human-like intelligence with intuitive theories of objects, agency, symbolic state descriptions, causal transition models and capacities for simulating hypothetical futures. EMPA’s architecture optimizes performance by identifying informative interactions; augmenting sparse rewards with theory-generated subgoals; as well as modelling object categories, interaction rules and conditions for terminating behaviour. It also includes a flexible exploration policy that seeks out the most informative aspects of hypothesis space, as well as a sophisticated planner and meta-controller that selects among planning modes over multiple temporal scales. EMPA matches human learning efficiency on a majority of games, outperforming deep reinforcement learning baselines by orders of magnitude in sample efficiency and exhibiting behavioural patterns that align with human exploration, including targeted interaction with task-relevant objects and rapid avoidance of dangerous ones. The authors show that EMPA can still achieve near-human efficiency in unfamiliar games, suggesting that theory-based modelling, exploration and planning may provide a general framework for understanding human-like learning across diverse environments. The cognitive architecture introduced by Tsividis et al. not only models the world with powerful conceptual abstractions but also refines those models (via exploration) and navigates the world with sophisticated forms of planning. They further leverage the same kinds of inductive biases and ‘core knowledge’ [4,40] that are believed to contribute to the remarkable ability of biological intelligences to learn so much from so little (c.f. the ‘less is more’ principle discussed by Krakauer et al.). It is exciting to think of what could be enabled by using such approaches along with foundation models such as LLMs [41], and the extent to which this could allow for agents with sufficient intelligence that they can be relied upon in real-world use cases. The conjunction of an LLM with a world-modelling agent architecture could be more similar to human minds in which syntactic processes are coupled to capacities for imagination/simulation (or embodied self-world modelling). However, to achieve human-like performance outside game environments, to what extent do systems need to have sufficiently human-like experiences such that they could develop similarly grounded self-world models [37]? Significant advances were made in this work by incorporating exploration-optimizing features, consistent with findings that intrinsic drives for curiosity are a powerful means of bootstrapping world models [8]. The following contribution describes how even more progress could be made by equipping systems with intrinsic drives for empowerment, so bringing them closer to the forms of agency found in biological intelligence.
It actually does when you realize how much of that stuff is just our brains modeling reality:
Nadav Amir, Stas Tiomkin and Angela Langdon present a computational framework in which the descriptive and prescriptive aspects of world models arise jointly from an agent’s goals, defined as preference relations over probability distributions of action–observation sequences, or experiences. They argue that standard reinforcement learning fails to capture many forms of purposeful behaviour, especially those driven by intrinsic motivations or temporally extended experiential features, because it separates state representations from value functions and assumes that goals can be expressed in terms of scalar rewards. Drawing on Buddhist philosophical accounts of concept formation, the authors propose that cognitive agents categorize experiences according to their relevance for achieving goals rather than according to veridical correspondence to external states. They formalize this proposal through the construct of ‘telic states’, defined as equivalence classes of experience distributions that are equally preferred under a given goal. A telic state representation is a goal-induced partitioning of the space of possible experience distributions, determining which experiential distinctions matter and which can be ignored when acting toward a desired end. The authors argue that telic states offer a unified account of behavioural, phenomenological and neural aspects of purpose-driven activity by integrating intrinsic preferences, subjective experience and representational structure. They then discuss implications for intrinsically motivated agents, preference-based learning, state abstraction, multi-objective decision making and the broader relationship between goals and world modelling in cognitive and artificial systems
I encourage you to read the article because it essentially proves the point I’m making.
Georg Northoff, Yasir Catal and Samira Abbasi propose that computing and AI can benefit from understanding how the brain uses its intrinsic neural dynamics to engage with the world’s physical time. The paper argues that the physics of complex systems provides temporality, understood as continuously unfolding change, and that the brain leverages spontaneous neural activity as ‘inner time’, which enables it to track, encode and align with the temporal structures of environmental inputs across multiple scales. Multiple empirical phenomena are shown to demonstrate that neural systems mirror the frequencies and periodicities of external inputs, that variability increases with input complexity and uncertainty, and that such dynamic encoding is actively modulated by individuals as they couple with the world. This context-sensitive world-mirroring, time-varyingly graded and active encoding strategy is interpreted as the basis for the organism’s ‘being in time’ and ‘being in the world’, enabling ‘tacit knowledge’ and flexible behaviour. In contrast, the paper argues that current artificial and natural computing systems lack this kind of alignment with time unfolding over multiple scales, leading to a ‘temporal locked-in syndrome’ in which they process inputs discretely and non-temporally, so limiting their intelligence and behavioural flexibility. Northoff, Catal and Abbasi’s work suggests that key principles of intelligence in complex adaptive systems may not be present in approaches to AI based on pretraining in which inference and learning are artificially separated. According to this view, natural systems inherit much of their intelligence from the properties of physics whereby coupling (near-critical) systems become mutually entrained in their dynamics. LLMs can receive feedback from their environment via post-training fine-tuning and reinforcement learning methods, but this is of a different kind from the temporospatial alignment that allows biological organisms to self-organize to better adapt to their environments. Adaptive capacity from entraining innerand outer-dynamics can also be thought of as a (potentially very) special case of the feedbackbased regulation and world modelling described by Alicea et al. as well as the emergent properties of scaling and criticality described by Krakauer et al. Even more, this perspective could be taken to suggest particular functional roles for ‘consciousness’—as integrated world modelling [17,18,20,47]—to the extent that subjective experience both affords and is afforded by the temporospatial alignment of dynamic relations between system and world [47]. The following contribution further explores the ways in which dynamical systems afford potentially conscious world modelling in the service of adaptive behaviour.
