timelets | Entries tagged with cognition

The transition from the ‘red’ of awareness to the ‘red’ of thought is accompanied by a definite loss of content, namely by the transition from the factor ‘red’ to the entity ‘red.’ This loss in the transition to thought is compensated by the fact that thought is communicable whereas sense-awareness is incommunicable.

-- Whitehead, The Concept of Nature.
https://www.gutenberg.org/files/18835/18835-h/18835-h.htm

This is different for AI. Even more specifically, Jeff Dean introduced a mode of "teacher-student" transmission where both weights and elements of the model transferred between entities.

The word ‘_perceive_’ is, in our common usage, shot through and through
with the notion of cognitive apprehension. So is the word
‘_apprehension_’, even with the adjective _cognitive_ omitted. I will
use the word ‘_prehension_’ for _uncognitive apprehension_: by this I
mean _apprehension_ which may or or may not be cognitive.

Whitehead. Science..., 1925.

----

If we modeled this prehension as a CT monad we could develop and algebra of prehension, with quantative thresholds that separate the uncognitive and cognitive.

https://ncatlab.org/nlab/show/state+monad

X -> [W, WxY]

Here the operation [W, Wx(-)) is the monad on the type system which is induced by the above adjunction; and this latter function is naturally regarded as a morphism in the Kleisli category of this monad.

“The genre has been defined as an art of “cognitive estrangement”. This means, in the first place, that science fiction distances us from our everyday cognitive assumptions and frames of reference – which is something that philosophy is also supposed to do. But the definition also implies, at least in some instances, that science fiction works to estrange us from the very possibility of being able to cognize our “immediate experience” at all. In science fiction narratives, cognition may fail because new technologies “alter sense ratios or patterns of perception” so radically that there is no evident pathway from here to there; or because the sort of subjectivity that we take for granted has broken down; or because we encounter alien forms of sentience that are not commensurable with our own.”

Steven Shaviro. “Discognition.”

====
LLMs can be trained to show that can't be shown, i.e. confabulate consistently on a massive scale.

“We can no longer be satisfied with the old stimulus/response model, according to which animals (and other organisms) passively respond to prior, incoming stimuli, and learn by means of conditioning (or associations among these stimuli). For this is only one part of the story. In addition, and much more importantly, biological entities are active reality-testers. They are always busy “probing the environment with ongoing, variable actions first and evaluating sensory feedback later (i.e., the inverse of stimulus response)”. Rather than just responding to stimuli, they exhibit ongoing activity that is self-generated, and only secondarily modulated by stimuli. Output tends to come first.”

Steven Shaviro. “Discognition.”

If a research experiment were well defined, it would be altogether unnecessary to perform it. For the experimental arrangements to be well defined, the outcome must be known in advance; otherwise the procedure cannot be limited and purposeful. The more unknowns there are and the newer a field of research is, the less well defined are the experiments.
...
And if after years we were to look back upon a field we have worked in, we could no longer see or understand the difficulties present in that creative work. The actual course of development becomes rationalized and schematized. We project the results into our intentions; but how could it be any different? We can no longer express the previously incomplete thoughts with these now finished concepts.

Cognition modifies the knower so as to adapt him harmoniously to his acquired knowledge.

-- Ludwik Fleck, Genesis and Development of a scientific fact.

similar to Bezos' "wandering."

Deficiency of understanding we call stupidity: deficiency in the application of reason to practice we shall recognise later as foolishness: deficiency of judgment as silliness, and lastly, partial or entire deficiency of memory as madness.

-- Arthur Schopenhauer, The World As Will And Idea.
https://www.gutenberg.org/cache/epub/38427/pg38427-images.html

What makes it the case that something counts as a form of cognitive success? For instance, why think that knowing the capital of Pakistan is a cognitive success, rather than just another cognitive state that an agent can occupy, like having 70% confidence that Islamabad is the capital of Pakistan? Not every cognitive state enjoys cognitive success. Knowing, understanding, mastering—these are cognitive successes.

[e.g]...we describe a person as “knowing” something as a way of signaling that her testimony with respect to that thing is to be trusted.

https://plato.stanford.edu/entries/epistemology/

Machines can now be trained to generate information traditionally associated with cognitive successes by humans. Today, we know relatively well what kind of cognitive success is achievable for a majority humans. We are still in the beginning of the process of building cognitively successful computers. We simply don't know what they are capable of when trained by a small number of highly successful humans.

ЖЖ-шные траповатники, ранее крутые специалисты по разоружению Северной Кореи, стали еще более крутыми специалистами по подсчету голосов, например, в Пенсильвании.

Все-таки интересно, откуда у людей берется эта уверенность рассуждать о вещах, в которых они ничего не смыслят? Как эта иллюзия понимания работает в мозгах физиологически? Наблюдая определенную активность нейронов, мы уже можем предупреждать приступы эпилепсии. Наверное, в будущем можно будет предупреждать приступы глупости.

I've got a crazy hypothesis that modern political polarization is caused by information overload. That is, people's cognitive abilities are so strained that they can't handle political complexity and resort to a primitive two-value system "us vs them." Furthermore, since politics for the vast majority of us is a rare, low-priority, low-skill activity with no real-life consequences, rational cognitive allocation should be equal to zero (as "investors" we are in no position to gain political capital at all). Therefore, the only return on politics is social (e.g. group cohesion) and/or emotional (e.g. BIRG) capital.

How would I test it?

I've always wondered why labeling something as "idiotic" or "very very bad" makes people happy. In rational thinking, labeling is considered fundamentally flawed because it simply shows how the labeler positions an object/event on his/her internal value map without explaining the reasoning why s/he does so. As John R. Searle would say, something is good or bad in virtue of something else, not because its intrinsic goodness or badness.

Yesterday, I found a review paper* that describes multiple psychological experiments on Schadenfreude. For example,

In the context of a real-world sports rivalry, Red Sox and Yankees fans report feeling pleasure, and show activity in reward- related brain regions (i.e., right ventral striatum including nucleus accumbens) when they watch their rival fail to score against their favored baseball team, and also against a less competitive team in the same league (i.e., the Orioles). Attaching positive value to outgroup members' suffering may provide motivation for inflicting suffering: People who show more reward-related activity when watching the rival team fail also report being more likely to actively harm the rival team’s fans (Cikara, Botvinick, & Fiske, in press). These findings extend to situations in which the rival fans themselves are in physical pain: Soccer fans exhibited reward-related activity (again, the right ventral striatum) when watching a rival team’s fan receive a painful electric shock; the magnitude of this activity predicted participants’ later unwillingness to relieve the rival’s pain by receiving half of the electric shock themselves (Hein, Silani, Preuschoff, Batson, & Singer, 2010).

It looks like, labeling facilitates Schadenfreude-based pleasure. In other words, a social activity that looks irrational from a purely logical perspective, has significant psychological advantages because it creates positive in-group empathy.

* Cikara, M., E. G. Bruneau, and R. R. Saxe. “Us and Them: Intergroup Failures of Empathy.” Current Directions in Psychological Science 20.3 (2011): 149–153. Web. 13 Apr. 2012.

Crossposts: http://timelets.livejournal.com/627878.html

395. There is a lack of clarity about the role of imaginability in our investigation. Namely, about the extent to which it ensures that a sentence makes sense.

-- Investigations.

I wonder whether AI is going to have a big advantage over humans in imagining things. Furthermore, two (or more) AI agents equipped with superior imagination can come up with a productive agreement/disagreement much faster than humans. All this comes on top of superior data access, processing, communications, and memory capabilities.

Crossposts: http://timelets.livejournal.com/599620.html

It just occurred to me that our perception is necessarily a process of splitting the world into parts and then combining them into various wholes. Biologically, we are made up of cells, which have their own preceptors. Therefore, different cells perceive different aspects of the world and then communicate to assemble them according to behavioral scenarios. I can probably say that we are biological expressions of the world.

Crossposts: http://timelets.livejournal.com/595249.html

Cognition early in training is fundamentally different from cognition late in training.

...information in an expert’s long-term memory...is organized differently from the information in a novice’s long-term memory. Experts don’t think in terms of surface features, as novices do; they think in terms of functions, or deep structure.

The novices ...[generate]... categories based on the objects in the problems.

...transfer [using known info to apply to new problems] is so difficult because novices tend to focus on surface features and are not very good at seeing the abstract, functional relationships among problems that are key to solving them.

...experts are able to ignore unimportant details and home in on useful information; thinking functionally makes it obvious what’s important.

--- Willingham. Why students don't like school? 2009.

Here's the teaching/learning dilemma:
- On the one hand, you want to teach "objects", i.e surface features, so that novices can learn.
- On the other hand, you want to teach functional relationships, i.e. abstract, deep structures, so that learners acquire useful skills.

In other words, if you are an expert you can't teach novices "the real thing" and must dumb down the material, so that it becomes learnable.

As a side note, Category Theory can be a good teaching/learning tool for experts because it focuses on functional relationships, instead of objects.

— counterfactual reasoning for past and future: ‘what if’ questions and mental scenarios in the context of preparation; ‘what if’ questions that seem to serve fantasizing; ‘what if’ questions that pertain to the past versus the future; and ‘what if’ questions that relate to self versus others. Are they all different? Perhaps not as much as it would seem,
with all promoting simulations and imagery, possibly creating ‘memories’ that can be applied in future situations.

http://www.oxfordscholarship.com/view/10.1093/acprof:oso/9780195395518.001.0001/acprof-9780195395518
DOI:10.1093/acprof:oso/9780195395518.001.0001

There must be a better way to present this information.

Contrary to popular belief, the brain is not designed for thinking. It’s designed to save you from having to think, because the brain is actually not very good at thinking.Thinking is slow and unreliable. Nevertheless, people enjoy mental work if it is successful. People like to solve problems, but not to work on unsolvable problems.
...
For problems to be solved, the thinker needs adequate information from the environment, room in working memory, and the required facts and proce- dures in long-term memory.

-- Daniel T. Willingham. Why Don't Students Like School. 2009.

An interesting piece of research on developmental cognition:

In a monumental study involving 709 children, Alloway et al. tested the development of working memory capacity with a number of tasks.48 They documented a linear increase in working memory capacity from age four to age eleven.
...
An increase of only one chunk in working memory could herald the emergence of an entire range of qualitatively distinct cognitive operations.
doi: 10.1111/nyas.12189

First, I wanted to write that a good pre- and elementary school would be essential for development. Then, I thought that we don't quite know what "good" means in this context. Maybe the Finnish education model works because they expose kids to a broad variety of tasks from an early age.