Does Stockfish Dream of Electric Knights?

After 5 minutes I gave up reading the article. My cognitive abilities fell deep down the rabbit hole... Amazing work though!

Absolute masterpiece of a blog. My brain lost the plot about halfway through, but I might go back and read it again another day lol

Awesome blog. You've made my day.

Good work. It's easy to underestimate the human brain when comparing it against recent technologies. I found this post quite hopeful in that regard.

@RuyLopez1000 said ^

One GOFAI history note: Melanie Mitchell should also be credited for Copycat alongside Douglas Hofstadter (the work was a joint effort).

Thanks for the tip! I edited the blog to reflect this.

Another note: in the text below the Kasparov-DeepBlue picture you wrote "GOFI" whereas it should say "GOFAI".

Great blog! I shared it in my Cognitive Science Chess group!

@derjosh said ^

One GOFAI history note: Melanie Mitchell should also be credited for Copycat alongside Douglas Hofstadter (the work was a joint effort).

Thanks for the tip! I edited the blog to reflect this.

Another note: in the text below the Kasparov-DeepBlue picture you wrote "GOFI" whereas it should say "GOFAI".

Great blog! I shared it in my Cognitive Science Chess group!

Curious about the Cog Sci Chess group - I'm a cog sci professor (visual recognition and development, primarily) and I'm always interested in meeting more people interested in chess & cognition.

Consciousness remains ill-defined and should probably be avoided in such papers aiming at being scientific.
From my understanding you miss the two main scientific theories of predictive processing and eliminativism (not to mention spiritual approaches where consciousness does not derive from the brain).
For the researcher, the mind is still a black box, we are limited to psychometrics (measuring the output of mind, not the mind) and self reporting. The other main method is reverse anthropomorphism, modeling the mind in computer programs and testing the results against psychometrics and self reporting. Added to neural correlates of consciousness (leading method of research) of trying to correlate individual acts with neural events.

Cognitive Architectures is now a field of computer science, hardly part of psychology, but still a leading area of research as reverse anthropomorphisms are currently the best method to study the human mind, although quite limited.

I still find chunking theory to be the best field of research, that can likely explain the difference between the human mind's approach to problem solving, and the computer's approach.

Another important philosophical precursor is the question of Platonism (ideas are true independent of mind), a postulate that most researchers want to reject. Can you mathematically / logically derive rules of mind and apply those to both the computer and the human mind? That question is still one of the most debated among mathematicians and philosophers, but not much discussed among chess researchers, as presumably most chess players are Platonists, and assume the answers to the Truth of chess exist outside of the human mind, and the computer and human mind both do their best to find / match the mind-independent Platonic Truth to chess.

I had mentioned in one of my blog posts 'Does one need a Theory of Mind to understand Chess Improvement?'
https://lichess.org/@/DIAChessClubStudies/blog/science-of-consciousness-conference--the-penrose-hameroff-hypothesis/2TSoKyIQ

But maybe will try a post on the metaphysical presuppositions to understanding chess improvement?

But first, are you a Platonist? I have noted studies and surveys that only about 35% of academic philosophers are Platonists (with Nominalism being slightly more popular than Platonism). As where among Mathematicians over 53% are Platonists.
But what would a Nominalist approach to chess mastery look like? My guess is that among chess academics the percent of Platonists is higher than the even mathematicians. But good research would need to be able to describe the findings in Nominalist terminology.

@DIAChessClubStudies said ^

Thanks!

Consciousness remains ill-defined and should probably be avoided in such papers aiming at being scientific.

Bernard Baars gave an operational definition of consciousness: verbal report of a stimuli that can be verified. This was just a first approximation to begin understanding the concept. Baars noted that you can be conscious of something without reporting. But we need a way to study the problem empirically and this definition is suitable. Pairs of stimuli can be given, where one stimuli becomes conscious and the other doesn't, so that conscious and non-conscious data can be contrasted.

He noted a number of features of consciousness such as the narrowness of consciousness where it only takes in a limited amount of items at time, but it somehow controlled by the vast amount on unconscious information we store, which can pop into consciousness at any moment.

He also noted how consciousness resolves ambiguous stimuli to form a coherent interpretation. One experiment he described had subjects to pay attention to one of two simultaneous audio streams presented through headphones (different for each ear). Sentences like 'They were standing near the bank' being interpreted either as a financial institution or a river depending on what sentences the subject was given in the unattended ear.

Consciousness also leads to adaption. New skills are originally conscious as they are being learned, but become unconscious and automatic as they are mastered.

These observations and Global Workspace Architectures from AI helped form his Global Workspace Model of consciousness.

Bernard Baars - Cognitive Theory of Consciousness Great book.

From my understanding you miss the two main scientific theories of predictive processing and eliminativism (not to mention spiritual approaches where consciousness does not derive from the brain).

I chose to cover Global Workspace because it is well supported by the neuroscience evidence and conceptual understanding of the roles which consciousness plays. I chose IIT as a foil to GWT.

For the researcher, the mind is still a black box, we are limited to psychometrics (measuring the output of mind, not the mind) and self reporting. The other main method is reverse anthropomorphism, modeling the mind in computer programs and testing the results against psychometrics and self reporting. Added to neural correlates of consciousness (leading method of research) of trying to correlate individual acts with neural events.

Neural events through fMRI/EEG and other methods doesn't seem like a black box? It gives us something.

Cognitive Architectures is now a field of computer science, hardly part of psychology, but still a leading area of research as reverse anthropomorphisms are currently the best method to study the human mind, although quite limited. I still find chunking theory to be the best field of research, that can likely explain the difference between the human mind's approach to problem solving, and the computer's approach.

Another important philosophical precursor is the question of Platonism (ideas are true independent of mind), a postulate that most researchers want to reject. Can you mathematically / logically derive rules of mind and apply those to both the computer and the human mind? That question is still one of the most debated among mathematicians and philosophers, but not much discussed among chess researchers, as presumably most chess players are Platonists, and assume the answers to the Truth of chess exist outside of the human mind, and the computer and human mind both do their best to find / match the mind-independent Platonic Truth to chess.

I had mentioned in one of my blog posts 'Does one need a Theory of Mind to understand Chess Improvement?'
https://lichess.org/@/DIAChessClubStudies/blog/science-of-consciousness-conference--the-penrose-hameroff-hypothesis/2TSoKyIQ

But maybe will try a post on the metaphysical presuppositions to understanding chess improvement?

I'd be interested in that.

But first, are you a Platonist?

My logic is that ideas must be represented by the brain. Therefore no brain = no ideas. Meaning that ideas do not exist outside of the brain.

There is also the question of what an idea is.

I have noted studies and surveys that only about 35% of academic philosophers are Platonists (with Nominalism being slightly more popular than Platonism). As where among Mathematicians over 53% are Platonists.

Interesting.

But what would a Nominalist approach to chess mastery look like? My guess is that among chess academics the percent of Platonists is higher than the even mathematicians. But good research would need to be able to describe the findings in Nominalist terminology.

We got some fMRI studies of brain areas that activate in chess. But that doesn't explain how chess develops and allow a cognitive model to be created.

Consciousness (Global Workspace) integrates and broadcasts information from and to numerous unconscious processors. It is the reason we can learn complex skills and understand very nuanced contexts. So a cognitive model of chess should try to utilize Global Workspace Theory to explain chess skill.

Consciousness in medicine is similarly narrowly defined to 'awareness / control' and only part of anesthesiology, as otherwise consciousness is not a medical term, and might not exist.

But in philosophy and psychology consciousness often if a category to encompass all mental activity, together with sense perception and active control. The mind is an ancient term, and implies a non-material entity, and was well defined by the Greeks and others with clear definitions of what the functions of the mind were, as a non material entity. Even though we can scan the brain, and correlate those to thought patterns (neural correlates of consciousness), at best that is just a correlate, and the mind aspect is still unknown (as Cartesian understanding of brain as a receiver, not producer of thought from Platonic realm).

If you are a Nominalist, like the majority of philosophers, researchers, implying that ideas only exist within 'brains', the question of universal ideas is debated, as ideas only exist within individual mind, and may or may not match with ideas in other minds. But implies the 'search for universal truth' is misleading, as there is no universal truth, just individual minds with some overlap in what individual minds produce.

Not enough chess researches for a survey of chess researchers to know the popularity of Nominalism and Platonism. But probably good that the few regardless of Nominalist and Platonist collaborate. But if there is a 'Universal Truth' to chess, does it exist within the mind, or separate from the mind, and what role does the computer play in that? I lean towards 'Monist' interpretations of reality, and would argue that computers also access a non material Truth from the Platonic realm.

Blessings.

Predictive Processing is important for this discussion, as PP implies that 'consciousness' works exactly like Stockfish. From the smallest aspect of sense perception, to higher levels of 'control' of the body, to cognitive functioning, every aspect of consciousness goes through some sort of algorithm / formula, to determine the best path forward, and automatically chooses that (with some minor possibility of free will, choosing between a few top options).

Eliminativism approaches to Predictive Processing, where there is no free will or decision. As the research of Fodor and Libet imply that the mind has a compartment that produces 'explanations', but the compartment of the mind that produces explanations actually involved in control and deciding. So self-reporting is largely useless, and even as a personal meditative enlightenment, one needs to recognize that the reasons our own mind tell us why we do something is not actually why we do things, but produced separately for different reasons.