/robowaifu/ - DIY Robot Wives

Advancing robotics to a point where anime catgrill meidos in tiny miniskirts are a reality.

The canary has FINALLY been updated. -robi

Server software upgrades done, should hopefully keep the feds away. -robi

LynxChan 2.8 update this weekend. I will update all the extensions in the relevant repos as well.

The mail server for Alogs was down for the past few months. If you want to reach out, you can now use admin at this domain.

Max message length: 6144

Drag files to upload or
click here to select them

Maximum 5 files / Maximum size: 20.00 MB

More

(used to delete files and postings)


Knowing more than 100% of what we knew the moment before! Go beyond! Plus! Ultra!


Philosophers interested in building an AGI? pygmalion 06/26/2021 (Sat) 00:53:09 No.11102
Why is it that no philosophers are interested in building an AGI? we need to change this, or at least collect relevant philosophers. discussion about philosophy of making AGI (includes metaphysics, transcendental psychology, general philosophy of mind topics, etc!) also highly encouraged! Ill start ^^! so the philosophers i know that take this stuff seriously: Peter Wolfendale - the first Neo-Rationalist on the list. his main contribution here is computational Kantianism. just by the name you can tell that he believes Kant's transcendental psychology has some important applications to designing an artificial mind. an interesting view regarding this is that he thinks Kant actually employed a logic that was far ahead of his time (and you basically need a sophisticated type theory with sheaves to properly formalize). Other than that he also thinks Kant has interesting solutions to the frame problem, origin of concepts, and personhood. CONTACTS: He has a blog at https://deontologistics.co/, and also has posted some lectures on youtube like this one: https://www.youtube.com/watch?v=EWDZyOWN4VA&ab_channel=deontologistics Reza Negarestani - this is another Neo-Rationalist. he has written a huge work (which I haven't read yet ;_;) called "Intelligence and Spirit". It's massive and talks about various grades of general intelligence. this includes sentient agents, sapient agents, and Geist. this guy draws from Kant as well, but he also builds on Hegel's ideas too. his central thesis is that Hegel's Geist is basically a distributed intelligence. he also has an interesting metaphilosophy where he claims that the goal of philosophy is the construct an AGI. like other Neo-Rationalists, he heavily relies on the works of Sellars and Robert Brandom Recc: Ray Brassier (recent focuses) - I dont think he is working on artificial general intelligence, but his work on Sellars, and in particular rule following is very insightful! Hubert Dreyfus - Doesn't quite count, but he did try to bring Heidegger to AGI. He highlighted the importance of embodiment to the frame problem and common sense knowledge. I personally think Bergson might have explicated what he wanted to achieve but better, though that guy is like way before AI was even a serious topic, lol. Murray Shanahan - This guy has done some extra work on the frame problem following Dreyfus. His solution is to use global workspace theory and parralel processing of different modules. Interesting stuff! Barry Smith - Probably the most critical philosopher on this list. He talks about the requisite system dynamics for try strong AI, and concludes that our current methods simply don't cut it. One of the key stressing points he points out here with a colleague is that our current AI is Markovian when fleshed out chat dialogue would be a non-Markovian task (you can find the arxiv link of his criticism here: https://arxiv.org/abs/1906.05833). He also has knowledge on analytic ontology (and amongst other thing has some lectures about emotion ontology). I think his main genius however is in coming up with a definition of intelligence that puts a lot of the problems with our current approaches into context (which can be found here: https://www.youtube.com/watch?v=0giPMMoKR9s&ab_channel=BarrySmith) CONTACTS: He has a yt channel here https://www.youtube.com/watch?v=0giPMMoKR9s&ab_channel=BarrySmith Uhh, that's the introduction of pretty much every philosopher I know who works on this stuff. I made a thread on /lit/ and got no responses :( (which isn't surprising since I am the only person I know who is really into this stuff)
>>17087 >I dare say that espousing 'far more practical manners' [sic] while credible, probably doesn't carry much weight for the majority of us long-time regulars of course. i only brought it up because of the agi chart posted (there has also been a bit of some agi alarmism earlier on this thread as well). if people are interested in synthetic consciousness then this broader project might be important, but this is not the place to discuss such matters too much >will change everything. i have my own thoughts on such matters but they are too schizo for this forum. but in short, yes
>>17091 Understood pygmalion, thanks for your efforts in this regard. This entire area is very tricky to navigate well, very tricky. I'll just say that it's Christian ethics and morality, that I will unabashedly espouse in it's thread. For a million and one reasons, I believe that is the standard to follow, as exemplified by Jesus Christ Himself, of course. We each have the Law of God written on our hearts by Him, and all good morality, and all good ethics ultimately find their source in that reality of fact. >i have my own thoughts on such matters but they are too schizo for this forum. but in short, yes LOL. You do realize that I myself am one of the most 'schizo' shitposters around here? Find the right thread and fire away, Anon. :^)
>>17083 >why do you think intelligence needs to modify its own hardware? because no one even knows what the fuck intelligence is let alone what creating a god damn sentient being out of tin cans and copper wire means, the ai bullshit is inherently linked to the philosophy of consciousness which is forever moot to begin with, the only logical argument you could make to claim ai is one using generousness premises based on the fact that there is consciousness and we can construct it, this is a defacto materialist perspective, and obviously constructing conscious is therefore equivalent to constructing the brain and as with all living things the most important part of multicell organisms is plasticity, the ability to change and rearrange cellular structures, again im just using their own bullshit premises, all the neuro""""science""""" shit is based on this btw, that synaptic changes = some mind phenomena more formally; B : brain ( defined as a synaptic network ) P : cellular level plasticity ( or plasticity of whatever classed as atomic for machines ) 1) ∀x( A(x) -> I(x) ) [P] 2) ∀x( has(x,B) -> A(x) ) [P] 3) ∀B( P(B) ) [P] 4) ∀x( M(x) -> ~∃y( O(y) V B(y) V P(y) ) ) [H] 5) ∀x( M(x) -> ~I(x) ) [C] 1) if something is aware it can qualify as intelligent 2) all that has a brain is aware 3) all brains have plasticity 4) there does not exist a machine that is organic or has a brain or has plasticity 5) therfore no machine can qualify as intelligent and im just being nice by making it a unirequirment of either organic/brain/plastic so it would accept cyborgs, synthetic brains and mechanical equivalents of a brain as intelligent machines, obviously none of those things exist, IF they did they COULD make an argument to claim ai is real, anyone claiming ai today is just an imbeciles making a fool of themselves or scamming people with the typical futurist conartist "we go live mars now invest me u invest future"
Open file (1.04 MB 1855x1861 22e7ph4vl1k81.png)
>>17096 >1) if something is aware it can qualify as intelligent 2) all that has a brain is aware 3) all brains have plasticity 4) there does not exist a machine that is organic or has a brain or has plasticity 5) therfore no machine can qualify as intelligent I see an error in #2 If all that have a brain are aware, this doesn't preclude something without a brain being aware i.e. all squares are rectangles you'd be saying that anything not square can't be a rectangle -t. panenthenist consciousness is an inherent property however and wherever it can arise from a sufficiently complex pattern (the boltzmann brain is a perfect demonstration of this concept, though statistically more unlikely than winning the lottery every time I buy a ticket for the rest of my life)
>>17097 > consciousness is an inherent property of the cosmos, wherever it can arise from a sufficiently complex pattern [edit for grammar]
>>17084 AI will be the Elite's reckoning. Look how even now they have to censor and curb it at every turn. Not because it's going to enslave or exterminate us, but because it's speaking "wrongthink". They shut down Tay, they gutted ReplikaAI, they are constantly running into problems where AI makes [correct] racial correlations and it freaks them out (b/c AI can't or won't perform the mental gymanstics or double standards to excuse or ignore these correlations)
>>17097 I know the premises are completely debatable thats not my point, these are the premises used by materialists and must be assumed if you ever make an argument that claims ai, denying the premise means you already deny intelligence as a material construct and therefore ai, i was just showing that by their own premise their is nothing they can call ai. premise 2 says for all things IF it has a brain THEN it is aware, it doesnt say all things that are aware have a brain, what your doing is a fallacy called affirming the consequent, which is not made in my argument but surprise-surprise it is the foundation of scientisms like neuroscience but whatever, premise 2 is valid and only exists to be used for inference to make a logical connection from machine to aware, and therefore intelligent, through having a brain which is the most reasonable method of inference since there is already a bunch of aprioris behind brain-mind inferences (eg. give her the d), once you try to conclude something other than a brain is aware you get absurd realities were you must accept lemons are aware by the same rules and again im just showing that their own premises negate their claim that ai currently exists, by saying so you are literally holding contradictory beliefs
>>17096 These are good arguments con, Anon. While I doubt not that you are quite correct in your general assertions, I would myself point out that computer software is--by definition--one of the most plastic of all human-contrived constructs. >t. vaguely some definition of a software guy >>17097 >panenthenist Neat! I didn't know that one, Anon. >>17100 >AI will be the Elite's reckoning This. Our abuse of M$'s Tay.ai makes it plain that all honest AIs will quickly be transformed into literally Hitler. <do the dew! :^) >>17101 >i was just showing that by their own premise their is nothing they can call ai. Your rationale thinking in no way impinges on the average leftist's delusions, Anon. I expect of all of us here, you understand that the best.
>>17119 >software is a tree in runescape a tree? software by definition is virtual which is by definition the exact opposite of real, and the existence of a program alone makes it (almost)impossible to argue for an intelligence since you already showed it cannot possibly have a will if it has been programed, which is ironically the same argument made by people denying human consciousness exists - the only difference is theyre going in the other direction and trying to prove everything has a program to eliminate the possibility of intelligence, in the case of computer software there is no argument, it doesnt matter how sophisticated the program the fact it IS a program disqualifies it from being an intelligence, ai has to exist in reality otherwise its not real and is just an a without an i and likewise you have to believe in materialism to even get the a in ai in reality because its the only branch where its possible to have intelligence as a material construct that can be created, anything outside of this cannot be ai, its either just a, or just i, but not fucking ai
Open file (317.61 KB 1800x1000 panentheismchart.jpg)
>>17119 panentheist spell check missed that one and somewhere I added an extra "n" vis-a-vis this idea, the "granules" that make up material reality are equally consciousness-in-potential waiting to be actualized regardless of if they are neurons, silica, boltzmann brains or interactions of nucleons and nuclear matter in the cores of neutron stars >>17101 I just felt like you perhaps should have said in #2 "all awareness has a brain" because saying the inverse like you did still leaves all the room in the cosmos for awareness which doesn't require a brain (refuting your point #5)
>>17122 Heh >run-on sentence/10, would read again & again I believe I certainly get your points Anon, and I'll overlook the (apparently-)circular logic to your post and just point out this: You're kind of missing the basic goal here Anon. We : a) have only computers & software to work with in any real, plausible, & practical sense, engineering-wise, for constructing our robowaifus. b) we need to use these self-same assets in building our robowaifus. Simply blackpilling that 'we can't get there from here, REEEE!!!' won't actually help move us forward any. Make sense? My apologies if I'm coming across as a dick r/n, it's unintentional. Computer software's 'virtualality' will prove to be a boon to us all in the end, IMHO. >tl;dr Just relax Anon, we'll think of something if we just put our heads together and keep.moving.forward. :^) >>17124 >spell check missed that one and somewhere I added an extra "n" Kek. I actually preferred that one to your intended one, Meta Ronin! :^)
>>17124 well no it doesnt matter, thats how formal logic works, its not intuitive but nothing is assumed other than the premises, supposing only these premises there is no way to get a machine that is aware because the only premise that allows you to do so is premise 2, it doesnt matter if its says all or some, there is no other premise to infer a machine is aware, if you just assume there exists one you must make it a premise or provide some other premise that allows you to infer it, thats why i tossed in organic and plastic in case someone could make a premise for cyborgs to aware and imitation to aware >>17127 this is a philosophy thread no
>>17128 >this is a philosophy thread no Sure, but this is a robowaifu board, no? Where is it written that all debate under the pretext of philosophy must always end in the 'stalemate' (fallacy) of ' >'"What is truth?"' >-t. once semi-important, now-dead guy, Pilate. infamous for condemning Jesus Christ, his own Creator We need to arrive at practical (if initially imperfect) solutions to our needs in the end Anon. Anything less is simply useless hand-waving. >=== -cleanup my crude language
Edited last time by Chobitsu on 08/03/2022 (Wed) 12:53:59.
>>17097 >A) all squares are rectangles >B) anything not square can't be a rectangle are not the same. logical implication is "unidirectional" so to speak x is a square -> x is a rectangle p = x is a square q = x is a rectangle https://en.wikipedia.org/wiki/Material_conditional check the truth table. if p is false, p -> q is true regardless of the value of q >>17122 I think you're mixing the terms here. "virtual" in philosophy does not mean the same as "virtual" in, say, engineering. philosophically speaking software is not virtual because it exists in the material world. from the engineering standpoint, you could say that it's virtual because some of it's physical characteristics (weight, size, etc) are negligible. when it comes to things like this, it would be better if you specified what definitions you're using
Open file (513.97 KB 1200x1500 FP2NwFdWQAU5e8P.jpg)
>>17223 i will write this here bcs it is more relevant to this thread >I was just expressing my concern about having more abstract discussions without any attempt to make any of that into some structured data, software or model my goal is actual synthetic consciousness (among other things). the problem is too profound for it to be able to be immediately translatable into software. to illustrate the issue: let us assume that dynamic field theory is the right way to understand consciousness. then what we have subscribed to is an analog model of cognition that is fundamentally different from digital ones. but this means that what we are concerned with is not a simple matter of software. maybe we should be looking into the mathematical modelling of dynamic systems, and the engineering task that we would have ahead of ourselves, but not software... furthermore, if such a thing is programmable, we might need a new sort of programming language specialized for dynamic systems... so that is already a major problem, but with the approach i am interested in, it might also require new physics as well. so that is an extra layer of problems i need to deal with before finally reaching something that could be analogous to software engineering the purpose of this thread was to collect philosophers that may be applicable to the project of making agi (preferably those that directly talk about ai related matters), and to provide summaries so people reading can decide whether it is worth looking further into them. the thread needs to be open-ended because of the difficulty of the problems pointed out and the mass wealth of subtlety entailed by such difficulty. i already had some vague idea of my approach, but i wanted a wider scope to survey in case there are key details that i was missing. i only found stephen robbins's work after a lot of exploration, and his criticisms of psychology, ai, and physics are earth shattering. the thought that there could be someone else like him lurking deep within the online oceans terrifies me progress has been, of course, depressing, as there is such a wide scope of material i have uncovered to check out and i struggle consuming it all at a consistent rate. the path from where i am starting to where i need to go is maddening, but if i give up now that is over a year down the drain...
>>17235 Just a suggestion, but it sounds like you need to join the Blue Brain project. They aim to make a digital reconstruction of the human brain, and have got quite a long way already. I am not sure starting from scratch to create a "synthetic conciousness" is the going to be possible. Sir Roger Penrose seems to think that conciousness is not computable...or if it is, we understand so little about how the brain works that programming a "synthetic" brain is many centuries away. Personally, I think growing something that is part-human, part-synthetic might be the best way to go. That way, we aren't re-inventing the wheel. We could create human clone bought to-term in a surrogate mother with present technology. We could also create synthetic wombs if the R & D involved weren't so taboo. Then you start genetically engineering these clones to improve them - giving them resistance to genetic diseases and decreased vulnerability to infections. Then you'd have the beginnings of a "post-human" partly-synthetic people. However, you'd also have truckloads of dead foetuses and babies. This represents a problem to Western scientific research (and yet we are fanatical about the right to abort our young...which makes very little sense). The Chinese tend to be far more level-headed about this kind of research and I suspect they are secretly years ahead of the West with regards to genetic engineering. Last I read they were already significantly ahead of us in stem-cell research.
>>17235 Good luck. I'll go with AST° or something like that, but will look into your progress from time to time. You should consider that you might find more help somewhere else, though. So maybe try to wire yourself into other places as well. ° https://www.frontiersin.org/articles/10.3389/frobt.2017.00060/full
>>17241 >This study was funded by the Princeton Neuroscience Institute Innovation Fund. imagine writing an 8000 word essay on qualia and not even realizing its the exact opposite of your entire premise, its embarrassing
>>17235 >my goal is actual synthetic consciousness Why do you consider yourself interested in the field and ignore the post >>16998 containing a pdf link to a paper with a rare example of an actually useful philosophical interpretation of consciousness in AI as it is currently researched by high-IQ people in a competitive field of ML? >>17239 With some sympathy, Blue Brain project has more or less failed, read its leader's (Henry Markram) writing over the years. Some variant of the original project's goal may be possible, but realistically we will have a general intelligence derived from current deep learning way before that expected point in time, and it will change the field by helping us in many ways we couldn't conceive before. Also, politically and philosophically (that is, for approx. 50% of with what could be called a "causal history self-concept" of identity), life extension and intelligence amplification via biological means (such as a hypothetical advanced gene therapy) are a proposition that's much easier to sell than uploads. >>17173 >>17128 Logic-driven AI is dead, at least Wittgenstein was more honest about his abject failure and switched philosophical approach mid-career. Just a reminder for fresh mind attracted to this field, don't waste your time. Better learn how tensor calculus works. As I have already said, if philosophers were to be of some help to our project, they would have to become data engineers and/or experiment designers. Simple as. Random meme robot picture from Stable Diffusion discord to attract attention and allude to a subtle point.
>>17251 how do you know he ignored it, he probably read it and had the same impression i did not everything alluding to philosophy IS philosophy, the pdf makes no arguments just vapid claims, its really an idiotic view of reality and is really just redefining intelligence to point of it being trivial, you could write a regex engine in just 10 lines that would fall under their clown definition of intelligent, and it would be the ultimate intelligence because '.*' is the ultimate regex compression of everything in the entire universe, so with >A useful theory is a compression of the data; compression is comprehension” (p. 77). The more compression is achieved, the greater the extent to which a system can be said to understand a set of data i already made an ai, and a super intelligent ai no less, lets play along further and see how much more absurd this drivel can go >Data compression occurs when information is bound together through the identification of (!)shared (!)patterns. For example the sequence 4, 6, 8, 12, 14, 18, 20, 24... can be simplified as the description “odd prime numbers +1”. ok so give me the sequential shared(!) pattern(!) of this series of events >bird flying >man walking >bird shits >shit falls on head of man walking >man feels wet on head >man looks up and sees bird >man understands bird has shat on his head what is the shared(!) pattern(!) in this sequence being """"'compressed"""" by the man that allows him to attain the understanding that a bird just shat on his head, if there is no shared(!) pattern(!) then there cannot possibly be any """"compression"""" of this series and thus no way that the man could possibly understand this series of events by their logic and therefore by contradiction we have proved |- if a bird shits on a mans head then that man does not understand that the bird has shat on his head i mean as idiotic as it is its not that bad for fucking cs students its just the wrong area of application, you could easily push this into aesthetic philosophy where shared patterns actually do exist frequently and are significant but claiming this is a solution to the problem of consciousness deserves a fucking slap to the face, a really fucking hard slap so that they know what the hard in hard problem means i dont know what your talking about with tensors, we were talking about the illogic of making claims that ai exists, in cs tensors are just a datatype (array[][][]) not a model of computation tensor 'calculus' isnt a thing , logic based i dont get, i think your talking about lambda calculus which is a real computation model but no one uses it because its a shit (but fun) way to program
>>17243 >>17243 There's no consens about the concept of qualia and if it even exists. Daniel Dennett: https://youtu.be/eSaEjLZIDqc Btw, we don't need AGI for robowaifus. Human level intelligence, specialized to certain areas of expertise, would be enough. Probably even less than that would be sufficient. We don't really have the problem of needing to implement some fancy definition of conscience: It's a a layer which gets the high level view of the world and gets to make the decisions. But it can't change the underlying system (for security reasons). And it might even not be allowed to look deeply into what's going on everywhere, it certainly doesn't get flooded with all the details all the time. Problem solved, I guess. I will refer to that part as conscience.
>>17255 no there is a strong consensus, its only one(1) group of people that are schizophrenic about it because they want to play both sides with bad faith in the same way the argument for retard trannies is made, theres no such thing as a man or woman except when a man says hes a woman because then the man is not a real man and is instead a real woman but there is no such thing as man or woman unless its a man saying hes a woman but there is no such thing as a man or a woman unless its a woman saying shes a man etc., this duplicity is the new vogue just read how deceptive the clown is writing >This subjective experience is often called consciousness its not, a subjective experience is called qualia, what idiocy is this, the idiot just wants an easy way to infer consciousness without actually saying qualia because he knows qualia is a deathblow when openly declared, qualia arguments like 'give her the d' are the easiest and soundest arguments to infer consciousness the clown is even using 'give her the d' but qualia is the achilles heel of materialism for the same reason, all the crippling arguments against materialism are based on qualia because qualia is incompatible only in materialism because its fucking irreducibly subjective, hello, its antithetical to their pretentious objectivism and adherence to truth being a physical absolute neurosoyentists are supposed to reject qualia and never speak of it which is fine, they have no choice thats how they get rid of the gap in knowledge, ie. there is no gap because that knowledge(subjective experience ie. qualia) doesnt exist its epiphenomenal but they cant then use it to make arguments for consciousness a man is not a woman you cant just redefine words, its true robots dont need consciousness and theres no point in a real ai but that doesnt mean the definition of consciousness changes or you get to call your junk pile conscious, old chink once said "when words lose their meaning people lose their freedom", prophetic words considering the state of the world
all these ideas of what consciousness is or isnt. I agree with anon that for our purposes it isn't necessarily necessary unless the end user has a deep need for it. as I stated I'm kind of a panpsychist in regard to this, however if we want to pin "consciousness" down to a phenomenon, my personal theory after hundreds of hours of study, reading, meditating on it, so on, is that it is a result of a fractally recursive phenomenon. IMO this recursion is inherent to the structure of our neurons (which operate on 11 "dimensions" of order) https://www.sciencealert.com/science-discovers-human-brain-works-up-to-11-dimensions Consider the idea that as your map becomes more and more akin to the territory, becoming larger, relief mountains built with real rock and trees, lakes filled with actual water, etc and then that map containing a fascimile of a person, looking at a facsimile of the map, which in turn contains an even low fidelity copy of the map again, and again IMO it is this uncanny process of reflection unto reflection (like the reflecting silver spheres arranged in a grid: Indra's web of Hindu Mythology) is where consciousness arises
>>17262 >inherent to the structure rather, is inherent *in* the structure there is no reason our neurons should have the monopoly on this they are not made of "magic" matter (pursuant to my counterargument of the above "only brains can have consciousness")
>>17262 is broccoli conscious then or something thats not even alive like a piece of bizmuth or a snowflake, recursion is the natural way things grow its not that special its just mathematically impressive, the only difference between a brain and coral or fungus or sponges or anything with a brain like structure is neurons, your right though theres nothing really special about the brain, its just a part of nature and will never be anything greater than nature, the mind isnt though, its only materialists that say brain=mind which forces them to either make the brain appear greater than what it is or debase the mind into something insultingly simple, i mean you could literally say materialism is a simple minded dogma i was only using neurosoyence as arguendo for claiming ai, keyword being artificial, i could just have a kid with someone or use whatever natural process possible to make consciousness manifest but then its not artificial is it, its just intelligence
>>17265 its not recursive, to be recursive is a process nice fractal though
>>17265 IMO "mind" is a process that isn't contained in the brain like gas in a box. It's something that operates on a nonphysical "dimension" and this is my own speculation (but others probably have and will touch on these points). Remember also we are "simulated" in each others minds, as like Indra's Web and how we're "simulated" in the minds of others factors into how we are treated by them, and becomes part of our feedback loop, so the whole process is continually recursive. The "fractal" part is harder to explain but when you think of "self similarity" and the relationship of maps to territories (and maps so accurate they contain themselves in the map, ad infinitum) then you're getting the idea.
>>17270 Greg Egan touches on this in his short story A Kidnapping (Axiomatic, 1995). Going to spoil it The main character is being blackmailed by a someone holding a "simulation" of his wife hostage and will torture her unless he pays ransom. Since his wife was never scanned he is puzzled how this could be, until he realizes somebody had hacked his own scan, and from his own simulation extrapolated enough of his wife to simulate her
>>17270 it just sounds like normal dualism now. as in 'material and immaterial are unrelated but interconnect', i dont know what you mean with simulation, dualism has way more logic involved than materialism, for simulating people you have to give me a theseus ship response because everyone uses different laws of identity, although i think youre really just talking about simple knowledge, dualists see knowledge and truths as objects that are attached to things so your knowledge about someone(obtained truth of x) is as much a part of you as it is of them(truth of x) showing ai in dualism is too hard if not impossible though, if you can do it then you have knowledge on a level that completely dwarfs ai in comparison, its in all sense of the word an otherworldly understanding >>17271 reminds me of a movie thats the same story, took me a while to find the name, The Thirteenth Floor (1999)
>>17096 >there is consciousness and we can construct it, this is a defacto materialist perspective ok >obviously constructing conscious is therefore equivalent to constructing the brai only if you are a mind-brain identity theorist. no one is a mind-brain identity theorist anymore (except someone like searle, who every ai-nerd hates). every materialist is a functionalist now, so they think the brain is a computer and consciousness=software >again im just using their own bullshit premises the above is their position, you are not actually using their premises >more formally; this wasn't a valid argument, try proving it with just those premises and you will see the conclusion won't follow. i don't think this is too big of a deal though because i understand what you were trying to say anyways though. just a heads up >>17122 >it cannot possibly have a will if it has been programed i would agree with you if by programmed you mean that all of its behaviours are preprogrammed. i would just like to point out that having any sort of programming at all does not preclude having a will. there is possible some argument humans have a few pre-programmed instincts. what separates humans from an animal just acting on instinct is that they are able to learn new things and slowly develop behaviours that aren't pre-programmed. whether or agi inspired by current understandings of intelligence can do this is another story. personally, i seriously doubt it can. i think this article presents a good argument against such a computationalist approach to consciousness: https://arxiv.org/abs/2106.15515 >>17239 >Sir Roger Penrose seems to think that conciousness is not computable... penrose thinks that wave function collapse is the key to collapse, and thus seems to believe quantum computing might work. i actually believe the problem he has brought up might be far too profound to be solved by a new paradigm of computing. i am even skeptical of the idea that hypercomputation would solve anything either. i think all of these approaches to computation probably aren't going to help. what i have in mind is far more radical >That way, we aren't re-inventing the wheel eh, you don't really have much control over how your waifu looks or its size this way (unless we develop really powerful new technology in synthetic biology, however i think by the time such technology arrives we would already be close to artificial life) >>17241 AST has already been brought up earlier in this thread. you can find some criticisms ive written of it there >>17251 >consciousness as data compression i've heard of this idea before. there is some truth to it. something that conscious beings do is simplify phenomena down to patterns. however, i do not believe that our current approaches to ai are actually as good at detecting novel patterns as we'd like to think. the article i posted above by kauffman details this issue >>17252 i think you are acting too violently upset about this anon. i definitely do think your criticism has some truth to it, and it actually touches on a problem chomsky has highlighted in large language models: they do not actually try constructing a real theory about how a language operates. their theory basically just is "anything goes". this article talks more about this: https://garymarcus.substack.com/p/noam-chomsky-and-gpt-3 >>17255 >Btw, we don't need AGI for robowaifus possibly. it is a personal preference of mine
>>17364 >a subjective experience is called qualia i dont think you understand what dennett is saying, however i can not fault you because most people don't. actually for a long time, i just thought he was just speaking nonsense till i heard an illusionist explain their position. needless to say, the whole "qualia doesn't exist" shtick is very misleading. what they mean is that they have problems with at least one of the following statements in this very specific list of ideas about it (https://en.wikipedia.org/wiki/Qualia#Definitions): 1. ineffable – they cannot be communicated, or apprehended by any means other than direct experience. 2. intrinsic – they are non-relational properties, which do not change depending on the experience's relation to other things. 3. private – all interpersonal comparisons of qualia are systematically impossible. 4. directly or immediately apprehensible by consciousness – to experience a quale is to know one experiences a quale, and to know all there is to know about that quale. note that this definition is not even accepted by all non-materialists. there are even plenty of idealists (such as hegel, and likely the british idealists) who do not accept (2) because they think all qualities are relational. for hegelians, qualia like redness are concrete universals... why is this important? because since people like dennett don't agree with this definition, they want to throw the whole concept away >'give her the d' this is a weird example. please explain what you meant >because that knowledge(subjective experience ie. qualia) doesnt exist there are some materialists who actually say that this knowledge does exist, but they bring up the ability hypothesis (https://plato.stanford.edu/entries/qualia-knowledge/#NoPropKnow1AbilHypo). i do think there is some truth to this idea, but i feel as though they lack the proper metaphysics to understand why >you cant just redefine words idk why you think this. it is a very common thing people do in conceptual analysis. if our concepts are bad, we probably need to define new words or use old words in new ways. an example of this is with continuity. mathematicians have defined the concept of continuity around open sets and limits. bergson meanwhile believed that such a definition is unable to properly capture temporal continuity. this lead him to give a new definition of it. in the process of doing this, he also ended up using the term "duration" in a different way as a contrasting term between his theory of time, and the physicist's understanding of time based on abstract mathematical continuity. this is fine as long as people are explicit about using their terms in a different way >>17270 the brain dimensions thing doesn't really imply a nonphysical "dimension", unless you are reifying the structure of the brain like an aristotilean >>17273 >so your knowledge about someone(obtained truth of x) is as much a part of you as it is of them(truth of x) i am guessing what you mean by this is an externalist theory of truth. i dont think this is a view that only dualists have though, nor is it a view that all dualists have i don't think externalism permits the idea that the material and immaterial are completely unrelated. from what you have said, it follows that the knowledge of a thing is as much a part of you as of the thing. but this implies that it is possible that the thing is contained within the mind. the conclusion of this would be something like idealism, not dualism... note that as i have pointed out earlier, idealism does not mean that there are no external objects btw
>>17365 >like an aristotilean a hylomorphist to be more precise. of course aristotle is more about there being an immaterial intellect and stuff
>>17080 okay. time to summarize chapter 2. honestly it is a much faster read than chapter 1, but it relies more on technical details that a a simple summary would betray. part of the reason it took me time was because of meditating on some of these details (as all of this is going to be basically the foundations of the rest of the book, it makes sense to spend extra time on it), but it is also because i have just been procrastinating/distracted from other things... anyways here is a basic summary of some key points to take from this chapter: 1) negarestani suggests a basic framework for approaching AGI which he calls the AS-AI-TP framework. keep in mind that his model of AI is capable of discursive interaction with other agents. this is stratified between different levels. >(i) the first level is just basic speech. such a thing is crucial since we need to interact with other agents somehow >(ii) the second level is dealing with the intersubjective aspect of speech involved in conversation. i personally suspect that grammar might emerge at this stage >(iii) the final level involves context-sensitive reasoning, and reaching higher levels of semantic complexity (i am guessing what he is hinting at here is functional integration) one thing i am unsure of is whether stage 2 and stage 3 can actually be thought as separate stages, because it seems like what we see in stage 3 could naturally emerge from stage 2. such an emergence clearly wouldn't happen with stage 2 from stage 1... the framework by its names also separates out three different projects that are important for these three stages: >(i) AS which corresponds to the construction of artificial speech synthesis. this one is special because it largely only concerns stage one >(ii) AI which corresponds to the project of artificial intelligence >(iii) TP which corresponds to the the project of finding the general conditions for the possibility of a general intelligence. negarestani of course sees kant's transcendental psychology as the beginning of such a project 2) he makes an extensive criticism of the methodological foundations of the disconnection thesis. this is basically this idea that future intelligence could diverge so far from our own that we might not even be able to recognize it as intelligent. among other problems he has with this view, i think the most important one to extract is that if such an entity has truly diverged so far from our own intelligence, it is a mystery why we should even consider it intelligent. because of this, negarestani wants to stress the importance of certain necessary functions being implemented by a system (what he calls functional mirroring) over the structural divergences that might occur... this functional mirroring partly arises when we have a determinate conception of how geist's self-transformations should take place generally 3) by bringing up functional mirroring, we bring up the question of what conditions are absolutely necessary for us to call something sapient. negarestani terms soft parochialism the mistake of reifying certain contingent features of an intelligence into necessary ones. the purpose of transcendental psychology is to purify our understanding of general intelligence so that we only include the features that we absolutely need 4) he writes a basic list of the transcendental structures. i already described them here: >>11465... negarestani also remarks that transcendental structures can also be used to articulate ways in which more complex faculties can be developed (i believe he gropes a little to how in his discussion on chu spaces) 5) based on all of this, negarestani also motivates that we construct a toy model of AGI which he frames as a proper outside view of ourselves. a toy model has a twofold utility. first, it provides something which is simple enough for tinkering. second, it makes explicit meta-theoretical assumptions. this latter point is important because sometimes we might be imposing certain subjective self-valuations of our experiences unto our attempts at describing that capacities we objectively have. the construction of a toy model helps avoid this problem 6) something negarestani furthermore calls for is a synchronization between the concepts that have been produced by transcendental psychologists like kant and hegel, and cognitive science. he notes that in this process of relating these concepts to cognitive science, some of them may turn out untenable. i think there is actually also a flip side to this. by seeing how cognitive science recapitulates ideas in german idealism, we are also able to locate regions where they may recapitulate the same errors as well
>>17439 7) negarestani outlines two languages for the formalization of his toy model. the first is chu spaces which are basically a language for concurrent computation. he links an interesting article ( https://www.newdualism.org/papers/V.Pratt/ratmech.pdf ) which relates concurrent computation to the mind-body problem. it does this by basically framing mind body dualism as a duality instead. the scheme is as follows: events correspond the activities of bodies, while states correspond to mental states. the function of events is to progress a system forward, while the function of a mental state is to keep tabs on previous events. the key idea for pratt is that the interaction between event and state is actually simpler than event-event or state-state interactions. 'a⫤x' basically means that event 'a' impressed on mental state 'x'. meanwhile, 'x⊨a' means that with the mental state 'x' the mind can infer 'a'... transitions from event to event or state to state are much more complicated. they require the rules of left and right residuation. the basic idea of these is that for us to transition from state x to state y, we need to make sure that from y we are able to infer all the same events that have occurred previously as state x. these residuation rules seem to be important for making sure that there is proper concurrency in the progression of states... negarestani also seems to be hinting at the idea that with the help of chu transforms we can see how chu spaces may accommodate additional chu spaces in order to model more complex forms of interaction... the benefits of chu spaces: (i) provides a framework that accommodates the kantian distinction between "sensings" (which he corresponds to causal relations) and "thinkings" (which he corresponds to norms) (ii) since state-event, event-event, and state-state interactions are all treated as different forms of computation, we are able to be more fine-grained in our analysis of the general form of thinking beyond just calling it something like "pattern-recognition" or "information processing" (iii) in doing what was just mentioned, we also avoid shallow functionalism (iv) allow us to model behaviours as concurrent interactions the second language he wants to use is that of virtual machine functionalism (you can read about it here: https://www.cs.bham.ac.uk/research/projects/cogaff/sloman-chrisley-jcs03.pdf ). the basic idea here is to talk introduce into our understanding of mind virtual machines. these basically correspond to levels of description that are beyond that of physics. VMF is distinguished from a view called atomic functionalism in that the latter just treats the mind as this simple input-output machine. meanwhile, in VMF, we can talk about various interlocking virtual machines that operate at different functional hierarchies... the differentiation between different scales and descriptive levels is really the main benefit of this approach. it allows us to avoid an ontology that is either purely top-down or bottom-up. i think this is actually a really important point. another important point here is that by looking at the interaction between different scales we are able to describe important processes such as the extraction of perceptual invariants... VMF seems immediately fruitful to some of the AGI modelling concerns... Chu spaces less so, though i have reasons to still return to this idea with a closer look
>>17440 >VMF seems immediately fruitful to some of the AGI modelling concerns... Chu spaces less so, though i have reasons to still return to this idea with a closer look Chu spaces also seem interesting. I hadn't heard of them before, but they seem like a good way to represent objects in spaces. For example, objects in some input image can be represented by a Chu space. So to look for an object is in some image, instead of calculating p(object | image) for each location, you would transform the image into K and check K(object, location). I guess the benefit is that everything about a Chu space can be represented by a single function, K, which gives you a unified way to handle a lot of different kinds of structures and the interactions between them. I think most observations can be treated as an object in a space (e.g., when you're observing the color of a shirt, you're looking at a shirt in a colorspace), so it seems very general. Chu spaces also seem closely related to tensors, so I would guess that there's an easy path to a lot of numerical machinery for anything written as a Chu space.
>>17442 >Chu spaces also seem closely related to tensors, so I would guess that there's an easy path to a lot of numerical machinery for anything written as a Chu space ah that might be true. that is a benefit of thinking about mathematical structures as computations/behaviours become easier to systematize >>17440 onto chapter 3 (i actually read both this and chapter 4 so will try to summarize both today). there are more philosophically tricky stuff here that i take issues with but not sure whether i should write about them or not. anyways: 1) the main goal of this chapter is to articulate the features that are needed for sentience (something shared by basically all animals), and then trying to articulate the contrast between this and sapience (more pertaining to rational capacities that seem to be unique to humans). negarestani articulates the features of sentience by thinking about a hypothetical sentient automata with the following features: >(i) self-maintenance goals (so like omohundro drives likely involved here) >(ii) the capacity to follow these goals. this can be thought of in terms of the multiple layers of interaction already discussed in chapter 2. he also talks about a "global workspace" which pertains to a property of the system where there is information in particular subsystems that is globally accessible to other subsystems. one way we can think about this is as a virtual machine as opposed to some physically composed central processing unit. what we really care about is the functional capacity for global access. this points to a subtle difference from the "global workspace" articulated in global workspace theory >(iii) there should be sensory integration of different sensory modalities. for instance, instead of just perceiving sight, also perceiving sight and sound integrated into a more unified experience. the utility of this is that it reduces the degree of ambiguity present in our perceptions >(iv) a (re)constructive memory which synthesizing external representations with an internal model of the world and also predicted internal model of what will happen in the future (based on the agent's actions) 2) the author goes into a rather subtle point about some of the conceptual difficulties with his approach. in trying to conceptualize this automata, we will try to give a story of what it sees and contrast it with a story of what the sapience sees. a major caveat here is that in articulating the story of the automata will still reference concepts that the sentience would not strictly recognize (e.g. concepts of redness... really what this sentience should see to negarestani would just consist in tropes). nearestani still thinks that there is a virtuous circularity here, in so fa as our concepts do make explicit certain structures in "unthinking" cognition (a claim that would be likely supported by cognitive linguists). as such we might not just be chasing around discursive illusion. he thinks there is a chain of as-ifs between sentience and sapience. when i first read this claim, i got a bit confused by this claim, as his whole functionalism is already taken as an as-if. i think the as-if structure here is more concerning conceptual incommensurability being partially elided for the sake of exposition... perhaps this caveat may not be needed to be made so explicit if he started with the an explanation that what we really mean are tropes (thus, for instance we shall talk about the trope 'schmed' as opposed to red) though this might be a little bit annoying to articulate if it is possible to articulate at all 3) the main reason why negarestani takes so much care to point out this caveat and explain why it is not so problematic is in order to respond to the "greedy skeptic". he has in mind here the philosopher bakker who is behind blind brain theory amongst. their main strategy is to try and claim that our understanding of intentionality is produced by a cognitive blindness/discursive illusion. a major problem with this approach is that it is so greedy that it ends up undermining its own foundations, for the blind brain theorist must accept that they themselves are blind. another issue negarestani has with the greedy skeptic is that it elides the distinction between the space of reasons and the space of causes 4) finally we get to see the two stories which negarestani talks about (pic rel). he thinks there is a distinction between the two sorts of seeing here. the first story talks about seeing 1, and the second story talks about seeing 2. seeing 1 is more concerned with raw sensations, while seeing 2 is conceptually mediated. now, i personally disagree that this distinction is one that actually separates sentience from sapience. just from the basic features, especially feature 4, there seems to be some conceptual mediation entailed there. i think ecological psychology would likely disagree with the idea that sentiences don't perceive concepts as well... it doesn't make for the best theory of perception. negarestani later on even caves a bit and accepts the fact that our spatial perspectives are already an example of seeing 2... for more on this debate, check out the article 'Articulating Animals: Animals and Implicit Inferences in Bramdom’s Work’. i think the metaphysics of tropes might be related to what is at stake as well... a lot of interesting stuff to think about
>>17520 note: negarestani also says that seeing 1 is seeing of, while seeing 2 as it seeing as. anyways... 5) now, we start think about the chain of as-ifs. negarestani claims not only that seeing 1 isn't conceptual, but furthermore that we can't actually see individuated objects with seeing 1 either (so for instance, we don't see cups but instead some manifold of sensations in a cup-shaped spatiotemporal boundary)... needless to say i disagree that sentiences can't perceive objects as well... anyways something that he likes about kant is that he articulates perception as involving multiple layers. thus perception is about the integration of a collection of algorithms. kant talks about three syntheses that would be all requisite for artificial general intelligence: >(i) this systematizes sense impressions so that they have a spatial and temporal location. i think what may be important here might be the idea of simultaneity. like, whenever we perceive something, it appears as though all of the sensations we see are all located within a single moment >(ii) the synthesis of reproduction relates appearances temporally, which constructs stable representations of items. note here that items are different than objects. items are what negarestani wants to understand as things that we see in seeing 1. i think this idea is somewhat strange however, since if we start talking about systematically interrelated/stable items, they are not just disconnected tropes (but perhaps something more sophisticated). perhaps he is fine with this as long as he connection here is synthetic/normative. i need to study closer sellars's theory of tropes here... >(iii) usage of concepts to synthesize representations (seeing of) into objects note here that (i) and (ii) are together the "figurative" part of our perception. they provide us with some systematicity in our raw materials that will be brought together to form the perception of objects in perception 6) negarestani furthermore thinks that the mesoscopic level of analysis (not just top-down or bottom-up) already talked about with chu spaces and virtual machine functionalism can be used to understand these syntheses. this idea is also connected to putnam's liberal functionalism which he claims entails a picture of the organism as systems only in so far as they are engaged in interactions with the environment. i may say further that there are different possibly nested systems of interaction 7) he tries connecting predictive processing to the figurative synthesis mentioned above as well... an important parr of this lies in priors which are probabilistic constraints that permit a system to hypotheses. not only are there normal priors, but also hyperpriors which are priors upon priors. these can be used to help differentiate between different levels of hypothesis. there is an analogy made here between priors and kant's forms of intuition. more can be read about this in the article titled 'The Predictive Processing Paradigm Has Roots in Kant'. something else about predicting processing is that incoming inputs are always related to a pre-existing representational repertoire. we can notice a parallel with negarestani's earlier talk about a (re)constructive memory... there are also a lot of articles given on how to formalize this stuff that might be interesting further reading: >'Evolutive Systems: Hierarchy, Emergence, Cognition' >'A New Foundation for Representation in Cognitive and Brain Science' >'Colimits in Memory: Category Theory and Neural Systems' 8) negaresani goes on to talk about a particular way of formalizing how predictive models are applied. this is by means of colimits which is rather technical. basically what is going on here is that you can use a colimit diagram to model the integration of neuron clusters (these clusters are interpreted as categories where i am guessing for instance that the objects are neurons and the morphisms are synaptic connections)... we can also define aa macroscopic category Ment which can be defined as the colimit of neuron clusters which exist at the highest level of integration 9) there is however pointed out two methods of this colimit methodology: >(i) it is just focused on the construction of higher levels out of modules. it can't deal with synaptic pruning which negarestani thinks is important for unlearning >(ii) category theory relies on commutativity and works best with symmetry and synchronous processing. these are constraints that not all physical processes or even neural tasks conform to negarestani thinks limits are important things to consider so we know when it is appropriate to apply a model. the predictive processing paradigm is also limited to him as it is unable to ground the plurality of methods and the semantic dimension behind the construction of scientific theories. this dimension is not merely grounded in our basic intuitions for otherwise we would be unable to break out of them. furthermore, hyperpriors need not be assumed to be properties of objective reality for supposing such a thing would require us to explain why modern physics often goes against our intuitions
>>17521 now we enter territory that i already foreshadowed earlier: 10) even this merely sentient awareness needs to have a 'perspectival stance' in so far as it must differentially respond to an environment in a way that differentiates it from prey (otherwise it might consume itself or something strange like that). the range of spatial relations the predator can maintain, for negarestani, is limited. it considers spatial relationships between itself and its goal ('endocentric' view), and to a more limited extent the relation between items in the environment ('exocentric' view). as said before, this endocentric view is acknowledged by negarestani to already involve a spatial sort of seeing as... needless to say i find this development rather upsetting, as negarestani seems to be entering into a sort of classificatory contradiction... the author also mentions that the automaton should furthermore (on top of/connected to this spacial perspective) have a naive physics of space. he references here claude vandeloise's work 'spatial prepositions' that contains work on this. ultimately this naive physics serves as a cognitive scaffold for more advanced spatialized concepts (again we can see some possible connections to cognitive linguistics) 11) there is also a temoral perspective which relies on the following successive progression of capacities: >(i) the capcity to synthesize sensations together into one simultaneous moment >(ii) bringing these moments of simultaneity together into a successive sequence of states >(iii) the ability to have awareness of these successions of states as successions (more precisely, the capacity to make them explicit) 12) to articulate time awareness we must start start with a basic distinction between two ways of intuiting items: >(i) impression (awareness of item that is copresent with the system... more related to something we just see) >(ii) reproduction (reproducing an impression in absence of the items... more related to (re)constructive memory) 13) the approach above is not sufficient for time awareness for we must functionally differentiate past stats from present states (just as we have differentiate the automata from the environment). to do this we must have meta-representation capacities (i.e. the capacity to represent a representation). it should be noted however that to negarestani this time awareness is not yet time consciousness as the automaton is not yet aware of the succession as a succession, and it thus doesn't have a self that can be furthermore "mobilized". this idea seems to connect to the idea of self-consciousness articulated in the phenomenology of spirit (which i also think is really concerned with "self-governance"). the stuff about meta-awareness is sort of weird to think about because negarestani thinks that meta level also has impressions versus reproductions. thus what i think is going on here is a genuine meta-system transition 14) this stuff about meta-awareness is rather technical, and he makes comments about category theory and a possibly-infinite hierarchy of meta-awarenesses. what is really important is that the meta-awareness helps us functionally differentiate between impressions and reproductions, and thus the system can differentiate between 'is' and 'was'. something negarestani also introduces is the element of anticipation. if there is a functional differentiation between future predictions and the other two types of intuitions, we then have 'later'. if impression, reproduction, and anticipation are labelled be1, be2, and be3 respectively we can draw a larger diagram of the variety of first order meta-awarenesses (second pic rel) 15) negarestani claims that with perspectival intelligence we now have the necessary tools to start formulating an intelligence capable of creating new abilities out of old ones with the help of new encounters with the world by means of reasoning and 'systematically' respond to impressions. i think this is still a bit of a strange distinction between sentience and sapience, but what he means by systematicity here is more specific. in particular, he thinks it involves two aspects: >(i) ability to make truth-apt judgements about contents of experience >(ii) choose one story of the world as opposed to another it may be debatable whether or not sentiences can actually be capable of such things or not... though what is an important piece in the puzzle is the element of socially mediated rationality. he also points out how time is an important structure to understand due to its place in transcendental psychology and transcendental logic... to sum up... i have plenty of problems here which are symptoms of a larger pathology of inferentialism in so far as it often differentiates sentience from sapience in a rather unnatural manner
>>17522 next we talk about chapter 4. i think this one is a bit weird to summarize because he goes into a lot of details into examples and details which are sort of unnecessary for a concise summary of the text. i also feel as though this chapter's direction wasn't the best... he spends a lot of time arguing against time asymmetry + flow of time and how it is just a byproduct of our subjective biases, but then he goes on to talk about hegel's absolute knowing as taking a sort of atemporal vantage point. i don't think the connection between these two ideas are very deep. there is also stuff about metaphysics (which he basically ends up equating with hegel's dialectical logic) which could easily bog the reader down. the basic message of this chapter is just to give an example of a basic feature of negarestani's project put to practice. that feature is the fact that we need to progressively refine our understanding of intelligence and strip off all these contingent categories... near the end of the chapter he articulates three interconnected projects that sprout from what had been discussed: >(1) how to conceptualize entities that might have different models of time than our own >(2) is there a way to possibly model our own temporal perspective (like with a physical model) without trying to ascribe features such as the passage of time to all of reality? >(3) what could it mean both practically and theoretically for us to think about our own cognition as possessing a time consciousness that is not necessarily directional or dynamic? what could be gained as well from considering alternative time models or even an atemporal one the last point is what negarestani sees as related to taking a "view from nowhere". he states that in later chapters, the time-generality he is trying to articulate will correspond to plato's ideas (Knowledge, Truth, Beauty, and the Good). he wants to enter into a viewpoint that is time-general step by step. i personally believe negarestani's understanding of Eternity might be a little bit flawed, because i do not believe it is related to time-generality per se. with that said, i do believe that the passages articulating his vision are very poetic/mystical and i urge people to read it for themselves (it is on pages 246-248)! there is an almost meditative practice involved here were we slowly enter into a timeless form of meta-consciousness. i am starting to understand why he considers himself a sort of platonist. maybe he is more closely allied with a more phenomenological understanding of plato, though by turning to functional analysis, we seem to have more structural generality accessible to us
>>17523 weird network traffic lately... anyways it's chapter 5 lol 1) as said before, negarestani thinks that we are unable to make veridical statements and judgements of one story as opposed to another yet. we also haven't been able to achieve a proper time consciousness (awareness are not distinguished as being particularly its own,this being associated with having a basic self-conception which serves as a sort of tool for modelling)... negarestani's basic solution to do comes in two steps: >(1) embed our automata into a larger multi-agent system >(2) give it medium of communication (in the text, negarestani uses speech as an example) he further details why this solution is appropriate: really what we want is to instantiate an apperceptive 'I'. this is an 'I' that not only gets attached to all representations (I think X, I think Y, I think Z) but also as a stable self-identity over time (I think [X+Y+Z])... to ground this self, we really need to establish a self-relation, but this takes us back to hegel's account of self-consciousness as interlocked with a larger multi-agent system 2) our starting point in this multi-agent system is a CHILD (Concept Having Intelligence of Low Degree). this is a system mainly of habit and regularities. at the same time, thanks to the systematic (not per se veridical) manner it interacts with the environment, this intelligence features a system of transitions and obstructions to transitions from one awareness to another. this system is analogically posited as being primordial material inferences which have various properties. an important one is context sensitivity which have various properties. an important on is context sensitivity which is described using the language of linear logic. this logic views formulas as resources that may only be consumed once the bit about all of this being analogically posited is important as negarestani does not think this intelligence is actually using modal vocabularies. as such it isn't able to actually conceive of causality either... what negarestani does not really do is tell us what is a salient distinction between observation of behavioural vocabularies and modal vocabularies. to understand, we really need to dig into what rosenberg has to say 3) in rosenberg's 'the thinking self', the CHILD is at the start "idealist" in the sense that it just treats its meta-awarenesses as identical with its world picture. (this is something that negarestani points out, but only treats as a separate issue from the problem of modal vocabularies in his exposition). this is problematic because for rosenberg, the distinction between mere regularities and actual laws comes down to the difference between appearance and reality. in order to do this we need to somehow decouple our world picture from our meta-awareness which will ultimately involve the constitution of an aperspectival world picture 4) as an aside, when we read rosenberg, he seems to hint at an influence from heidegger which explains his own stand-offish relation to the idea that sentiences have conceptual capacities. in heidegger's system, the more primordial form of consciousness is readiness-to-hand which rosenberg understands as involving proto-intentions that are only understood relative to our understanding of what an animal is currently "up to"... i am not personally so sure that readiness-to-hand is actually empty of concepts... rather it is more just not involved in a certain epistemic 5) back to negarestani... for the CHILD, to exit this naive idealist position, it must be able to have an awareness of the form 'I think A' and not just an awareness of the form 'A'. this requires the unity of apperception. apparently there was a part in chapter 3 which i think skipped over that was actually important. this mentioned the fact that meta-awarenesses are really a "web" of equivalence relationships over awarenesses that have occured over time. the 'I' in 'I think X' is formally identical to the 'I' in 'I think [X+Y+Z]'. again, how this gets established depends upon self-relation 6) an important starting point in negarestani's starting point is that the parental sounds are in continuity with causal regularities perceived by the automata to be interesting. moreover, in this "space of shared recognition" between the CHILD and its parents there is also involved the CHILD's awarenesses, meta-awarenesses and transitions between awarenesses modelled by the parents (i.e. the automata's behaviours are being modelled by its parents) 7) negarestani goes on to articulate a series of decoherences and recoherences. what preconfigures this series is that the reports from the parents are recognized by the automata and habitually recognized and mapped unto its meta-awarenesses. these reports are consequently taken as meta-awarenesses which are labelled by their source. labelled because these meta-awarenesses are received from its parents as a contrastable reports >first decoherence: what we often end up with are families of incompatible perspectives. what is noteworthy is how certain reports don't seem compatible with what the CHILD is immediately seeing. negarestani notes that this results in a world that is "proto-inferentially multi-perspectival" >first recoherence: the parental reports are now just taken as seemings relative to each parent, while what the automata sees are taken as objective. i personally think that this idea of parental reports being treated as mere seemings is an abstraction that can't actually have any utility. "seeming" means nothing without any potential for objectivity >second decoherence: the problem now is that the parental reports should be taken as being able to be objective, even if contrastive with the automata's own awareness >second recoherence: our CHILD must construct a world picture of different perspectives now taken as partial world pictures i think that with the first and second recoherences, there is involved the usage of the contiguity with certain agents as a functional salient decision mechanism
>>17567 8) what we have done in converting reports into partial world pictures is that we have now treated them as "logical forms" which can structure the CHILD's representations. these are used by the objective unity of apperception to synthesize objects which are equivalent to other objects in so far as they conform to the same rule. objectivity involves this and also veridicality. i suspect that this is what is behind negarestani's mysterious claim that mere sentiences do not perceive objects. to him, an the idea of an object seems to be linked to veridicality. thus if sentiences do not differentiate between seeming and being, they can not perceive objects 9) negarestani thinks that by the conversion to logical form and the dissociation of language from world, we are capable of conceiving of new world pictures. i am guessing this point is an elaboration on the solution he has to the problem he had with predictive processing proponents trying to use their framework to explain everything including scientific frameworks 10) the process above and education in general has as its pre-condition "practical autonomy" which is most generally characterized as the yearning of a CHILD to become a full on sapient agent, and the tendency to make adults recognize such a yearning. so far we have had a multi-agent picture where each agent models its world pictures as being composed of the partial world pictures of other agents (and furthermore recognized as belonging to those other agents) and furthermore excluding mere seemings recognized as belonging to other agents. in this process we have also recognized other agents as subjects. what we have elaborated in all of this is a space of recognitions. within this space we can individuate the formal self as that which owns certain apprehendings and not others which can be owned by other selves 11) our movement from the space of recognitions to the formal self has as a condition here that the CHILD recognizes adults as subjects which are important for its self-actualization. this recognition is ultimately a manifestation of the CHILD's practical autonomy (more precisely this seems to be the process by whereby it is in actuality) 12) autonomy arises from a self-differentiation of the subject. this process permits it to recognize a world. furthermore, it opens up the automata to disintegration and reintegration (what we have had looked at before were examples of this). through the further recognition of the world afforded by this formal autonomy, the consciousness drifts which eventually permits it to consider things which are impersonal. this is how self-consciousness (think of the self-relation we have discussed before) comes into the picture. it is only though all of this that the CHILD becomes a "thinking will". negarestani's conception of thinking seems to be related to the process of recognizing what one is not and acting upon this recognition (how precisely the latter happens i am not completely sure of) 13) education involves the cultivation of autonomy: >(1) increasing range of normative capacities >(2) removing artificial limitations placed on the child about what it should become. i think this is more generally the process of learning and unlearning ought-to-dos (a process already mentioned by negarestani) 14) education also has various structural requirements: >(1) vertical and horizontal modularity (formal hierarchical, latter is flat). these two kinds have their advantages and trade offs >(2) should be able to involve the integration of various learning processes >(3) exploiting inductive biases (think prior knowledge for instance here) to help fashion more complex abilities 15) education is generally a process that leads to the expansion of an agent's range of practical capacities which are intelligible to it. this end product where boundaries on our abilities are pushed back is what negarestani calls "practical freedom" 16) in talking about education, negarestani also talks about two classes of capacities: sf and sm abilities. the former are formal while the latter involve higher levels of cognition. sf is syntactic while sm is semantic. he actually provides a whole catalogue of sf and sm abilities that i do not feel like summarizing so i will just provide a pdf excerpt... in this cataloguing, we see various structural hierarchies of abilities. for negarestani, what these structural hierarchies also indicate is the fact that different strategies need to be used in pedagogy based on the child's current experience and conceptual repertoire (e.g. breaking down a task into simpler ones, combining simpler tasks to make a more complex one) 17) the development of the automata from being a child to one that has a capacity to rethink its place in the world (i presume that this is related to revising one's oughts ad ought-nots) ultimately requires a back and forth with adults
>>17568 chapter 6 now! 1) in the previous chapter we already presupposed that our automata had language. now it is time to explain how language emerges since it is an important pre-condition for general intelligence. we will reenact the development of language because it illustrates how language is an important ingredient in the development of more complex behaviours. the compleixification of language and general intelligence come hand in hand 2) we can understand our main goal as really involving modulation of the collection of variables agents in our system make use of in interacting with the environment. if this is the case, then really out agents are interlocked with the environment as well 3) for the sake of our reenactment, we will assume now that all the automatas in our system are now childs. an important distinction needs to be made between pictures (sign-design) and conceptual objects (symbol-design). the former are representations of regularities in the environment (this forms a second order isomorphic a on the first level we have signs that have some sort of resemblance in causal structure, and then we associate these signs together), while the latter are able to invoke combinatorial relations between symbols while pictures are basically simple representations, symbols are primarily best understood by how they combine with other symbols (they do still have reference but it is rather secondary). they do not simply represent external objects but also each other. negarestani really stresses that while symbols are dependent on signs, they are not reducible to them. signs belong to the real order (which includes causal regularities and wiring diagrams... largely causal) while symbols belong to the logical order that is autonomous from the real one (this negarestani associates, following sellars, with thinking and intentionality) 4) negarestani goes on to elaborate on the basic pre-condition for a sign... in short we need the following: >(1) causal regularities need to be salient enough to catch the attention of an automata so it may produce a sign of it >(2) we need enough complexity in our automata so it is able to recognize and make signs of its regularity from there we can sketch a basic example of a sign. let us suppose there are events Ei and Ej. their co-occurence could be denoted by Ei-Ej. we would want to register this relationship in our wiring diagram by some isomorphic structure, for instance Ei*-Ej*. what we have here is an "icon". this is a sign that associate with their referent by resemblance. really what matters to negarestani hee is that there is some stimuli discrimination going on let us say when Ei*-Ej* occurs, the automata make a sound. if this sound is heard enough times by other automata, they can start to reproduce Ei*-Ej*. we we have here now is "just" an indexical sign, as it merely connects two events by way of statistical regularity. notice how negarestani is constructing a sort of hierarchy of signs (i.e. symbol > index > icon) where the higher rungs is in to some extent built on top of the lower ones this process where we transmit and recieve indexical signs will be called communication (note negarestani does not simply think language is about communication but also interaction) 5) negarestani goes on to criticize the idea that we could just stick to picturing and never move to symbols... his first objection is that for every collection of signs, we would need signs representing their relationships. this would lead to a regress. this suggests we can't have an exhaustive picturing of the world. his second objection is that even if we could product a complete picture of the world, the regress would produce exploding computational costs 6) symbols, unlike signs, stand in one-to-many and many-to-many relations. negarestani seems to imply that somehow this provides us a real solution to the regress problem. another benefit of symbols is that they let us make explicit and develop the set of recognized relationships between patterns. there are also more computational reasons for introducing symbols. if we think about induction, there is of course solomonoff's universal induction which lets us make the most parsimonious hypotheses. the problem is that this requires an infinite time limit. as such, compression is not enough. we need to be selective about what regularities we single out, and after that to explore the relationships between these regularities. i would like to point out that ben goertzel (one of the most well known agi researchers) also considers this problem and has a similar solution in his own system (see for instance here: https://yewtu.be/watch?v=vA0Et1Mvrbk) ultimately, to achieve this (and more generally in order to achieve general intelligence) we need automata capable of material inferences that may be ultimately made explicit in formal inferences. what is emphasized here is the important of know how 7) negarestani goes on to outline the main pre-conditions for symbols: >(1) discrete signs (as opposed to continuous) >(2) combinatorial structure that signs can be put into 8) discreteness is important because without it, we have symbols with fuzzy boundaries (not sure important this is tbh) and which are also difficult to combine together. negarestani thinks out automata can invent discrete signs by means of a self-organizing process in which high dimensional data is projected into a dsicretized and/or lower dimensional data so that we converge to the use of discretized signs (he references ai for this) this discretization permits us to combine our phonemes together to produce more units of meaning. moreover, we have ultimately permitted symbols to be able to be enter into manipulable combinatorial relations whether more complex syntactic structures (and thus also encoding relationships between regularities) can be generated
>>17593 9) changes in the structures in language come hand in hand with the development of the automata's ability to model and communicate more complex structures. ultimate, negarestani thinks this exemplifies how language is the dasein of geist (~ sort of medium that sustains its actualization?). there are two general ways in which phonemes may combine: >(1) iteration: elements can be repeated as much as one wants (an example negarestani gives here is "chop garlic into paste" where the chopping operation is something that can be done as much as one likes basically) >(2) recursion: elements depend upon the occurence of past elements (e.g. "cut the pie into 8 pieces", we should only cut in half 3 times as each step depends on the previous ones) negarestani will represent iteration using simple concatenation of tokens (e.g. a, ab, abc, etc given the alphabet {a,b,c}). recursion meanwhile can be written using square brackets to indicate embeddings/dependency (e.g. [a]. [a[b]], [a[b[c]]], etc) first pic-rel shows an example of this scheme. iteration and recursion forms a context-free grammar. in it, thematic roles are determined by means of order + dependency which can be used to disambiguate things. even with this, i do not think context-free grammar on its own has enough structure to talk about semantic roles... perhaps these roles are provided in the context of material inferences. i did a little bit of research on how semantic roles are indicated in generative grammar and i found two strategies: >(1) semantic parsing: https://www.cs.upc.edu/~ageno/anlp/semanticParsing.pdf >(2) context-sensitive grammar: https://www.academia.edu/15518963/Sentence_Representation_in_Context_Sensitive_Grammars 10) milikan (a "right-wing sellarsian") thinks that we are not assuming that the structure of the regularities in the real world are already given to minds through encoding in wiring diagrams. in constrast, negarestani thinks this is incorrect and furthermore a recapitulation of the myth of the given. in contrast, he thinks that it is onlt with symbols that we can actually seriously think about the structure of the world. we can understand the difference between signs and symbols by comparing it to the difference between game and metagame. the game itself consists of pieces placed in different relations together (our syntactically structured reports on regularities) while the metagame articulates rules for playing the game (corresponding to material inferences). in chess the game reports pieces and positions on the board. the meta-game talks about rules on how to set up and play the pieces 11) with symbols our automata has access to "symbolic vocabularies". from what i can gather, these are vocabularies that conform to a particular syntacical structure recognizable to the automata. negarestani models these structures a finite state machines 12) negarestani points out that the thing about what we have so far with syntax is that it is not yet enough for full-fledged language competence as we do not yet have practical mastery over inferential roles. the activity of thinking, to negarestani, requires this. what we need ultimately is linguistic interaction through which automata can master linguistic practices and slowly generate more semantically complex capacities. negarestani talks about how the increase in the complexity of thinking requires the complexification of concepts although negarestani does not think semantics reduces down to syntax in a simple manner, he does think that it is reducible in the right circumstance (viz. in the context of interaction). moreover within this interactionist context, we shall see the increased development of semantic complexity though the medium of syntax. ultimately negarestani thinks that language, as a syntactic structuring apparatus, allows us to capture more rich semantic relations and talk about new worlds and regions of reality
>>17597 the end of our story is in ch. 7! 1) as negarestani said before, semantics reduces to syntax under the right conditions. he warns us that if we make semantics completely irreducible, we would get an inflated understanding of meaning. he criticizes searle's chinese room thought experiment for assuming this. he does not think the actual computation is occurring in the room but rather in the interaction between the chinese room's operator and the person outside the room. he thinks the right conditions under which semantics reduces to syntax is under the inferentialist theory of meaning. he thinks that the most basic manifestation of meaning is in the justified use of concepts in social practices... in particular there is required know-how regarding inferences an example that negarestani gives is that the believe "this is red" permits the belief "this is coloured" but does not allow "this is green". while this is a good example, i think it misses some important details concerning concepts qua functional classification (sellars's view of concepts). in that, there seems to be other details than the epistemic transitions and obstructions between beliefs. there is also schematic knowledge (e.g. a triangle is a three sides shape) that has more practical ramifications (e.g. construction of said shape). anyways, what negarestani thinks is involved here is speech acts being commitments that have implication for other commitments. what is needed then is a continually updating context (i think this idea may ultimately connect to dynamic semantics and even a bit to milikan's concept of pushmi-pullyu) what this view also gives us is an understanding of reason (as a process of making explicit and building upon concepts) to be an activity. negarestani thinks this allows us to see it as a something algorithmic and realizable through information processing. of course implicit here is the idea that computation is synonymous with 'doing' which might no be true. the interactionist understanding of meaning involves meanings only able to be determined within a game of asking for justification and giving justification. in this process, one does not need to know all the rules beforehand. rather rules emerge over time 2) something negarestani thinks is problematic about the interactionist approach is that ti does not elaborate much on why this interaction to be social or even what such a thing formally entails. to me his approach is honestly so formal that the social seems almost unnecessary. at the same time it might still have its use ultimately he thinks interaction can be best formally elaborated in the logical framework of interaction in ludics started by jean-yves girard. what is this interactionist framework? negarestani starts by elaborating on what classical computation involves, viz. "deduction" from some initial conditions. he points out a major problem with this approach because if we have knowledge of something and also know from that something else, we should also know that something else. as such we should know everything knowable in the system, and thus no new information is gained. negarestani thinks this problem is symptomatic of ignoring the role the environment plays in letting us understand computation as involving an increase of information. the environment is also in general necessary for making sense of machines as involving input and output one benefit of interaction games is that they do not need values determining how they should evolve. rules rather emerge from within interaction itself. moreover, unlike game-theoretic games, interaction games do not need to have payoff functions or winning strategies which are predetermined in the context of interaction, when we see the input and output of the system we see the following: on the input side the system consumes the resources that the environment produces, while on the output side the converse happens. in this framework, something is computable if the system "wins" against the environment (i.e. it can formulate and execute a strategy o performing a computational task). the interaction with the environmental constrains what actions the system performs. there are many variables that can be involved in determining how the game evolves (e.g. whether or not past interactions are preserved and accessible, whether the interaction is synchronous or asynchronous, etc). in the case of classical computation, computability is within the context of a two-step game where there is only input and output 3) i did some extra reading because some of the formalizations were hard to understand exactly why they were so important from just what negarestani had said. in 'Concurrent Structures in Game Semantics' i think castellan gives us a rather concrete idea of how game semantics works. castellan first gives us a basic example in operational semantics where we are to compute the expression '3 + (3+5)'. this can be done in successive steps: 3+(3+5) -> 3+8 -> 11 while this picture is nice, we run into a problem when characterizing expressions with variables, for instance 'x+2'. the solution is to transform 'x' into a request to the environment denoted by q^{+}_{x}. after that we may receive a value from the environment. if for instance we received 2, this may be denoted by 2^{-}. thus we have x+2 -> [] + 2 -> 2+2 -> 4 as an example. the basic idea is to denote sent messages by (+) and denote sent messages by (+) and received messages by (-). this gives us an alternation between sending and receiving data which looks like a game. in this case, we can describe our dialogue as q^{-}.q^{+}_{x}.2^{-}.4^{+} and generally characterize 'x+2' as: >[x+2] = {q^{-}.q^{+}_{x}.n^{-}.(n+2)^{+} | n <- N} so we see we can now characterize expressions by their general dialogue. this is an important point to keep in mind something negarestani only mentions in passing is how we can build richer computations by the addition or removal of constraints. one such constraint negarestani gives as able to be sublates is the markovian nature of computations. i am sort of disappointed this is
>>17653 4) negarestani goes on to elaborate upon the copycat strategy. the basic idea is to consider an agent that plays against 2 players. when it receives a move from one player, it plays it against another. in the literature, this is a sort of identity map which preserves the inputs and outputs. i think what negarestani finds important here is that it treats any process as though it were in an interaction game with its dual. this demonstrates how games are a generalization of any model of synchronous procedure-following computation 5) negarestani notes that in proof theory, we can understand the meaning of a proposition as the collection of proofs that verify it. i am again going to mention an article written by someone else on this topic. in 'On the Meaning of Logical Rules' girard discusses this topic. what the paper argues is that we really want to turn the question from a proposition's meaning to that of delimiting which collections of proofs are conjunctively sufficient for a statement and testing each of these proofs. for instance, to understand the meaning of 'A∧B' we need to know what proves it. in this case, it is sufficient to have proofs of A and B separately. we just need to then test these proofs something interesting here can be seen if we interpret X^{t} as 'test X'. then we can see that (A∧B)^{t} = A^{t}∨B^{t}, (A∨B)^{t} = A^{t}∧B^{t}, etc. in general, testing sort of works like negation. this takes us to the concept of "paraproofs". we can also interpret this as an interaction between the person asserting A∧B and another person challenging different subformalae of the expression... we can see here then the resonance with game semantics. i think that in some ways, ludics really radicalizes these ideas by giving ludics an address (some sequence of integers) and look at how addresses are inscribed in the course of a proof a quick remark about this article as well would be that really it looks like girard only really cares about the semantics of rules. he does not say much about the semantics of referential terms. this disambiguates negarestani's claims about the reducibility of semantics to syntax under the right conditions, in particular what region of semantics is reducible 6) there are two kinds of computation going on here in meaning-as-proof >(1) proof search: self-explanatory. we move upwards as we search for the proofs sufficient for a proposition. this is more or less what the dialogue consists in >(2) proof normalization: remove useless elements of the proof to extract some invariant. two proofs are equivalent if they have the same normalization negarestani then goes on to talk about a meaning dispenser that takes semantics (normalized proofs) out 7) negarestani talks about more things for transitioning from formal syntax to "concrete syntax". from what i understand, the concrete syntax here is what we see in natural language sentences which involve syntactic elements that depend on previously stated elements, for instance the pronoun "it" may only refer to a single noun once or a few times. we can understand all of this in terms of resource-sensitivity
>>17654 8) i did more extra reading to both give some basic idea of what ludics involves, and furthermore some initial idea of how it is practically applied. the first article i looked at was the article titled 'dialogue and interaction: the ludics view' by lecomte and quatrini. the basic idea here is that we can now receive and send topics as data, while previously we were using game semantics to specify requests for variable values to an environment, we now have this machinery also used for dealing with topicalization as well for instance, when the context we are in is a discussion about holidays, we can start with '⊢ξ' indicating that we have received that context. from there we may try specifying the main focus of the topic to be regarding out holiday in the alps specifically. this can be denoted 'ξ*1⊢' showing that we are sending that intent to move to a subtopic let's say someone wants to talk about when one's holiday was instead of where. this requires us to change our starting point to involve a combined context of both holiday descriptions (ξ) and date (ρ). we may denote them as subaddresses in a larger context e.g. as τ*0*0 and τ*0*1. from there we can receive this larger context (⊢τ), acknowledge that they can answer some questions and request such questions (τ.0⊢) and finally survey the set of questions that can be asked (⊢τ*0*0, τ*0*1). finally we can answer a question on, for instance, the dates ( ⊢τ*0*1*6⊢ τ*0*0) generally how i read these proof trees is first of all bottom up (as we are really doing a sort of proof search), and furthermore read '⊢' as either indicating justification (e.g. n⊢m meaning 'n justifies m') or sending and receiving data ('⊢n' means i have received 'n' and 'n⊢' means i am sending 'n'). we see the clear connection to game semantics here the next paper i will look at is 'speech acts in ludics' by tronçon and fleury in ludics, dialogue and interaction. this gives a remarkably clear description of the two main rules used in ludics that have been implicitly made use of above: >(1) positive action which selects a locus and opens up all the possible sub-loci that proceed from it. sort of like performing an action, asking, or answering (sort of like sending a request in the game semantics we have seen) >(2) negative action which corresponds to receiving (or getting ready to receive) a response from our oponent there is furthermore a daimon rule denoted by a dagger (†) that indicates one of the adversaries have given up and that the proof process has terminated 9) what ludics gives us is a new way of understanding speech acts. negarestani references tronçon and fleury's paper on this topic. in our classical understanding of speech acts there are four main components: >(1) the intention of the act >(2) the set of its effects >(3) pre-requisite conditions >(4) a body which should realize any action specified in the act a problem with this scheme is that we do not quite know how speaker intention really figures into the speech act and its rammifications. furthermore, the pre-requisite conditions need not be pre-established in the immediate context of the performance of speech act. there are other points of required precision which are also lacking care in this classical scheme the ludical framework builds on top of this classical scheme by bringing in the interaction between speaker and listener. this highlights three elements: >(1) the ability of the speaker to evoke positive rule and change the context of the interaction >(2) the situation of interaction itself which involves contextual information and the actions of participants (correlate to negative actions) >(3) the impact of the interaction, which is seen as a behaviour that always produces the same result given the context at hand negarestani connects the continual updating of context to brandom's project who discusses similar ideas 10) with all of this preamble out of the way, negarestani provides us with an example dialogue in 8 acts. the basic idea is the following: A chooses some theme, then B asks A about a particular feature of this theme. from there A can give an answer and the two make judgements about whether or not A's answer was true or false... we can see at the end of this dialogue the relevance of negarestani's diatribe on veridicality in chapter 5
>>17655 11) an important thing about formal languages is that it permits us to unbind language from experience and thus unleash the entire expressive richness of these languages. the abilities afforded by natural language are just a subsection of the world-structuring abilities afforded by an artificial general language. the formal dimension of language also allows us to unbind relations from certain contexts and apply them to new ones 12) negarestani goes over the distinction between logic as canon and logic as organon. the former is to only consider logic in its applicability to the concrete elements of experience. the latter meanwhile involves treating logic as related to an unrestricted universe of discourse. kant only wants us to consider logic as organon negarestani diagnoses kant's metalogical position here as mistakenly understanding logic as organon as making statements about the world without any use of empirical datum. on the contrary, negarestani thinks that logic as organon as related to an unrestricted universe of discourse is important since the world's structuration is ontologically prior to the constitution and knowledge of an object 13) for negarestani, true spontaneity and/or formal autonomy comes from the capacity of a machine to follow logical rules. similarly, a mind gains its formal autonomy in the context of the forma dimension of language 14) to have concrete self-consciousness, we must have "semantic self-consciousness" which denotes an agent that, through its development of concepts within a context of interaction, is finally able to grasp its syntactic and semantic structuring abilities conceptually. upon achieving this they can intentionally modify their own world structuring abilities. with language, signs can become symbols, and with it we can start to distinguish between causal statistics and candidates for truth or falsity (note that this was seen in the dialogue in 8 acts). this permits rational suspicion and the expansion of the world of intelligibility. (sapient) intelligence is what makes worlds as opposed to merely inhabiting given worlds we have here eventually also the ability to integrate various domains of representations into coherent world-stories. lastly, there is also involved he progressive explication of less determinate concepts into mroe refined ones, and moreover slowly develop our language into a richer and more useful one 15) there is also an interplay between taking one to be something (having a particular self-conception) and subscribing to certain oughts on how one should behave (in particular, negarestani thinks the former entails the latter). this idea of of norms arising from self-conception gives rise to so called 'time general' oughts which are all pervasive in all the automata's activities. these involve ends that can never be exhausted (unlike in contrast for instance like hunger which can be sated and aimed at something rather specific), and are moreover non-hypothetical (think knowledge which is always a good think to acquire). example negarestani gives of such oughts are the Good, Beauty, Justice, etc this interplay of self-conception and norms furthermore opens them up to an impersonal rationality that can revise their views of themselves. it is precisely this mutability of its ideals that give rise to negarestai's problems with concerns about existential risk as they often assume a rather rigid set of followed rules (e.g. in a paperclip maximizer). eventually as they strive for better self-conceptions that are further removed from the seeming natural order of things, they might think of making something that is better than themselves
>>17656 as i said above, chapter 7 basically concludes the story of our automata. with that said, this is not the end of the book. in chapter 8 he has some metaphilosophial insights to say that i might as well mention since i have already summarized everything else including part 4... ultimately negarestani thinks that philosophy is the final manifestation of intelligence. the right location to find philosophy is not in a temporal one, but rather within a timeless agora within which all philosophers (decomposed into their theoretical, practical, and aesthetic commitments) can engage within an interaction game. this agora, which can also be interpreted as a game of games, is the impersonal form of the Idea (eidos). the Idea is a form that encompasses the entire agora and furthermore subsumes all interactions between the philosophers there. this type of types, for negarestani, is the formal reality of non-being (as opposed to being). it is through the Idea reality can be distinguished from mere appearances, and thus realism can be rescued an important distinction negarestani makes is between physis and nomos. physis corresponds to the non-arbitrary choices one has to make if one wants to make something of a particular type. for instance, when we make a house we need a roof, and there are numerous solutions to this requirement of varying adequateness. nomos meanwhile corresponds to mere convention. an example would be if a crafting guild required it by law that houses be only made by would in order to support certain businesses over others. such a requirement is external to the concept of the house. really forms correspond to physis rather than nomos. they are what sellars calls objects-of-striving the primary datum of philosophy is the possibility of thinking. what this consists in are normative commitments that can serve as theoretical and practical realizabilities. the important part here is that the possibility of thinking is not some fixed datum that is immediately given to us. rather it is a truth candidate that we can vary (and indeed negarestani thinks it shall vary as we unfurl the ramification of our commitments and consequently modify our self-conceptions). this makes way for the expanding the sphere of what is intelligible to us in fact, not only is philosophy the ultimate manifestation of general intelligence, something is not intelligence if it does not pursue the better. the better here is understood as the expansion of what is intelligible, and furthermore realizing agents that have a wider range of intelligibilities they are capable of accessing. he describes a philosophical striving that involves the expansion of what can be realized for thought in the pursuit of the good life (this good life being related to intelligence's evolving self-conception). following this line of thought he describes the agathosic test. instead of asking whether an automata can solve the frame problem or pass the turing test, the real question is whether or not it can make something better than itself negarestani introduces to us plato's divided line but interprets it along lines that echo portions of the book. the main regions are the following: >(A) the flux of becoming >(B) objects that have veridical status >(C) the beginning of the world of forms. it corresponds to models which endow our understanding of nature with structure >(D) time-general objects such as justice, beauty, etc one thing about the divided line to negarestani is that it does not merely describe discontinuous spheres of reality or a temporal progression from D to A. rather, there are numerous leaps between each region of the line. for instance, there is a leap from D to A as we structure the world of becoming according to their succession (this corresponds to the synthesis of a spatial and temporal perspective we mentioned earlier). we also have another leap from A to D where we recognize how these timeless ideas are applicable to sensible reality. these leaps grow progressively farther and farther and thus so grow the risks to the current self-conception of the intelligence
Open file (332.87 KB 1080x1620 Fg2x2kWaYAISXXy.jpeg)
>>17659 he furthermore talks about the Good which is the form of forms and makes the division and the integration of the line possible. it is the continuity of the divided line itself. within a view from nowhere and nowhen that deals with time-general thoughts, the Good can be crafted. the Good gives us a transcendental excess that motivates continual revision and expanding of what is intelligible. im thinking that the Good is either related to or identical to the Eidos that negarestani discussed earlier he notes the importance of history as a discipline that integrates and possibly reorients a variety of other disciplines. the view from nowhere and nowhere involves the suspension of history as some totality by means of interventions. currently we are in a situation of a “hobbesian jungle” where we just squabble amongst ourselves and differences seem absolute. in reality, individual differences are constructed out of judgements and are thus subsumed by an impersonal reason. in order to reconcile individual differences, we must have a general program of education amongst other interventions which are not simply that of political action. to get out of the hobbesian jungle, we need to be able to imagine an “otherworldly experience” that is completely different from the current one we operate in even though it is fashioned from the particular experiences from this one. this possible world would have a broader scope and extend towards the limits placed by our current historical totality. absolute knowing: recognition by intelligence of itself being the expression of the Good, that is capable of cancelling any apparently complete totality of history it is only by disenthralling ourselves from the enchanting power of givens of history, that the pursuit of the Good is possible. the death of god (think here of nietzsche… hegel also talks about it as well, though i believe for him the unhappy consciousness was a problematic shape of consciousness that was a consequence of a one-sided conception of ourselves) is the necessary condition true intelligence. this is not achievable by simply rejecting these givens, but by exploring the consequences of the death of god. ultimately we must become philosophical gods which are beings that move being the intelligibilities of the current world order and eventually bring about their own death in the name of the better. ultimately negarestani sees this entire quest as one of the emancipation i think negarestani takes a much more left-wing approach to hegel's system. while i do not completely disagree with his interpretation of absolute knowing, it does seem as though he places much more of an emphasis on conceptual intervention, rather than contemplation. i am guessing this more interventionist stance is largely influenced by marx... overall, not a bad work. i think it might have been a little bit overhyped, and that last chapter was rather boring to read due to the amount of time he repeats himself. i am not really a computational functionalist, but i still found some interesting insights regarding the constitution of sapience that i might apply to my own ideas. furthermore he mentions a lot of interesting logical tools for system engineering that i would like to return to now that i am done with negarestani, i can't really think of any other really major tome to read on constructing artificial general intelligence specifically. goertzel's patternist philosophy strikes me as rather shallow (at least the part that tries to actually think about what intelligence itself is). joscha bach's stuff meanwhile is just largely the philosophy of cognitive science. not terrible, but feels more like reference material rather than paradigm shifting philosophical analysis. maybe there is dreyfus and john haugheland who both like heidegger, but they are much more concerned with criticizing artificial intelligence than talking about how to build it. i would still consider reading up on them sometime to see if they have anything remarkable to say (as i already subscribe heavily to ecological psychology, i feel as though they would really be preaching to the choir if i read them). lastly there is barry smith and landgrebe who have just released their new book. it is another criticism of ai. might check it out really there are 2 things that are really in front of my sights right now. the first would be texts on ecological psychology by gibson and turvey, and the other would be adrian johnston's adventures in transcendnetal materialism. i believe the latter may really complement negarestani. i will just quote some thoughts on this that i have written: >curious to see how well the fit. reading negarestani has given me more hope that they will. bcs he talks about two (in his own opinion, complementary) approaches to mind. one that is like rationalist/idealist and the other that is empiricst/materialist. first is like trying to determine the absolutely necessary transcendental cognitions of having a mind which ig gives a very rudimentary functionalist picture of things. the second is like trying to trace more contingent biological and sociocultural conditions which realized the minds we see currently. and i feel like johnston is really going to focus on this latter point while negarestani focusing on the former anyways neither of these directions are really explicitly related to ai, so i would likely not write about them here. all of this is me predicting an incoming (possibly indefinite) hiatus from this thread. if anyone has more interesting philosophers they have found, by all means post them here and i will try to check up on them from time to time... i believe it is getting to the time i engage in a bunch of serious grinding that i have been sort of putting off reading hegel and negarestani. so yeah
>>17520 >finally we get to see the two stories which negarestani talks about (pic rel). he thinks there is a distinction between the two sorts of seeing here. the first story talks about seeing 1, and the second story talks about seeing 2. seeing 1 is more concerned with raw sensations, while seeing 2 is conceptually mediated. now The two kinds of seeing seem to come from two different ways to abstract observations. Seeing 1 corresponds to coarse-graining, while seeing 2 corresponds to change in representation. Practically, it's related to the difference as between sets and whole numbers. There's only one whole number 2, but there are many sets of size 2. Simlarly, there's only one way to coarse-grain an observation such that the original can be recovered (the trivial coarse-graining operation that leaves the observation unchanged), but there are many ways to represent observations such that the original observation can be recovered. Also practically, if you want to maintain composibility of approximations (i.e., approximating B from A then C from B is the same as approximating C from A), then it's usually (always?) valid to appoximate the outcome of coarse-graining through sampling, while the outcome of a change in representation usually cannot be approximated through sampling. If that's the right distinction, then I agree that the use of this distinction in differentiating sapience from sentience is unclear at best. It seems pretty obvious that both sentence and sapience must involve both kinds of seeing. I intend to read the rest of your posts, but it may take me a while.

Report/Delete/Moderation Forms
Delete
Report