(untitled)

Hypothesis: At the most basic level, our decisions and ability to detect patterns can be explained mathematically. Hypothesis 2: Understanding how we do this will be instrumental in developing artificial intelligence. So can you explain how humans are able to detect patterns /hum/? Bonus points: in mathematical terms, how do we adapt to situations and solve problems?

Sigmoid activation function that acts upon biased inputs from other neurons doing the same function on input they get on sensory inputs.

>>8216 aren't you suppose to prove these or at least make a proposition? This is equivalent to spewing sokyaf. Well since I don't like math that much I would say your hypothesis' are wrong I think we will be using a combination of neural networks and hand written logic to make AI capable of deducting new information from old ones.

>>8217 Can neurotransmitters be described on the Cartesian plane? I'd think since there are many of them, the function would be in a higher dimension. >>8218 We can trace everything back to elementary particles in physics. It's not unreasonable to predict all these particles' behavior can be described mathematically. Ideally, we can abstract it out to neurons like >>8217 suggested; it's vastly complicated and may be incomprehensible to us if we cannot use abstraction.

>>8230 Current machine learning techniques can help us bridge the gap into something we comprehend

>>8218 He doesn't have to prove a hypothesis, since he never claimed it to be a theory or fact. It is a first draft you make before actually trying to find scientific evidence. >Well since I don't like math that much I would say your hypothesis' are wrong I think Emotion is never proof.

>>8230 The disturbing thing about that is: are we capable of describe pecfectly the function of our mind by using our own mind?

>>8239 Maybe not but >>8231 Also, through observation it's possible we can abstract it to something ridiculously simple like e=mc^2. In this equation we don't know why or how it the most granular parts of energy or matter bubble up to this equation, yet we know it works.

At the topic of AI. When do you think second AI winter will occur? Or can you even treat current neural network boom as hype of early AI research?

>>8239 >The disturbing thing about that is: are we capable of describe pecfectly the function of our mind by using our own mind? Not going deep enough. As Carl Sagan said, we are the universe's way to explain itself. I hope one day we fully can. Stardust forming a planetary body, life spawning, then this life itself learning about everything. Truly an amazing thought.

>>8217 it goes deeper than that though. i think the real question here is "can consciousness be approximated?". that's what the the convolution neuronal lattice model is trying to do, but it may very well be the case that consciousness depends on complicated nonlinear dynamics "all the way down" past the atomic level. it could also be that the "consciousness attractor" is unstable, and thus any minute error in the approximation will eventually blow up and drown out all signal.

Let me present an idea on the relationship of mathematics and reality and why we'll never be able to completely explain/predict/describe the universe. Let's have things and their representations. ooooooooooooooooooooooooo 25 os ooooooooooxooooooooooooooo 10 os x 15 os or 25 os divided 2:3 by an x Notice how the original string of os is kinda harder to read, you are prone to counting it wrong. But its also dead simple, just one character and you don't have to interpret or understand anything: it is just what it is. Meanwhile "25 os" is an abstraction or representation that is significantly more complex. You can compress it well because the thing in itself is simple, you pay with a larger set of characters and the need to understand the context (that "25 os" is not the string itself, it's something you have to decode). Something with less patterns and simplicity would be harder to compress, while something that is completely random and lacks any pattern can't really be compressed at all - hence why very high levels of abstraction may seem like random bullsoykaf until you finally get it. On a small scale, using stuff like "25os", basic arithmetic, abstraction with variables can get you ahead. When you get a clear understanding (or system of axioms) of what you can and can't do, you might be able to infer truths you could not infer from the things themselves. Nobody every tried a scenario where they had a hundred thousand cows, killed twelve thousand by smashing their head with elephant bones, and burned half of the remaining ones to death to see how many cows are left. Yet we can still know there's 44k left. This kind of activity is limited by the small scale you are on, which is defined by how complex your brain is. The brain is pretty fuarrrking complex. It's made of <big number> of neurons which themselves are complex. It is able to handle a large amount of similar things (differentiate colors, pictures, names) while it can also handle several levels of abstraction. It needs abstraction to handle very flat stuff (thousands, millions of the same thing) while it needs flatness to handle very difficult abstractions (put it down on paper or in code, think only about a small part of it at once). There are however things which in their raw, flat form or after any layers of abstraction (decompression or compression) are still both too much and too abstract for you to understand. Since your brain has some jobs besides understanding itself (its full capacity is shared between your memory, urge to breathe, motor skills, etc), you can rest assured that your brain is unable to understand itself in exact detail and full scope. You can't argue for a bigger size, because then whatever has to be understood would also grow. You also can't argue for more abstraction, because the brain to be understood would become even more abstract. If this sounds counter-intuitive to you, think about computer memory. A Kingston 16GB Desktop RAM DDR4 - KVR24N17D8 weights about 100 grams. It can hold 8 x 16 x 10^9 bits of data, while is made of over 10^23 atoms (gross underestimate). While a good portion of that is wasted on wires that connect the RAM to a motherboard and also internal structure for flipping those bits, it still requires orders of magnitude more atoms to represent a given number of bits. Meanwhile, to represent an atom requires a good many bits. So you cannot represent it atoms by atoms on itself, that would require 10^23 x "good many" bits of data, while you only have 129 x 10^9, and it certainly isn't intelligent enough to understand the blueprints (abstraction) of itself. So their own brain is an upper limit of what a human can understand. Now let's add the rest of the body. Or the planet. The overall shape/picture might change with these additions, but the brain is still a subset that doesn't fit in itself. This limit will eventually apply to AI as well, regardless of its size and complexity, it won't be able to understand complexity that is exactly itself. If the AI is more complex than humans, humans won't understand it and thus understanding humans won't help it any more than understanding dogs helps us. If humans still understand the AI, it's less complex than humans. If a human and an AI understand something as a whole that is beyond the understanding of either, they won't be able to exit this symbiosis without losing the understanding, and their symbiosis is limited to itself in complexity. The only question that remains is whether the entire universe does or doesn't understand itself. Maybe the idea I presented above holds and applies to it. Maybe it breaks down, like all abstractions of the universe do on a large scale, because the abstraction is simpler than the universe that holds it.

>>8259 That is a very soykafty analogy, especially since real data like DNA sequences can be compressed quite well. Especially since start and stop codons exist you can even ignore the parts not encoding anything meaningful at all. Perhaps you should look up the CRAM algorithm first and read about asymmetrical numerical systems before pulling stuff out of your soykaf cavity. If your hamfisted example only had any relevance to the topic before you start rambling into an all offtopic showoff what has been known for years about storage theory and data entropy. Literally wikipedia recapping tier. So where is that data that you speak of which is so "random" that we cannot understand it? Neuroscience is just full of scammers living of highly remunerated budgets. The data is not random, we are simply using the wrong tools. Which was Q.E.D.'d in the linked article below: https://arstechnica.com/science/2017/01/can-modern-neuroscience-understand-donkey-kong/

my citation is tripping balls but I dont believe anything in the world is truly patterned. I think our brain patterns things because its efficient and convenient. Some things appear very patterned, some things have a pattern. But not innately, perhaps on some level even the things we design with a pattern arent truly patterned. when you look at your wall paper even if its just spackling you already have an idea of what it looks like. I would reckon right now you dont see your wallpaper as it actually is. In fact you dont see most things as they are. Hence why someone can put some new item in our fridge and we can miss it everyday. People say this is because its easy for us to not see what we dont expect, and id call that or attribute it to patterning. Perhaps it even has something to do with why our world appearing so stable when in fact its wibbly wobbly. I would say im 90% confident in everything I said. And I think if we wanted to make a really solid artificial intelligence perhaps it might be better if it didnt pattern everything. A different system that allows it to see things as they really are.

the problem with "we can explain it mathematically" is that it may be technically true, but just because you *can* explain something mathematically doesn't mean it makes much *sense*. If the brain takes 20 exabits to store or something, a human simply cannot derive meaning from that kind of data without using some structure to interpret it. In many ways a computer program is the only way we can reason about such a thing rigorously. We can show that the program is correct, and then use the program to analyze the brain in some way that gives all that data meaning to us. Basically the entire field of neuroscience is trying to build structures that work for analyzing that insanely complicated structure. It's not easy, but great strides are being made. Come back in 100 years and maybe there will be some good ideas, for now we're still mostly in the dark.

>>8267 >my citation is tripping balls but I dont believe anything in the world is truly patterned. Moving the goalpost but I'll bite. The periodic table is patterned itself, not just because we arranged it like that, but since the electronic structure of electron shells repeats. Celestial bodies are on orbits around a central star. Pulsars emit patterns based on their size and age. So pattern recognition is useful. 1) logarithmic hearing 2) optical illusions forming patterns where none exist, given how the visual receptors and the cortex are wired 3) filter mechanisms >>8275 You assume things again here that are not relevant. The brain of a cat is the size of a walnut. Storage is not important. Some insects based on spinal ganglia exhibit the most impressive flight patterns and they run on even less nerve tissue. You just want to keep bringing up this hamfisted argument about storage. You don't have to fuarrrking save the data at all, it is just good enough if your animal of choice can process it. With what? Pattern. Recognition. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3084041/ READ: "If, for instance, a vertical dark stripe is presented to a fly in a fronto-lateral position and is slowly oscillated around this position, the fly's yaw torque shows a characteristic modulation. During the front-to-back movement of the stripe, the fly generates a strong torque response in the same direction as the stimulus motion..."

>>8279 yeah, from what i know of neuroscience, memory isn't actually stored anywhere in the brain. it's just sort of "encoded" in the trajectory of your brain-state.

>>8280 Whilst you are right that there are probably no memory proteins or molecules, your brain still heavily remaps synapses during learning processes.

This is old news. The mathematical underpinnings of neural nets: https://youtu.be/uXt8qF2Zzfo They really are amazing. It's important to note that the math is just an abstraction we use to help reason about how the brain works. Nature doesn't have to conform to an equation, it just works.

>>8279 I know its very useful. Im more going off in lala land and saying even things when they appear to be extremely patterned fundamentally arent. That the structure of the universe can't actually be predicted by patterns but for the purpose of survival and growth it can for all intents and purposes appear very predictable. And I also predict that at some point when we have things extremely figured out there will still be questions we just cant seem to solve and it will be a result of that. And a new way of "patterning" will have to be made. Again my only citation is that I am extremely confident this is how my own brain works. And that as a result of patterning I miss many many things. And im assuming my brain is mostly similar to all peoples.

>>8285 I don't have to disagree with everything to feel great, but I see a structure in the cosmos. Even though the Millenium Simulation is just that, an artificial recreation of matter distribution in our cosmos based on gravitational laws and expansion. Some insane people see patterns everywhere: https://www.youtube.com/watch?v=jo18VIoR2xU I still believe everything is governed by patterns.

>>8286 its possible im not one of those insane people but do see more patterns then normal and so in response i assume the world to not be actually patterned because of when my overpatterning fuarrrks up.

A mapping between patterns perceived by our senses and patterns in our gray matter, a network of concurrent processes in real time. Such concurrent processes have a memory of sorts implicit in it's transmission medium (stronger or weaker synapses), though the function could be as simple as a sigmoid. I used buzzwords for mathematical terms, if you don't mind.

>>8279 >Storage is not important. it is if you want to fully describe the system. That's what I was talking about, not the sensation of memory or whatever. >You just want to keep bringing up this hamfisted argument about storage. that was my first post in this thread. I was talking about our ability to meaningfully interpret the way minds work, even if all the data (the mathematical explanation you hypothesize) is known. As a human, what matters is that I could understand it, not that "it is known".

>>8318 oh, and to be precise: it's pretty obvious that the human brain is very high entropy. So storing it, or detecting patterns(aka compressing it) will be difficult.

>>8318 >>8321 >fully describe a system when you already fail to describe the inputs of it and the filters that is some nice soykaf

>>8216 The best place to start looking is our genes. We must understand them, and how they vary in strength across one's life. Humans have a ridiculous amount of layering that dictates what they do. Take the OSI model for example, while technically elementary particles control it, we abstracted those particles into binary. From Boolean logic we were able to build up in layers from bits to something abstract such as transmitting video in real time.

bump

>>8322 it's a hypothetical for a reason, dumbass

>>8216 That wasn't a serious question, but here's a serious answer: https://www.fil.ion.ucl.ac.uk/~karl/The%20free-energy%20principle%20A%20unified%20brain%20theory.pdf

>>10592 What other resources do you have that can be applied to computational neuroscience

>>8216 we make decisions by predictive repair. both perception and action are accomplished by neurons approximating variational bayesian analysis to bet on what states will keep us homeostatic -- the primary drive is to reduce uncertainty. all of our qualia are measures of how far off our bets were. every sensation is a predictive mismatch being corrected -- top-down schema information was compared to bottom-up sensory data, and the schema was adjusted slightly. this is accomplished at the synapse level by a control loop that uses dopamine levels to measure AMPA vs NMDA receptor activity and catalyze the production of more of whichever receptor type is needed, thus preparing for a firing pattern based on previous input. mammal sentience is a quirk of attachment heuristics -- young mammals have to stay warm and safe, as they age they have to get more paranoid to survive. thus, mothers establish a pattern of care and then gradually withdraw it, inducing infants to mistrust the world and assume it is full of hidden variables. early on we crave certainty as sensory extinction (feed me and let me sleep), later we crave certainty as active experimenters (let me explore what's out there). violations of schema -- bad bets -- are inherently uncomfortable. thus, making predictions about your own stress level is inherently circular. the illusions of time and agency come from human intelligence is a quirk of bipedalism and hairlessness -- narrow pelvises, premature infants, long helplessness. no hair means no chimp-cling, so we're the only great apes who alloparent. evolution conditioned human infants to be experts at exploring the world using only our own facial expressions, since we can't move on our own for months. face mirroring and triadic intersubjectivity based on the infants' need to distinguish and manage caretakers let us exapt single-word language for displaced reference about 1.2Mya, allowing us to recruit conspecifics to distant corpses (only others who can are ravens and bees). implicit topicalization and synonymy avoidance built up an implicit productive syntax, and around 70Kya a new mutation gave us recursive prediction. this mutation would have died out in any other species, since it encourages insane apophenia, but was so useful in the context of our existing syntaxless protolanguage that it caught on. i blame scholz's star, their oort gave our oort the herp. and there you have it. https://www.fil.ion.ucl.ac.uk/~karl/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3562439/ http://faculty.spokanefalls.edu/InetShare/AutoWebs/SarahMa/Hrdy.pdf https://dohdqafsiel.wordpress.com/2017/05/12/category-tweaks-a-case-for-clarifying-recursion-and-enlarging-fln/

>>11978 Considering recursive prediction, similar things have been implemented to axioms in an attempt to achieve Godel completeness. Can a human brain be considered Godel complete? Can our brains be implemented via a Turing machine? I understand that emotions can be a heuristic for decision making. This could be induced by a genetic algorithm but I don't know how to efficiently have this communicate with our rational counterpart. >the illusions of time and agency come from Don't leave me in suspense like that :(