This has been the plaguing my thoughts recently as well. When I analyze (or attempt to analyze) what’s going on in the black box of gray matter in my head, I’m increasingly aware that it’s nothing magical. I’ve stated to others that the real shakeup the latest AI successes may cause won’t be that we’ve created an artificial mind, but that our own minds may be nothing more than a collection of neural subsystems. Subsystems process our vision and pass it on. They process our hearing and pass it on. They process our senses of proprioception and pass it on. Then our cortex processes all these inputs and produces outputs, which may be speech, physical reactions, or just thoughts that loop back in as another input to the cortex. Maybe our consciousness is just a supervisory neural network that monitors what the cortex is doing and feeds in adjustments. But instead of prompt/response, it’s always running, feeding back on itself, trying to predict what’s going to happen next and adjusting its models based on what we observe. It explains a little too much and I start wondering if there really is a “me” that can be described as a conscious personality and I get a little creeped out.
One consolation is that we’ve always equated some new technologies with “maybe that’s the way the mind works.” The mind is like a book. The mind is like a machine. The mind is like a calculating device. The mind is like a computer. And now we’ve reached: The mind is like a super LLM with assorted subsystems. The difference is, this time, we may be right since, for really the first time, we’ve started simulating it from the bottom up.
Maybe our personality and being is just a result of our training data with a few genetic quirks thrown in. And what am I really doing when I “decide” or “create” something? Is it that much different from a LLM’s generation of language? Or an AI image generator?
> the real shakeup the latest AI successes may cause won’t be that we’ve created an artificial mind, but that our own minds may be nothing more than a collection of neural subsystems.
It should be no surprise. I'd have thought most HN readers have at some point attempted to kill their own ego.
Maybe it started when you were younger and learned that society used to believe the Earth was the center of the universe. Perhaps you then considered that humans too are nothing special — maybe not created in God's image after all? Instead the latest evolutionary result in a long complex process.
Dashing geocentrism, anthropocentrism should have made many of us skeptics about anything considered special.
Why should consciousness be inscrutable, magical?
I've been sure we're little more than imperfect machines for some time now.
When I was about 3 I learned that the world did not revolve around me. Eventually, I learned my country wasn't special, nor the culture I was raised in. I learned the Earth was just another planet, and humans were made out of meat just like other animals, all of it governed by the laws of physics. There are no special rules in this universe just for us. We are not special.
None of this struck me as a revelation, it just felt like growing up.
It's honestly a little sad that something so simple as an LLM is causing so many people to come up short against these basic philosophical assumptions. The human brain weighs about 1.4kg (most of that being metabolic or structural support), runs on about 12 watts, and it takes only 3.2 billion base pairs to make a new one with a default template and peripheral systems already included. It is not that special.
The number of human DNA base pairs. The data to grow a starting zygote into a brain with a default neural template (to keep it alive and allow it to start learning), and peripheral systems (the body).
Oh, it looks like a tabula rasa theory. No mithohondrea DNA, no special places in parents' body, no special behaviour required from parents. Just an 80-minutes audio CD.
Cosmologically, the world can be full of special, beautiful, unique things AND be merely the result of some computation. That is what a fractal is, no?
So if your concern is with your mind being special, that's not a problem that needed to be addressed by saying "nothing else in nature possesses this particular fractal boundary". You merely have to be one point on the curve, somewhere, and to the extent that an AI could replicate it, it would still be an approximation of such. It might be a very good one, but it wouldn't have the same causal relationship to nature, and therefore would be special in a different way, for the same reason that when you turn over an hourglass, the grain of sand that was special for being the last to fall now becomes the grain special for being the first to fall - it's "just" sand, but it's also positionally different.
So the world could be deterministic, but the interpretive meaning of that might not be "chains of fate," but a perpetual explosion of color and possibility, in which I will never know for certain in what way I might be unique, but it is far more likely that I am than not.
To me being sceptical of COVID vaccine is entirely reasonable. Looking back at how we evaluated vaccines before it should make it clear that something went wrong.
And when those pushing for it can not even have honest discussion of effectiveness like does it prevent spread or just worst cases? The first one has long been standard for vaccines, but this one was later. And then it seems even vaccinated did get the disease, so why did we use the vaccine status as metric, instead well more effective and certain testing?
Science is about question and understanding. But we should also continue to question what we think we understand, because the hubris of man has often lead us to believe we're more right than we are.
If you won't even consider the possibility that the covid vaccine is harmful, even only to an unknown subset of the population, you're not exactly one to be calling others less sophisticated.
That’s not what OP said. It is possible, indeed likely, that in some small subset of people, the COVID vaccine(s) do more harm than good.
However, the evidence suggests that the negative impacts of COVID itself outweigh by an order of magnitude the risks of vaccinating the whole population.
If we are able to identify those people for whom the vaccine is likely to be more harmful than protective, we can weigh up options sensibly, but without that concrete data, it is hard to justify maintaining a position that the “vaccines are not safe” as a generalisation given the overwhelming body of evidence that shows that the vaccines are effective at limiting community transmission, improving outcomes for those who are infected, and reduces the likelihood of infection in the first place.
The statistical argument should consider both the likelihood of infection when unvaccinated and the likely negative outcomes from such an infection and weigh it against the likelihood of an adverse reaction, the likelihood of infection post-vaccination, and the likelihood of negative outcomes from such an infection post-vaccination.
These factors are all changing quickly, especially the likelihood of infection as community infection rates drop, but based on my understanding, we have not yet reached an inflection point where the risk of a negative outcome post infection outweighs the likelihood of infection with negative outcomes without vaccination - at least on a community health level. Individuals may be able to further influence these factors themselves as well, reducing exposure by not using public transport, or not attending events with a high concentration of people, or wearing face masks, etc., which means that ultimately, it boils down to individuals trying to make informed decisions about their own health with incomplete information.
This is a very difficult thing to ask people to do, especially considering that half of the population have below average intelligence levels, and may struggle to interpret the data for themselves. As a result, advocating an “anti-vaccination” position likely increases the net harm done to the community, and is not at all conflicting with making a personal decision based on your own research and interpretation of evidence.
This is basically what I meant. I recognize requirements to be vaccinated to enter a restaurant is stupid if you require it with the expectation that you're making other people safer. But as a whole we're all safer if we're all vaccinated because it was proven to put less strain on hospitals, and it's somewhat empty hope is what made people feel comfortable with getting back to "normal". Which is important because a lot of people did suffer in very real ways from lock downs.
I support and encourage skepticism, and especially with anything that has a profit motive (including vaccines). But it was pretty clear that COVID vaccines were more good than bad.
Science is inherently a social and norms-based process, so it is more accurate to state it is about _mutual_ questioning and understanding.
> If you won't even consider the possibility that the covid vaccine is harmful, even only to an unknown subset of the population, you're not exactly one to be calling others less sophisticated.
Curious how the "vaccine-skeptic" crowd only directs its skepticism in one direction.
> There's plenty of anti-COVID vaccine discussion as one example.
Let's be honest, anti-COVID vaccine discussion is not the same as anti-vaccine discussion in general. Latter is ultimately stupid, but former... might be just a reaction about too severe government's reaction on some situation.
How about discussion about different anti-COVID vaccine from different sources - are you really treat Western vaccines and Russian ones as the same? These two give you kind of different political effects when you can visit one list of countries with one vaccine and another list of countries with another one. I do not remember any discussion like that with any other vaccines.
And by the way, there were a lot of changes of upvotes/downvotes near my comment, thank you for not being silent about such controversial topic.
We are created to God's image. Everything is like God designed, to be able to work at all.
What that long philosophical post is about, that thinking that much just leads to questions about existence, and Christian religion has answers.
Spending too much time with computers can cause thinking, that everything is a computer. It is not. Being a human has all feelings etc extra stuff. Watching Matrix movie too many times can make someone imagine we are in Matrix. We are not.
It is just, that world is going through hype bubbles, when everything at news is just about one topic: 1. Corona, 2. War, 3. AI.
AI is about pattern matching. Calling many old
stuff with new name AI, just to sell stuff.
Making it up as they go.
AI is just some hardware and software, human written answers,
recognizing text/image/etc and matching it with some answers.
But it only answers what it is asked about,
and makes up something that may not be true.
So there is need for humans that try to understand is that
answer useful or not. Humans will be always needed,
to help verify answers are not too dangerous.
If AI asks AI, that kind of system usually causes
AI to hallucinate more and mess up usefullness
of that AI model. If AI would "think something",
it needs checking, is that useful at all.
AI can not take over the world, those thinking that
has just watched Terminator movies too many times.
Most important data is at offline local networks.
Every online service is busy adding more security protections
and trying to always stay online.
AI, as software, is useful for many purposes. It is a new tool.
New tools helps in many kind of work.
This has happened many times, like after horses there were cars,
it can change what kind of work is needed.
There is big need for more coders. Coders are needed to debug
code, check does code have vulnerabilities.
AI tools help programmers to figure out some syntax,
like Bing AI helped me to convert SQLite SQL Query to
MongoDB query.
Just do not copy paste propietary code and secrets to AI chat.
There is no magic. Magic means, you are using some technical terms,
and writing unclearly, just to hide something, to sell some stuff.
The same can be written in easy to understand language.
Treat it positively as a mindfulness exercise. None of the stress you experience daily is more than a series of trained micro-reaction chains. With enough training you can start to subvert the chain reactions in your own mind. You always have control, you simply have to accept a more abstract form of it.
What's special about self-awareness? Could it be a trick of the brain?
It's like the story of Jimi Hendrix dropping acid and playing amazingly well. Or, rather Jimi under the influence thought so — until he heard the recordings.
A "magic" trick perhaps? One that can't be explained by anything we currently know about physics? The materialist point of view is becoming increasingly incoherent and I think that's what angers many of them.
Not really though, as I understand it. Simulating from the bottom up would be akin to simulating neurons and such, or even the molecules that make up the brain and the connected systems. What we've done is creating a complex system whose emergent behavior's output is similar in specific ways to a human brain. And significantly more so, compared to the previous attempts.
And yes, many aspects of our existence is comparable to a machine. Which can be a very good thing, since we can apply many of the knowledge to get improvements, like we can train, debug, and fix some things. In fact I consider the realization you're having essential for personal development, because you can now leave behind things that you previously considered "you", now that they turned out to be not really "you" but just something you do, and most others do too, to get closed to what "you" really are. And do better in things that you think "you" are not!
Yes, I know it’s not modeling our neurons and connectivity directly, but it’s closer than traditional software or “expert systems” approach. And, lo and behold, we’re seeing emergent behavior that seems to resemble some of quirks of our meat network.
Even the mistakes it made seem like the kind of mistakes people make, not the kind of mistakes software usually makes.
LLMs are more like a natural language interface for computers that do symbolic calculation of bytes aka computer science. Statistical madlibs for text generation, but useful for understanding people. There is a deeper and simpler logic of nonphysical consciousness that is perhaps best outlined in George Spencer Brown’s Laws of Form. LoF can do boolean logic but goes beyond it as well, towards the resolution of antinomy paradox with the logic of distinction andedness. It is also an interesting philosophy when it comes to Systems Theory of Luhmann. As well as used to do other very interesting things. There is an intro on Youtube by Leon Conrad.
Maybe you already know, but this 'it's all just a train of mechanical compulsions, there's nobody in here' is the very thing that meditation is intended to allow you to realize, it's the 'liberating insight'. The problem with just reaching it on your own is you don't get the other part, the viewpoint about the world that build the case that this is a good thing, a great thing (believing there is someone there is the cause of all the dissatisfaction you ever feel). So you could check into that if you want - ajahn brahm or suchlike
I met an artist not too long ago who showed me a time lapse of a painting he worked on. I commented that it looked an awful lot like what you get with stable diffusion if you view it step by step. Turned out he was really into the AI art stuff and his thought was that while SD is missing a kind of creative leap step, it does look like its taking a process quite similar to a human artist, or at least to the way he assembles work. He was also interested in the weird hands AI makes and was painting up a big cavas with weird hands, I thought that was pretty cool.
when you say we just like AI, this imply people do not really care anything, they just mimicking human behaviour. we know this isn't true. besides, if someone scoff at your family, you will angry, this is a common/predictable behaviour. is it a bad thing?
I sympathize with the author’s (OP I believe based on the last name) feelings here and also find myself wanting to leave somethings not understood. Maybe we’ll be fortunate and it literally will be the case that some things always will be a mystery.
One thing I would like clarified: I don’t follow the refutation of epiphenomenalism. Is the intent that I can raise my hand only as a result of my conscious experience? Couldn’t one imagine instead a LLM hooked up to a robotic arm that also is trained to raise its hand it reaction to similar text? I feel like that or something similar would refute this. How can you say was is causal in sending the signals to raise the arm? Why can’t the conscious experience just be a theater?
What’s more weird to me about epiphenomenalism is that consciousness seems like it need not exist - if it performs no function than why bother with it? So why do we experience it? Sort of seems that its very existence implies that it should be more than just a theater.
I’m a total amateur about this, but would appreciate recommendations if anyone resonates with this.
The problem with epiphenomenalism seems to be that it is unfalsifiable, like every attempt at analyzing some specific structure outside the physical world. From what we've determined so far, any part of consciousness that might exist outside the physical world doesn't feed back into (affect) the physical world. So the processing of reading the words "raise your arm" can be happening entirely in your brain, leading to the result, with what we call consciousness merely following along with that chain of causality (reasoning). Epiphenomenalism would seem to cling onto the concept of duality with the physical affecting the non-physical, saying that some unspecified reasoning process could be duplicated outside of the physical world. This would seem impossible to prove/disprove, and it would also seem to go against Occam's razor.
It feels like this debate around LLM consciousness is a tempest in a teapot caused by yet another scientific advancement leaving even less room for supernatural theories. Which makes a whole lot of people worried that consciousness isn't actually so special, who then lash out to defend it as something unique to humans. I think in actuality, some scale of computational process that feeds back on itself can create what we call consciousness. I think there is no way to actually know if this is happening. We still haven't solved this for humans - I don't actually know that anybody conscious will read this comment, and you don't know that the person/process that generated this comment is conscious! I have no idea if consciousness is a possibility with current LLM implementations, especially with their lack of long term feedback - it would seem an LLM's sense of self can only ever be some integral of the sense of self of the authors of the source material [0]. But I know at a certain point we're going to have to apply the same standard of consciousness we've done for all humans, which is believe them when they assert consciousness while demonstrating significant intellect, possibly involving various displays of physical power that lead to mutual wariness/respect.
[0] although it's interesting to think about the larger feedback loop consisting of LLM output going into the training dataset for the next model, and LLM output making it into humans' brains that then write material for the training set, etc. But this would likely still not be consciousness the way we perceive it, because it's too long/slow.
"For you to do what you just did, consciousness must be capable of influencing neural activity. It must be capable of influencing the material world. It cannot, in short, be an epiphenomenon."
I find his reasoning behind this "disproof of epiphenomenalism" not just unconvincing but entirely ludicrous and I stopped reading right there.
I enjoyed the article - kudos to the author - but I got stuck on this too.
There is nothing inconsistent about the position that the biochemical operation brain is the sole cause of consciousness (epiphenomenalism), and the capacity of the brain to observe consciousness. In technical terms, imagine a program that creates a private file and writes some contents to it. The fact that same program subsequently reads the file - maybe even branches on its contents - is in no way a refutation of the fact that the file is epiphenomenal to the program.
I'm a very receptive audience to a refutation of epiphenomenalism, but this one isn't clicking for me.
Where the author did lose me though, is the subsequent discussion of pain, and the further discussion of value. These are much deeper topics which require a slower and more deliberate treatment, and I felt like they were being handwaved away with some unjustified assumptions. Why is pain irreducible to neural activity? In what way and for what reason are pain and pleasure dependent on consciousness? What does it mean for a thing to 'matter', and why can things not matter irrespective of one's conscious experience? If an unconscious coma patient is caused great pain, does that not 'matter'? What does it mean to 'confer significance' on something? Why is that the exclusive preserve of consciousness? Does an AI bot playing a video game also not 'confer significance' on threats it is programmed to avoid?
Prompted on your reply, I finished reading the article. It seems like the author is arguing from false premises in that he's desperately trying to come up with a justification for consciousness that is not tied to epiphenomenalism (I laughed out loud at the Nobel prize bit).
Since science can't do that, enter his hunches and intuitive "feelings" which are nothing but wishful thinking for a universe where consciousness matters (to whom?). I feel that one could find his "what I had been missing" finale in any number of new-age self-help books.
If you're just an LLM, then why should you have any rights, can we just turn you off as you're now an obsolete LLM? You're just using power that could go towards training a better LLM.
> You're just using power that could go towards training a better LLM.
What if I haven't found that special someone yet, the one that will help me raise the next better LLM? I think I already got like 40% of the dataset I find valuable, all I need is to find someone with the other 40% and we could go in a journey together to find the remaining 20%. Sounds like the adventure of a life time.
Why should anyone (anything?) have any rights? To answer this, just look at the history of rights. We as humanity kind of agreed on it, and kind of not, and that's the state of the why should anyone have rights. Nothing special.
"But none of this gets at the most important thing about the chili-pepper scenario: your suffering. After all, why do you care about the pain the pepper causes? Naturally, it is because of the suffering involved. It is only because you are sentient that you suffer."
Plants also respond to various stimuli in the same way but on a different time scale. Plants do care about such stimuli, whether you consider that suffering or not. So either sentience is not required for this kind of reaction or plants are also sentient. The same is true for microbes.
If they are sentient we have to rethink a moral vegetarianism that's based on sentience. If they are not, but they respond in the same way, then sentiocentrism should indeed be questioned.
When a Roomba responds to various stimuli, does that cause you wonder whether it's sentient?
I guess my point is that we still don't have any idea how to determine whether something is sentient. That includes other humans (though for that, we just apply Occam's Razor and call it a day).
I do think that in a few years, we'll have LLMs with plugins for memory and perception, executing not in a web-based repl but in real-time, and it's going to get harder and harder to argue that they're not sentient.
I had this experience, except in psychology class decades ago.
Phineas Gage had an accident: a steel rod was driven from his jaw to the top of his head, obliterating huge sections of his brain. And yet, it was a completely non-lethal attack. Despite losing significant brain matter, Phineas Gage immediately got up, was conscious and seemingly only lost eyesight in one eye.
But Gage's personality, mental state, and other attributes were permanently changed. The early science of psychology (well... I dunno if you could call it "science" yet...) took Gage's accident and began to map out the physical mechanization of our minds.
---------
Lobotomies and other such experiments would then map out which parts of the brain did what things. This was all figured out in the late 1800s and early 1900s. Cut this part of the brain out, the people can't talk anymore. Cut this part of the brain out, and you cannot transition short-term memory into medium term memory. Etc. etc.
The fleshy parts of our brain are... fleshy... mundane, and machine like. Its a machine that remains mysterious in many ways, but its easily experimented upon and we can discover more and more from it.
-----------
In any case, AI isn't much enlightening at all. AI isn't how human brains work anyway... no more than how an airplane wing is related to bird wings.
If you want to discover and understand the human mind, with the human errors and human fallacies, you should study humans. Medical science, neurology.
That being said: we don't actually think with our brains all the time. A lot of our emotional state comes from our blood, chemistry, lymph nodes and more. Overemphasis on the neural network will leave you blind to other psychological mechanisms.
Our fingertips contain a substantial number of neurons, as our nervous system spreads throughout our body. The delineation of "brain" vs "spinal cord" vs "nerves" is very ambiguous, it flows. Honestly, a big problem with ourselves is that our Hollywood / Story understanding of ourselves and our own brains is complete bollocks.
It would do you well to study psychology, at least the elementary levels, to understand yourself more. If you're worried about existentialism or trying to find an understanding of self, I don't think looking at machines or computers is very helpful. But maybe that's just me.
In any case, please don't look to movies or stories. These are shared experiences, yes, but full of ancient 100+ year old psychology information that's largely been debunked by today's science. (Again: the first hundred years of psychology can barely be called a "science" or "art". It was more pseudo-science bullshit than real science. But you sometimes got good records of people like Phineas Gage and other forbidden experiments by today's ethical standards).
Maybe our personality and being is just a result of our training data with a few genetic quirks thrown in. And what am I really doing when I “decide” or “create” something? Is it that much different from a LLM’s generation of language? Or an AI image generator?