The Cognitive Short-Circuit of 'Artificial Consciousness'
scienceandnonduality.com
scienceandnonduality.com
I think about a hard drive with a fresh version of Ubuntu installed to it. If an alien civilization found the hard drive, perhaps they'd be able to measure all of the physical properties of it. But they'd have no idea that an operating system was an emergent property of the drive because they didn't have an understanding of, at minimum, the following:
- Binary representation
- Processor architecture (including the specific processor for which this was compiled)
- Big-endian / little-endian representation
- Disk formats
- File systems (hell, the concept of files in general)
This is really the point the author is making. There is no reason to be certain that consciousness (the reflective, emotional, inner experience that is distinct from logical intelligence) emerges out of the hardware of human (and other animal) bodies and brains, as no empirical evidence for such a process has yet been found, after several decades and vast sums of money invested looking for it.
Just as an operating system is a separate entity that needs to be installed onto the computer for it to be able to work, it (in the view of the author) is plausible that consciousness is something distinct from the physical matter of the body and brain.
It's entirely possible that we could observe and measure consciousness if we knew what to look for.
Well yes, exactly, and Kastrup's point is that the conventional fixation on matter (which to date has proven fruitless) is a distraction from other possible answers.
Perhaps I am using the term "emergent property" incorrectly, but your argument does not seem to contradict my point. That the operating system exists because many programmers intentionally developed it while interacting with one another and society at large, and that the OS only functions when the drive is connected to a functioning computer is immaterial. Intelligent consciousness was crafted by evolution rather than human hands, and it requires the hardware of a human body to function (as far as we know) rather than a computer, but the result is the same, is it not? That is, under the right conditions, a unique property appears that was not readily apparent from the existing components.
Incidentally: I should have said, "...an emergent property of the particular arrangement of bits on a hard drive" to be more precise, but I think you knew what I meant.
> Just as an operating system is a separate entity that needs to be installed onto the computer for it to be able to work, it (in the view of the author) is plausible that consciousness is something distinct from the physical matter of the body and brain.
Ok, consider the example of just the core components of a computer; CPU and RAM, and the hardware for them to communicate. This can be thought of as a very complex circuit from which the emergent property of computation arises when interpreted correctly. Without an understanding of clock timing, the CPU's instruction set architecture, binary and many other concepts, the ability to measure the voltages of various components does not make the large operation or organization apparent.
Ultimately my point was that the author was incorrect in stating, "We can deduce the characteristics of all known emergent properties by examining their constituent elements", and I wanted to give an example of a system for which we would likely not understand the emergent property had we not built it ourselves. Although I agree with the author's point that we cannot be certain of the origin of consciousness, I disagree strongly with his certainty that it is not an emergent property.
Here's an new counter-argument: self driving cars are conscious. They're aware of their surroundings, in considerable detail and with useful understanding. They're not passive, waiting for questions; they have an ongoing stream of intellectual activity, so they have an "inner life". They learn from experience. They have goals, some of which are more important than others. They try to predict the future ("will that car ahead try to make a left turn?"). They interact with others (lane merging requires this.) It's not human-level consciousness, but it's consciousness.
So there.
The more interesting question might be: "What is the nature of our experience of consciousness?" and from there "What is the experience of a self driving car, if any, and at what degree of complication (or other metric, interconnectedness, etc) does this experience exist in a way familiar to us"
Similarly, "what is the experience of a rock?".
The author seems to assume that experiencing the world is somehow separate from the mechanisms of that experience. Just because he feels like a little homunculus doesn't mean he is. If we had nerve endings in our brains allowing us to feel the thoughts we were having separate from thinking, then I don't believe this artificial distinction would be as prevalent.
For me it was the opposite. It was one of the only redeeming qualities of the article.
>It's a false dichotomy to assume that consciousness and intelligence aren't the same thing, if one cannot meaningfully articulate how they differ.
I think the difference can be articulated clearly. Intelligence is the capacity to create structures and processes that are isomorphic to the world "out there". It's the capacity to grasp the world. Consciousness is the phenomenological space of experience. It might turn out that it's a false dichotomy but it's a logically valid distinction. It's much easier to logically talk about the former than the latter. We can build logically consistent models of the former while only providing hand wavey explanations of the latter.
Characterizing Turing's argument this way misses the mark: "We can test whether a machine is intelligent or not purely by observing its behavior in the environment."
Turing (and Ex Machina) argue that this is the only possible test, largely because arguments about consciousness dissolve into circular meaninglessness. Daniel Dennett dispenses with this best in his essay Quining Qualia: http://cogprints.org/254/1/quinqual.htm
True but for the same reason, models of consciousness dissolve into circular ontologies of subject and object. I agree with you on Spinoza. Seeing consciousness as a qualitative property of the universe avoids this problem and ties up alot of loose ends. That doesn't imply artificial consciousness is impossible. It implies on some level artificial consciousness will always be a blackbox system to us. One we can maybe replicate and improve on but still beyond full logical comprehension.
What is love? How can an uncaring chunk of bones and flesh feel love? But we do fall in love, so it must exist in some form. Hence, love is the universe itself. Checkmate
'Love' or 'justice' or any other x that would work in this context depend themselves on consciousness and therefore the only silly part of the resulting argument is that "x is the universe itself" only works if x is the broadest aspect of the concept under discussion - i.e. consciousness itself.
I'm not saying the argument isn't flawed. It's possibly a hand-waving, question-begging, non-explanation. But it's a natural place to reach if, like me, you think a Descartian "I have qualia therefore I am" can't be hand-waved away as easily as some claim.
This is why thought experiments about zombies and chinese rooms are interesting. It's about exploring what would be left out (if anything) once a full physical explanation had been achieved.
The usual metaphysically couched argument that consciousness is a primary substance and matter is secondary, ergo scientists/engineers pursuing a materialist explanation/synthesis of consciousness are fools barking up the wrong tree, all topped off with this <sarcasm>gem of a</sarcasm> conclusion:
"The computer engineer’s dream of birthing a conscious child into the world without the messiness and fragility of life is an infantile delusion; a confused, partial, distorted projection of archetypal images and drives. It is the expression of the male’s hidden aspiration for the female’s divine power of creation."
Do you see a contradiction? You're broadly calling a bunch of people "arrogant", who may have a very well informed and valid opinion on a topic.
Saying "I'm not arrogant enough to have a strong opinion either way" about any statement A is itself a pretty arrogant thing to say (you are making a statement about the collective knowledge of everyone else on the topic of A not being too far from yours).
Here's an example relating to your experience. I could claim: "I'm not arrogant enough to use RoR for serious web development".
On the surface, I'm making a claim about myself / my beliefs. You shouldn't feel upset over my statement, it's not about you or any other people...
But you're still not getting it, if you actually went back and read my replies and attempted to understand them maybe you would see, but you don't seem to have any interest in specifics, only in generalization. Still, I'll feed your troll on more time.
> Here's an example relating to your experience. I could claim: "I'm not arrogant enough to use RoR for serious web development".
> On the surface, I'm making a claim about myself / my beliefs. You shouldn't feel upset over my statement, it's not about you or any other people...
This is a fallacy to assume that just because I am attacking an idea, that no one will be offended. Of course someone can be offended if they have a belief that I think is arrogant, that is not in question, it's a phenomenal strawman you've set up there.
Also, it's absurd to equate the belief that it's arrogant to use some mundane tool to the belief that the nature of consciousness is outside the bounds of human knowledge. Is it the case that you are fully aware of the nature of consciousness, and I'm just an ignoramus who is clueless about the state of human knowledge? Okay, enlighten me then, I'm open to arguments about the nature of consciousness and why we be certain about it.
Your assertion boils down to the idea that it's arrogant to believe that any belief somebody else holds is itself arrogant, regardless of how preposterous that belief may be. Fine, you got me, go get yourself a cookie.
One of the more interesting papers I read on this topic involved neural nets with a set of neurons associated with 'pleasure' and 'pain' such that firing a neuron in either the 'pleasure' or 'pain' group would externally activate a strong enhancement signal for learning and bind in a positive or negative association. So that future things which triggered values in that learned net would then get modified by the enhancement signal in proportion to how strongly they correlated with the net. The math was very obtuse but as far as I could tell it boiled down to creating an out of band signal from the activation of specific neurons within the matrix which could influence the weights of future associations.
The pleasure/pain internal path is basically adding additional stimuli sources within the network itself. So you can modify the training just by re-running the previous training.
A made up example of how that might work is lets say you had a 'funny' neuron. Basically anytime its weight was adjusted by the network you got an immediate feedback on that association as 'funny'. Now you re-run a previous association that shares weights with the one that was tagged as 'funny' and that weight back propagates through the network giving a previously 'not associated with funny' association a proportional association.
What I got out of the paper was building neural networks that were pre-disposed, though an internally hard wired linkage of some neurons, to learning a particular form of association. The idea that an identical set of training values would train differently on nets based on this sort of internally forced messaging for some neurons, is the closest thing I've seen to the 'inner life' notion of consciousness.
Maybe this could be simulated by concurrently running the AI's existing models in feedback mode on internally generated stimuli, themselves sampled from those models, with some variations, but with limited ability to control the AI's actuators.
The biological analogy for this might be dreaming, where our neurons seem to engage in activity (hence experiences) which are similar to but also different than our waking mental models, and does so while we are in a state of self-induced paralysis - so we don't actually jump off a bridge.
We have very little in the way of evidence on the question of consciousness, so we can only really conjecture very simple ideas. The idea that everything is consciousness is a simple idea. Metabolism is not. What is even meant by metabolism? Creating the magic ATP spirit molecule?
My method consists of giving the person a basic education in a Artificial Intelligence or Philosophy of Mind and then asking them whether they believe that either consciousness can potentially be explained using concepts we already possess or that in fact it isn't even in need of an explanation.
If they answer yes, then they are probably a zombie.
Edit: I apologize for implying the natural world is superior. Hurrah for infrastructure, bureaucracy, and humanism!
How would this guys argue against this way of thinking about consciousness:
Michael Graziano, evolution and non-magic description of consciousness
https://www.youtube.com/watch?v=peHcu8LEgEE
Ah! I guess he simply questions, right out denies materialism. This is borderline religious, right?
This started out interesting, then went absurdely tentative, and finally turned into Sokal affair material with the ending (the male’s hidden aspiration for the female’s divine power of creation, lol).
I think there should be some variant of panpyschism that disguishes between the consciousness of a mammal and that of a rock.
But the way the author describes consciousness it sounds like a bad thing. Like a shared delusion. Maybe the robots are better without it.
> If that is the case, there is absolutely no sense in talking about creating consciousness, since consciousness is already there from the start. It is what there is. It can’t be created for it is that within which all creations unfold.
> If something is fundamentally beyond all forms of experience, direct or indirect, it might as well not exist. Because all knowledge resides in consciousness, we cannot know what is supposedly outside consciousness;
Just because there are limitations on our own ability to perceive the physical world does not mean that there is a fundamental limit on what can be perceived. For example, the Hubble space telescope can perceive light that is beyond both the ability of us to observe inside of our atmosphere and the wavelengths which we can see with our own eyes.
>According to my system, reality unfolds in one stream of subjectivity that I call ‘mind-at-large.’
>As such, what we call ‘conscious entities’ are merely dissociated alters of mind-at-large. An image of that dissociation is a human body.
>Therefore, our feeble attempts to engineer an entity with a private, subjective inner life similar to our own aren’t really attempts to create consciousness. Instead, they are attempts to induce dissociation in mind-at-large, so to create alters analogous to ourselves.
After creating a veil that is unpierceable by human cognition, the author conveniently hides his metaphysical "mind-at-large" creation behind it. It is as if some great mind is doling out consciousness to each animal according to its share. Because we are but fragments, mere images of this great mind we can never hope to be able to recreate anything more than cheap reproductions. Sound familiar?
> Going further down the chain of biological complexity, it isn’t unreasonable to infer that metabolism itself – that process common to all life – is the most basic image of dissociative processes in mind-at-large.
Engineers design chemical processes that harness chemical energy broken up into individual steps to produce some useful product, just like metabolism does. Indeed, chemical engineers use enzymes in some of their steps. In fact, there is even work on creating artificial metabolisms. We are better at replicating this mind-at-large better than the author is aware.
>Based on this understanding, do we have any reason whatsoever to believe that the mere mimicking of the information flow in human brains, no matter how accurate, will ever lead to a new dissociation of mind-at-large? The answer to this question can only be ‘yes’ if you think the kidney simulation can make the computer urinate.
False equivalence much? Creation of a physical substance from thin air is impossible. However, given the right inputs, the Cloaca machine can create poop from food by simulating the digestive process.
*edit - formatting