Robotics News, Knowledge & Insight since 2008

Can AI Ever Become Self-aware?

Introduction

The “hard problem of consciousness” has not been solved – there is still no explanation of how physical processes produce self-awareness, the “self”, the observing first person, what we call consciousness. This is not our opinion, it is a fact, as of the writing date of this article.

First let’s make clear the terms we mean in this article which we may use interchangeably but we mean the exact same thing with all:

Consciousness = self-awareness = the observing first person = the “entity” that experiences, chooses = the “self” = the “soul”.

Note that in the definitions above, by the first one, “consciousness”, we of course don’t mean reacting to physical input, like a photo sensor reacting to light, but what we explained with the equal signs above.

Our stance on the matter: AI will never be self-aware, consciousness is real, in other words, “the self” does exist, but it is something non-physical and it will never be demonstrated that physical processes will produce consciousness. Yes, AI will imitate consciousness very well, to the point of AGI and far beyond as ASI. We are not debating that and are looking forward to it, and to all the scientific progress, and we are all for science. In fact AI will imitate self-awareness so well that we will not be able to distinguish it at all. But imitation is one thing, and being the genuine self-aware observer is another. In the end, still atoms will be moving other atoms, none of which are conscious themselves or something that can produce it by simply arranging them in one form or the other.

We know that some people will disagree with various places in our article, and it is perfectly normal, because whether consciousness is physical or not is already a highly debated subject and many people believe that science will eventually figure it out and show that the self-awareness, the observing first person emerges as a result of physical activity in brain. Our goal is just to share our view, our thought exercise, with the hopes that it will be an interesting read for you at least. We are not trying to convince anyone. We didn’t do any research for this article, nor is the author a philosopher. Thoughts belong to the author, many of which must have already been discussed and debated for a long time. We also did brainstorming ideas back and forth with multiple AI (which is ironically not self-aware) and the article improved in the process while trying to convince it, so it was a fun thought exercise too at the same time actually.

Now let’s dive deeper into our stance.

The Brain

First let’s make it very clear… We of course know and accept the brain’s “relevance” to consciousness. The brain not only manages the automatic functions such as reflexes, body regulation, sleep, but also serves as the hardware, in other words, the means through which the observing first person functions as we know it. But this doesn’t mean the brain is the self-awareness, the observing first person, in other words the “self”. As we said in the beginning, no physical process has been described that produces “the self”.

Brain damage, which the physicalist view likes to give example of, disturbs regulation and expression, but not “the self” in our opinion. A person with a mental disability has the same kind of “self” as anyone else. Only the hardware is limited or malfunctioning which affects the function. Even when the damage changes how a person judges or behaves, the self still observes and chooses; the damaged hardware distorts what it receives and how its choice is carried out. Brain activity correlating with experience is also not a mechanism to “self”. Correlation doesn’t explain how one becomes the other. Explaining neural activity, is only explaining how the hardware works. It does nothing to explain how and why those neural activities must be the observing first person entity in the first place.

“Building Self-awareness”

Just assume for a moment that we somehow figure out how to design self-awareness, “the self” on paper, and draw its design on a CAD file etc… Then continue to assume that we also somehow build it 99.9999999% physically and then plug in the last wire. Without that wire, no self-awareness. With it plugged in, suddenly an inner “self” just emerges? Let’s call this “the last wire” question.

Or we build it 100% but don’t place the battery and it is off, and then we place the battery and bam, we got a self-aware entity? By that battery moving atoms and turning transistors on and off? – or whatever more advanced technology you can imagine which must still ultimately be a form of (mathematical) calculation? And sudden or gradual does not matter; the point is how. Electrons merely start flowing, and unconscious calculations, matrix multiplications or whatever the form of calculation we will have with our future advanced tech happen. Where is the consciousness here? Nothing we can do in the physical world will produce that in our opinion. Experience, the observing entity, is a different kind of thing from physical arrangement. No amount of structure bridges that gap.

Plus, self-awareness is binary: you have it or you don’t. There is no partial. Here the physicalists will argue that consciousness “emerges” (somehow of course without the slightest description whatsoever so far) “gradually” as complexity increases, rather than appearing suddenly with the last wire. But that argument has a fundamental flaw: Every gradual addition still needs an explanation for that “partial self”, “partial awareness”. If any stage has even the slightest subjective experience, it is already conscious, self-aware, when it wasn’t a moment ago. In other words, every step is also binary. If a step has no consciousness, how does the next physical change produce subjective experience? Every degree of genuine self experience is still consciousness, a binary step. Experience may vary in richness, but having it at all is a yes or no. And what about the 100% complete machine without the battery and with the battery – what does that do to the gradual emergence claim? Physicalism may say that power-on is itself a process with many steps, so emergence could still be gradual but the binary argument above already answers this, since each step of the start-up either has experience or doesn’t.

If arrangement can’t produce the experience, such as the observing first person in an AI, your brain can’t be the source of your inner self either. Look at it also from this angle: Even with the tech of 1000 years from now when we build a brain that is 1000 times better than ours in every way, this will not change because that brain will also be just an arrangement of nonconscious matter and some algorithms, rules that it runs with.

Physical descriptions tell us what matter consists of, how it is arranged, and how it behaves. None of these descriptions explains why an experiencing subject should exist at all. The problem is not merely a missing technical detail, but the apparent categorical difference between physical processes and subjective experience. The problem is fundamental. Categorical. Not just an engineering or scientific challenge in our opinion.

Flipping Burden of Proof

The materialist, physicalist view claims that either our consciousness is an illusion or it is real but, even if it hasn’t been explained yet, in the future we will explain how physical processes in the brain produce consciousness. In this sense it remains faith, the very thing materialism criticizes believers for: believing in something with faith, such as God or whatever belief someone may have, without being able to physically show it.

As for calling consciousness an illusion: an illusion still has to appear to someone. If there is an appearance, there is someone who has it. Calling it an illusion does not remove the observer; it only gives the problem a different name.

Expecting future science to explain consciousness entirely through physical processes when we still don’t have the slightest hint for it, is only hope. Faith. Those who confidently claim that consciousness is entirely physical carry the burden of proving, describing how physical processes produce subjective experience. And no, brain dysfunctions don’t explain it, it is only the hardware as mentioned. It doesn’t prove how or why the brain activities must be the observing entity.

So now the burden flips. Anyone who defends the physicalist view and asks “prove your (nonphysical) faith / prove God exists etc..” must now prove that consciousness is physical, by showing how a physical process becomes a deciding, experiencing, observing self. And for the physicalist view, the chance of winning is getting narrower with each passing year, as our science and technology advance at an accelerating rate, and we not only learn more about the brain and still have nothing for describing “the self”, but also we build increasingly more complex and capable AI, which is also still not self-aware and shows nothing beyond imitation, even if it is / will be AGI or ASI.

Various challenges and answers:

“Your view names the self but doesn’t describe it”

A critic may say that calling the self non-physical only gives it a label, without saying what it is or how it works. But “how does it work” asks for a mechanism, and mechanisms belong to physical things. We do not owe a physical description of something we claim is not physical. Nor is the self undescribed. It is the one who observes. It is the one who tells good from bad and chooses between them. With 100 dollars in my hand, I can spend it, save it, waste it or give it to the poor, and a minute later I can choose differently. Everyone knows this entity, in other words their self, directly, more directly than they know anything in the physical world and looking at it this way, it is the thing that needs the least description. I can be wrong about what my brain is doing, and a scan or a theory may surprise me. But I cannot be wrong that I am experiencing something right now.

“Where does the self come from?”

Our claim does not depend on the origin of the self. Consciousness is here, whatever its source. If asked about its source, then we can say the question cuts both ways: the physicalist cannot say where the universe and the matter in it comes from, where its laws come from, or where the rules that evolution follows come from, the rules that, according to their view, make the physical self possible. So, if an unexplained origin does not sink physicalism, it does not sink our view either.

And if the physicalist answers that the universe itself always existed and needs no source, then the question becomes: is that eternal universe conscious or not? If it is not, it cannot be the source of consciousness, as this article argues. If it is, then its consciousness must be shown physically, and it has not been.

The physicalist may then say that consciousness came from the unconscious universe through evolution. If a blind, unconscious process like evolution could produce consciousness without even trying, even if it had billions of years, then we, intelligent designers who are actively, selectively trying with all our knowledge, should be able to do it too, at least have some ideas. Yet we have no idea how, not even a first step.

“You also say ‘it is just a basic feature of reality’”

A critic may say that when we call the link between self and brain a basic feature of reality, we make the same move we criticize as faith. It is not the same move. Physicalism claims that everything has a physical explanation. When a physicalist says experience simply comes with brain activity and nothing more can be said, he breaks his own rule. We never made that promise.

“Life used to seem mysterious too.”

Physicalism will say people once thought life needed a special non-physical force, and it turned out to be chemistry. But life was about what bodies do, while consciousness is that someone is there as observing first person. Explaining what the brain does still leaves that unexplained.

“How do the self and brain interact, link?”

Physicalism may ask how a non-physical self interacts with a physical brain. We don’t have to answer that. The question assumes physicality because it demands a physical explanation, and we do not owe a physical proof for a claim that is non-physical in the first place. We state openly that the link between self and brain is a basic feature of reality, not something with a mechanism, so the absence of a physical mechanism is exactly what our view expects. The one who owes proof is physicalism.

A critic may say: “If your self makes your hand rise, then at some point neurons fire because of something non-physical. It should show up as neurons breaking physical laws, and we don’t see that.” But this objection assumes that every physical event must have a fully physical cause, so that physics is a closed system. That’s not something anyone has measured. Nobody has traced a free choice through the brain and shown that physics alone accounts for every neuron. It’s an assumption that already takes the physicalist side. So the objection doesn’t refute us; it assumes what it needs to prove.

The self may be a kind of energy, but not energy in the physics sense because it is also able to observe and choose as first person and has the ability to be “self”; it is independent of the body, is not measured by physics, and since it is not physical, our logic sees no reason that it doesn’t continue after the body dies.

“Your case rests only on physicalism’s failure”

A critic may say that showing physicalists have not explained consciousness does not prove it is non-physical. That would be true if their failure were all we had. But this whole article makes a positive case.

“Some consciousness in matter”

Another view is that there is “some” consciousness in matter, and that bringing it together produces our kind of consciousness. Does that mean atoms have experience? Molecules? Electrons? If electrons, would it differ between atoms with different numbers of electrons? Why or how would an electron or proton or any part of matter possess consciousness at all? How do they make consciousness if we bring them together in what form? There seems to be not the slightest path. What about quarks? Again, how? And even if they had self-awareness, then describe not only how it works, but also describe how we must bring them together to produce self-awareness as we experience. Plus, consciousness is binary as we discussed above. A “little consciousness” can’t even be described. Saying “elementary particles possess extremely primitive forms of experience rather than human-like consciousness” doesn’t explain it. Even the most primitive genuine experience is still experience. Calling it primitive does not explain why a particle should possess it, or how separate experiences could combine into a single bigger experiencing self.

We measure matter with extreme precision and find mass, charge and spin, but not the slightest trace of experience. And even if every particle had some experience, billions of separate tiny experiences would not add up to the one “I” that observes and chooses; they would simply be billions of separate experiencers.

“If the brain is matter too, why can’t a machine have a self?”

A critic may ask: if the human brain is also matter and still has a self, why can’t a machine? We already showed that no physical arrangement can produce self-awareness, and a machine is in the same category as all other matter. Asking why a machine has no consciousness is like asking why a pencil has none. But since the brain is matter too, the question still deserves an answer.

Consciousness is non-physical, and I did not create my own. It was given to me. So my consciousness needed a source, and the source must exist. This follows by logic, not religion. (Regressing infinitely, we also need a source that doesn’t need a source, which is beside the point here but is at the foundation.) That source enabled my consciousness on a physical hardware that can run it and enable its function. It also made that hardware possible and linked my consciousness with it, so that I can function. And it made it that way only in living beings that can function, not in a dead human, for example. In addition, the source who gave our non physical consciousness must already have immense knowledge and control over our body and consciousness, to be able to link our non physical consciousness to our physical body which adds to the coherency of our argument, whether a critic may agree or disagree.

A machine is different. It is something I make, and I have no idea how to give consciousness to it. I have no reason to expect the source to give consciousness to a machine I made, any more than to the chair I am sitting on or the stone on the ground.

Conclusion

Our conclusion is that AI will never be self-aware. As time goes by and technology and science advance, we believe that this conclusion will withstand the test of time, as the problem is not an engineering or scientific problem but categorical, fundamental in our opinion.

A qualification: As a personal belief only, and beyond the argument presented here, the author may also accommodate the possibility of a physical self involving currently unknown physics, ONLY IF it is shown to exist independently of the body and to survive bodily death as this article describes the self. This article sets that unknown physics possibility aside and argues for a nonphysical self.

We presented our view. Some of you may not agree and that is perfectly fine, our goal was not to convince anyone but just to share our view, our thought exercise with you. We didn’t do any search and presented only the ones that we constructed/came up with. Whether you agreed or not, we hope that you found this interesting. Although it may feel a little off topic for our robotics site, it is still related to AI and we hope that it was a valuable addition.