Penrose’s Philosophy- Consciousness and Computation
‘There is something going on in human understanding which is not computational, it’s different from what a computer does. So when people call it artificial intelligence, I think that’s a misnomer. Artificial cleverness perhaps.’ -Roger Penrose
What is meant exactly by ‘artificial intelligence’? Or better yet, what is meant by ‘intelligence’. To Roger Penrose, this term is what differentiates us from computers and why computers will never achieve consciousness.
It is a common assumption these days to say our brains are like computers, and to a certain extent they are. Both receive input. Both process information. So why is it that we seem to possess ‘consciousness’ and computers don’t? Penrose puts forth a striking case and it has to do with the relationship between consciousness and computation. But in order to better understand his insights, let us first define the terms.
While there is no universally agreed upon definition of what consciousness is, Penrose begins by explaining his own interpretation and the importance of the terms ‘understanding/awareness’ ‘intelligence’ and ‘consciousness’.
‘I do not think that it is wise, at this stage of understanding, to attempt to propose a precise definition of consciousness, but we can rely, to good measure, on our subjective impressions and intuitive common sense as to what the term means and when this property of consciousness is likely to be present… To be conscious, I seem to have to be conscious of something, perhaps a sensation…’
‘My own use of the term ‘understanding’ certainly implies that a genuine possession of this quality would require some element of awareness to be present. Without awareness of what some argument is all about, there can surely be no genuine understanding of that argument.’
‘It would also seem to me to be unexceptionable to use the word ‘intelligence’ only when there can be some understanding involved.’
Although these definitions can seem frustratingly vague, what Penrose is outlining here is this; consciousness requires understanding, understanding requires intelligence. In order for me to be conscious, I have to be aware of my own consciousness which requires my ability to understand the concept of awareness which, in turn, requires a certain level of intelligence. By skipping over the need to articulately present a detailed enthrallment of what each of these terms means, Penrose is able to quickly move on and address the primary focus of his argument which, as stated earlier, is the relationship between consciousness and computation. So let us move on, keeping in mind the outline of the terms we just discussed.
‘It seems to me that there are at least four different viewpoints- or extremes of viewpoint- that one may reasonably hold on the matter (of consciousness and computation).’
A. All thinking is computational; in particular, feelings of conscious awareness are evoked merely by the carrying out of appropriate computations.
B. Awareness is a feature of the brain’s physical action; and whereas any physical action can be simulated computationally, computational simulation cannot by itself evoke awareness.
C. Appropriate physical action of the brain evokes awareness, but this physical action cannot even be properly simulated computationally.
D. Awareness cannot be explained by physical, computational, or any other scientific terms.
A is where supporters of artificial intelligence, or strong AI as they are sometimes referred to as, reside. As with all viewpoints, there are varying degrees of what is accepted and what is not, but for the most part A defines consciousness as entirely computational. The famous Turing Test is an example of this. Say you have a robot that is controlled by a computer. When questioned, it answers entirely in the same manner as a human would. It can show feelings and emotions, it can learn, it can even contemplate questions of the ‘self’. So, can we say this robot is conscious or possesses consciousness? A would say yes, it behaves as if it is conscious, therefore it must be considered as such. If the successful imitation of human consciousness can be achieved by purely computational processes in a computer, then how can we deny the label of ‘conscious’ being given to such a computer? Critics, like Penrose, emphasis that just because a computer can behave as if it were conscious, that does not mean it possesses it. ‘Like playing a chess computer by only playing chess… you can play it to find out if it makes good moves but it won’t tell you if it knows it’s playing chess and it won’t tell you if it knows what chess is.’ If we remember back to our terms, this is where Penrose emphasized the importance of the relationship between consciousness, understanding, and intelligence must be adhered to. There is no way of proving the computer has a real understanding of its own awareness and according to Penrose, intelligence without understanding is a misnomer. More is needed in order for any sure confirmation can be made whether consciousness is present or not.
It is at this point I would like to interject with my own thoughts for a moment. This I feel is where Penrose’s background in science and physics tends to get him into hot water with some philosophers. Although Penrose himself rejects A (we will eventually get to what viewpoint he falls into), he is not saying it is entirely impossible. ‘The operational argument does, it seems to me, have some considerable force, even if it is not entirely conclusive…If all the external manifestations of a conscious brain, including responses to continual questioning, can indeed be completely imitated by a system entirely under computational control, then there would indeed be a plausible case for accepting that its internal manifestations- consciousness itself- should be also considered to be present in association with such a simulation.’. While philosophers tended to be contemplative of the more abstract questions pertaining to consciousness, Penrose is searching for a scientifically based solution that can be proven as to what consciousness is. It is for this reason that he cannot support A. As of yet, there is no scientific way to prove a computer understands consciousness or that it possesses it. It cannot prove its awareness of itself (keeping in mind Penrose’s terms). ‘While I do not myself believe A is correct, I do regard it as a serious possibility that is worthy of considerable attention.’. So while there can be endless philosophical debates about whether a computer that passes the Turing Test really is conscious, Roger Penrose wants to see the actual scientifically provable solution showing this. So the next question becomes, well how do we prove that then? We need to know how our own conscious processes works in order to be able to compare a computer’s to see if the results can be duplicated with the same outcome.
*This is a working paper. I plan on explaining the other three viewpoints in depth.
**My apologise my footnote citations did not transfer over when I copied and pasted. Works cited are below.
Works Cited
Garland, A. (Director). (2014). Ex Machina. Universal.
Penrose, R. (1989). The emperor’s new mind : concerning computers, minds, and the laws of physics. Oxford University Press.
Penrose, R. (1994). Shadows of the mind : a search for the missing science of consciousness. Oxford University Press.
"A. All thinking is computational; in particular, feelings of conscious awareness are evoked merely by the carrying out of appropriate computations.
ReplyDeleteB. Awareness is a feature of the brain’s physical action; and whereas any physical action can be simulated computationally, computational simulation cannot by itself evoke awareness.
C. Appropriate physical action of the brain evokes awareness, but this physical action cannot even be properly simulated computationally.
D. Awareness cannot be explained by physical, computational, or any other scientific terms."
The problem with any and all of these theses,seems to me, is that they beg the question as to what may ultimately emerge from machine learning. After all, we don't have much of a clue as to what triggered our ancestors' own conscious awareness of themselves as intentional agents in a world separable in thought from themselves. Do we really know for certain that a sophisticated machine with a syntax, or an ability to process symbols in a systematic and rule-governed way, could never acquire a semantics, or a contextual understanding of the meanings of those symbols when related to a world? A hypothetical observer of human evolution might well have said, of our forebears, "they'll never get beyond the stage of rote symbol manipulation." But how could they have known that? And how can we, when it comes to the early observation of evolving machine "cleverness"? Evolutionary processes are open-ended and unpredictable, so far as we can tell.
So I agree with you: "We need to know how our own conscious
processes works in order to be able to compare a computer’s to see if the results can be duplicated with the same outcome." And we don't know that. So...
Maybe we need to reverse our focus. Study computers in hopes of gaining insight into how our own minds work.
Actually, we need to remain focused on both ends of inquiry and we need to resist the temptation to declare resolution of these questions by high definition. We really don't know what machines may learn to do, or may become, any more than an ET observer could have known the human destiny by watching primitive proto-humans and extrapolating from their unsophisticated absence of self-knowledge.
This is fascinating stuff, and while I remain wary of Penrose I do sympathize with humanistic attempts to mark out a special preserve of human based on our subjective consciousness (see for instance Jaron Lanier's "You Are Not a Gadget"). But I also remain impressed by the criticisms of Pinker, Dennett, Carroll, et al. I look forward to seeing how you respond to (or otherwise acknowledge) them.
This helps me think better about life and everything around us
Delete