Subscribe to:
Post Comments (Atom)
-
Harnad, S. (2003b) Categorical Perception . Encyclopedia of Cognitive Science . Nature Publishing Group. Macmillan. http://www.youtube.com...
-
(10a. Comment Overflow) (50+)
-
Psychology PSYC 538, Winter 2018: Categorization, Communication and Consciousness Time : Tuesdays 2:35-5:25 Place : STBIO S3/3 ...
I thought of cognitive neuroscience reading the article on physical symbol system. Cognitive neuroscientists precisely define the functions of neural circuits in order to predict human behavior, or even simulate it on a machine. An example of application is the intelligent prosthetics. I wonder if a low level of abstraction crucial to model human intelligence? The more details we know about the biological system of humans, the more precisely we could model it and simulate its behaviors. With a simple task such as the delivery man example mentioned in the article, higher level of abstraction of symbol representations might be sufficient to generate accurate outputs, but this representation may not generalize to other tasks. In the case of artificial intelligence, is the range of abilities it has dependent on the level of abstraction? Are both knowledge level and symbol level crucial to biological and computational entities, because without knowing the reasoning strategies of an agent, the replicable outputs to be regenerated must be simple and limited?
ReplyDeleteWhat is a Turing Machine?
ReplyDelete“It is a remarkable fact that none of these computers can outdo a Turing machine. Despite the Turing machine's austere simplicity, it is capable of computing anything that any computer on the market can compute”
I was not aware of this. My limited knowledge of the Turing machine saw it as the first step in computer creation, assuming that it’s followers evolved to do more but to learn that the Turing machine still remains as powerful as any other computer out there speaks to what a remarkable creation it was.
“But since a Turing machine is an idealised device, it has no real-world constraints on its speed of operation.”
Perhaps I’m not grasping the concept well enough but I’m a little confused as to how to comprehend the Turing machine. If it is “idealized” it means it’s not physical? Therefore, not tangible? Then how is it practically used? Is it something just to imagine in the mind?
Representations
“Computers and human minds are examples of physical symbol systems”
Isn’t it still up for debate whether the human mind is “physical”? Computers are tangible but is the human mind likewise?
“The term physical is used, because symbols in a physical symbol system are physical objects that are part of the real world, even though they may be internal to computers and brains. They may also need to physically affect action or motor control.”
So the term does not describe the computer or human mind itself as physical but the symbols that they use as such. Glad there was an explanation because wording can cause ambiguity or misunderstanding.
“Much of AI rests on the physical symbol system hypothesis of Newell and Simon (1976):
A physical symbol system has the necessary and sufficient means for general intelligent action. “
What does “general” mean? Does it allude to specific actions or tasks that are markers of “intelligence”?
“It means that any intelligent agent is necessarily a physical symbol system. It also means that a physical symbol system is all that is needed for intelligent action”
I don’t think the hypothesis is that sweeping because the way I understood, it says that a physical symbol system CAN produce general intelligent action but does not HAVE to. Thus, it is not saying that it is the only means for general intelligent action, it does not certify it as the sole producer but only says that it can be a producer of general intelligent action.
“Although no level of description is more important than any other, we conjecture that you do not have to emulate every level of a human to build an AI agent but rather you can emulate the higher levels and build them on the foundation of modern computers.”
So what does the “higher levels” consist of? Both knowledge and symbol since neither is deemed more or less important? Since the symbol level is “about what goes on inside an agent to reason about the external world” and it is the higher levels of the human mind trying to be emulated, we come back to what remains one of the main questions in neuroscience: how does the human mind work/reason? Sure we have modern computers with amazing capabilities but it is difficult to “emulate” something we still have so much to learn about.
What is Computation?
ReplyDeleteThe reading outlines the notion of ‘behavioural equivalence’ which suggests that as long as the right answer is generated as the final output, the person/system is considered to have solved the problem - regardless of what procedure they used. This means that we can substitute different computational systems for others, as long as they are behaviourally equivalent and generate the same correct answer. This section got me thinking about how much behaviourally equivalent systems can then vary in their most reduced and basic computational steps, if they must still produce the same and correct output. Is there a minimum finite number of steps that ALL systems can be reduced to, to produce ultimately the same output? In humans, do we all use different computational steps to produce our reliable outputs, or do these steps have an underlying equivalency between each of us? This ultimately leads me to ask how much computational variability can exist between behaviourally equivalent systems/persons?
“Simulation also raises questions about personal identity. There’s a long tradition in Western culture of identifying the self with one’s thoughts. But if our brains, and thus our thoughts, can be simulated, to what extent does that mean that we ourselves can be simulated? […] As artificial agents become more life-like, will we start to view them as real people?”
ReplyDeleteI have some issues with the concept of identifying the self with one’s thoughts. We most definitely tend to believe that who we are has something to do with the thoughts that we have, but I think this interpretation of personal identity is too narrow and omits crucial aspects of our lives that we use to define ourselves. For instance, part of someone’s unique identity has to do with their emotional responses, or the way in which they react to situations. When faced with a dilemma, some people’s immediate reaction is to get angry, while some get stressed, and others face these situations calmly. These immediate reactions are not always preceded by conscious thought and deliberation. We all know people that are short-fused or always put together, and these aspects of their identity are not reducible to their thoughts. Rather, they seem to be products of unconscious processes. Moreover, even people with salient personality traits can surprise us and act in unexpected ways or change their approach to situations. Therefore, I do not think that programming an artificial agent with human thoughts would give it an identity. In order to give an artificial agent an identity, we would have to identify all the aspects that make up identity (conscious thoughts as well as unconscious processes), program them into the agent, but also give the agent the possibility to act unexpectedly or evolve, which may just be a task that is too complex to execute.
This comment has been removed by the author.
ReplyDeleteGood afternoon, Dr. Harnad, I hope you are doing very well,
ReplyDeleteI have just registered for this course, and look forward to spending the semester in your course.
I apologize if my skywriting is not as elaborate as you’d wish for it to be, given that I am unfamiliar with it; however, I wish to improve them as I get accustomed to them.
Thank you very much, and have a blessed day.
My thoughts: from my understanding of what a Turing machine is, and what it is able to perform, the Turing machine can be used to understand the limits of what is computable, meaning that anything that can be done on a computer can be down equivalently on a Turing machine. This machine permits a model so a model for understanding numbers, functions, and computers.
Everything one does on a computer can be described by an algorithm (series of steps taken in order to calculate a number or create something from a set of inputs). Like solving a sudoku puzzle, a rubik's cube, brushing your teeth, commuting to work in the morning. It would then be logical to think of algorithm as a computation. The solution to many things can be computed. A few examples I am able to think of are, say a Rubik’s cube. These computations can be transferred over to computers. A second example I can think of would be Google Maps, which is able to compute which route would be the quickest and more efficient depending on my current location. Computers also possess programs capable of solving Sudoko puzzles or playing chess against. Based on my understanding, these are all computations. Given that the Turing machine is one that computes based on the input it receives, and per the reading, it is able to perform operations just like a marketable computer is able to do, if not more (given that there are no real-life constraints on the speed of operation of the Turing Machine, as eloquently noted in the reading ‘’What is a Turing Machine.’’). In other words, anything that is computable is not only able to be processed by this machine, but also can be described and stimulated. From my understanding of this, I would thus assume that a Turing machine capable of computing any and all algorithms could be built. I could further apply the notion of algorithm to one’s daily life and routine. For instance, the routes we take to go to university, or home, or the way in which we get dressed in the morning or make our morning coffee. If one was to write down a series of steps for each of these computations, then a Turing machine would be able to process and compute them.
This comment has been removed by the author.
ReplyDeleteOne of the things I found most interesting about these three readings was the lack of the term “consciousness”. Specifically, as it pertains to “What is Computation?” I found the boundaries drawn between the terms “mental operations” and “thought” to be poorly defined. The paper said that thoughts were traditionally believed to be what separated animals and humans. However, if “thought” is just a mental operation, and computers are capable of mental operations, does this mean that computers are conscious or might one day achieve consciousness? Is it possible for something that has been believed for decades to be unique to humans can be encoded in binary code for a computer to process? The paper goes on to say that anything can be encoded using binary; text, pictures sound and programs. With examples given such as computational chemistry and biology, it seems likely that if given the proper programming, a computer can surpass human intellect. Is this equivalent to them being conscious? Taking into account the idea of behavioral equivalence, is consciousness relevant to problem solving if both a human and computer arrive at the same answer? I think that this question also brings up interesting moral and ethical implications. For example, is it possible to give a computer a moral code equivalent to a humans’? The imitation game also illustrates an interesting perspective on the concept of behavioral equivalence. The paper dances around using the word “consciousness” by replacing it with “intelligence” but the same concept can be applied to both. If a computer is able to fool a human into thinking that it was human, is that computer equipped with a program for a conscious? And if consciousness is irrelevant, is there some other innately and uniquely human trait to separate us from computers? Reiterating again that computation is an idea in flux, I think that factoring the notion of consciousness into the definition may help to give a clearer shape to the construct.
ReplyDelete“In Western culture, we tend to take our capacity for thought as the central distinction between ourselves and other animals […] we view our specific thoughts and feelings as being one of the major constituents of our personal identity. So thought is constitutive both of our collective humanity and of our individual identities.”
ReplyDelete“Simulation also raises questions about personal identity. There’s a long tradition in Western culture of identifying the self with one’s thoughts. But if our brains, and thus our thoughts, can be simulated, to what extent does that mean we ourselves can be simulated?”
A few things cross my mind when I see these two quotations, then a few more when I consider them together. Firstly, I find it interesting that the first quotation appears to differentiate between thoughts and feelings, then quickly lump them back together by saying “as being one of the major constituents”. The rest of the article then goes on to explore computationalism (thought/cognition = computation), without addressing feelings again. Is the author suggesting the two are one and the same or are they saying feelings as a factor doesn’t matter for the sake of the query (whether or not cognition can be simulated and what that would mean)? It would seem to me that yes, thought is important for our “collective humanity” and “individual identities” but that feelings are separate and vital in their own right – not to be lumped in with “thought” as one thing when considering humanity or personality. It is then strange to me, to imply that a simulated brain, thus simulated “thought” should raise any questions at all about personal identity, because that would suggest either that feelings are equivalent to thought, that feelings don’t matter in this consideration, or that the author is simply not addressing feelings. I understand that a simulation of a thing is not the thing itself – but if that’s what we’re talking about why even consider “collective humanity” or “personal identity”?
I found these readings especially interesting because of how simplistic the actual Turing Machine is while it is able to compute even the most complex of equations. That said, I found it surprising that in the "What is a Turing Machine?" reading, the author notes that not all real numbers are computable. Shouldn't a number be in the realm of possibilities of things which a Turing Machine can compute if it is in fact real? Also, if a Turing Machine is able to compute anything that more complex computers are capable of, what is the benefit of creating more complex computers?
ReplyDelete“That is, an entity is intelligent because it behaves intelligently, not because it’s composed of some special substance such as living tissue or a soul. Moreover, intelligence is a computational phenomenon, amenable to computational analysis.”
ReplyDeleteThe argument that intelligence is a computational phenomenon, agreeable to computational analysis seems obvious when thinking of the types of intelligence that have been discussed thus far into the paper, however it is often considered – especially in psychology (think Gardener) – that there are numerous forms of ‘intelligence’ that can be expressed. For example, one can be argued to have ‘artistic intelligence’ or ‘interpersonal intelligence’ or ‘intrapersonal intelligence’. Take intrapersonal intelligence for example, this is an intelligence thus far only associated with humans – the notion of introspection and self-reflective capacities (or knowing about one’s own feelings). If intelligence is argued to be a computational phenomenon, and considering that intrapersonal intelligence is in fact ‘intelligence’ then how can this be? If the mere ability to know what it feels like to feel (cognition) is arguably one thing that suggests that cognition is not computation, how can intrapersonal intelligence exist – or does it not at all?
This comment has been removed by the author.
ReplyDeleteI apologize for doing these responses late because I don’t get the luxury of revisiting them to observe growth/learning, and the reality is I don’t know how I would have responded to the question of “what is computation?” three weeks ago. A small advantage I do have, perhaps, is that three classes in, we’re still dwelling on/revisiting that question, and it’s been impressed on me just how important is to understand the answer. Computation involves physical symbol manipulation based on a set of rules, and although the output could be interpretable (it’s really only useful if it is), the manipulation is based solely on syntax/what the squiggles look like. In short, it could be described as what mathematicians do. That’s a rather formal definition using the condensed bullet points provided to us, but does that help me understand why we keep revisiting the question? Not yet. It was said to us that to fully appreciate the limitations of computation, we must first understand its power (my paraphrase). What’s so alluring about computation? The first idea that comes to mind is that it could simulate just about anything. If you could simulate a neuron, you could simulate it firing, connect it to other neurons, and build your way up to a human brain. Who’s to say you haven’t simulated consciousness? Have you learned anything though? Could you explain where consciousness comes from by building up the brain up, atom by atom? I think what we’re trying to attain is an understanding of consciousness, rather than achievement (and we probably wouldn’t be able to attain the latter without the former anyway). Back to the simulation. It’s important to remember that these simulations are all occurring symbolically. I’ve mentioned this in class, but I believe that Hollywood has gotten the better of us in terms of visualizing simulations (especially Netflix’s Black Mirror). We imagine that there are these little people running around in the program, feeling and experiencing things, and it’s all rather homuncular. Although you may get a textual (or even visual) representation describing the output of a synaptic transmission, the only thing that is actually occurring (on a low level, assuming you run the simulation on computers as we know them) is 0s and 1s being converted into other 0s and 1s. The way I understand it, nothing actually happens independent of US being capable of interpreting it. We believe a simulated neuron fired because we interpret the squiggles as representing the neuron having fired. There are some uncomputable things that demonstrate the limitations of computation, but the real objection to computation probably lies in semantics - meaning.
ReplyDeleteIs the ability to simulate any known computer a byproduct of having Turing completeness or is it a requirement for Turing completeness? On page 12 it is noted that “[Turing machines] can compute anything those computers can compute” but then later, within the same paragraph it’s also stated that Turing completeness is the “property of being able to simulate Turing machines, and therefore being able to simulate any known computer”. I’m confused by this, because unless I’m misunderstanding the text, this is suggesting that the Turing machine, and any other machine with Turing completeness must be able to at the very least simulate anything a computer can do? Surely a computer can do more than a Turing machine?
ReplyDeleteI think you understood correctly. The Turing machine is hypothetical device that's not constrained by finite memory, time, etc. which a regular computer is (i.e. Turing machine can have infinite memory and states). A Turing machine can compute anything computable. And if (Strong C-T thesis) it can simulate almost anything in the universe, that certainly encompasses anything a computer can do. A Turing complete thing means if something is computable, then you can use that thing to compute it. Turing completeness can be found even in programming languages (they take any program written by us as input and can give you an output. Not considering the amount of time it takes to execute the algorithm, if you can compute it using one Turing complete language, it doesn’t matter which Turing complete language you choose, they’ll all get you the answer), but the storage/time limitations of computers mean they’re not really Turing complete. So Turing machines can simulate computers, but computers can’t really fully simulate Turing machines. If you're still confused, think about what a computer might be able to do that a Turing machine can’t. A computer runs on the same basic principles as a Turing machine (read, write, move, change state, etc). There's nothing going on in computers that's not computation.
DeleteSo fully complete Turing machines have more simulating power than computers? I had it backwards before but that makes a lot more sense now, thanks!
DeleteThis comment has been removed by the author.
ReplyDeleteWhat is a Turing Machine?
ReplyDelete“Commercially available computers are hard-wired to perform primitive operations considerably more sophisticated than those of a Turing machine--add, multiply, decrement, store-at-address, branch, and so forth”
Are commercially available computers considered Turing machine since it mentions that they are capable of more sophisticated operations, are we to assume their functions covers what a Turing machine can do? It is also mentioned commercial computer cannot outdo Turing machine. In what way? Also, does a Turing machine exist in reality?
“But since a Turing machine is an idealised device, it has no real-world constraints on its speed of operation.”
If Turing machine is not a device based on reality, what real world implications can we really draw from it?
What is Computation?
The author defines computation as a kind of question answering (behavioural equivalence) not related to the procedure but whether it works and can come up with the right desired answer.
The author does not provide any clear or satisfactory definition for computation by stating that “it is what computers do” in terms of his pragmatic definition. But to his credit, he elaborates that computation does not necessarily have to do with numbers but rather the process of producing some desired outcome without any bias towards the procedure it is implemented through.
"What is a Turing Machine" I thought the proposition that a Turing Machine can do more than any physical computer was an interesting concept to play with. We are introduced to the Turing Machine as the almost the most simplified model for a computational device possible. The Turing machine is described as a simple input/output machine with only a handful of fundamental operations. Despite the Turing Machine having such a simple design its unbounded memory capacity and unlimited speed of operation allows it to exceed the processes of any real computer. I think Turing uses this example to strip down a computational device to the bare bone elements that are required of a computational device in order for it to meet the high bar that is set by the idealized Turing Machine.
ReplyDelete“knowledge is encoded by a system of symbolic codes, which themselves are physically realized…it is the physical properties of the codes that cause the behaviors in question”
ReplyDeleteWhen I read this as a student with mainly a neuro-psychological background, I begin to think of ways a symbol system can be physically realized in an organic brain, and how this code can have a causal relationship with the organism’s behavior. The easiest way for me to conceptualize this relationship would be to think of the symbolic system acting on functional nodes in an animal’s brain. This node would consist of a vast number of neurons that all deal with a particular stimulus the animal has encountered, if this is in fact how the brain is organized (in a domain specific manner). The knowledge gathered by the animal could be encoded (symbolically) through plasticity in the connections between neurons at the synapse, constantly updating the overall system and “state” of the symbolic system as a function of physical interaction or thought of the stimulus. As is stated, if this neural node (the physical manifestation of the referent) was activated, the system could trigger a behavioral output. For example, if the node were wired to the amygdala (a stereotypical “fear centre”) perhaps the animal would exhibit a fear response, or a joyous one if it were wired to dopamine centres etc…
This comment has been removed by the author.
ReplyDeleteThis was an introduction for myself to the Turing Machine. Although I’ve came across a couple of examples in some of my classes; I’ve never went in depth with what it actually does and what its implications are. I find it fascinating that if a machine passes the Turing test then it is indistinguishable from a human in forms of imitating artificial intelligence.
ReplyDelete“Imitation of unobservable internal processes”
Since computation has helped us understand internal processes of human cognition and brain I think that a computer that passes the Turing test is important in order to have more plausible insight on how the brain and cognition mechanisms work.