| Free Will |
[Jun. 17th, 2009|11:35 am] |
I have only heard a single plausible defense of (non-deterministic) free will, and it went something like this: We cannot say in certainty whether the causes of our thoughts/our "will" fully determine what we choose to do I would normally assume that free will is something along the lines of "our choices cannot be explained [via causal reasoning] by deterministic causes." However, the above framing of "free will" does not make or entail this claim. Now, as tempting as it it to argue with my framing of the stance for free will, it is important to realize that doing so would be a fallacy -- I'd be arguing against a straw man.
In other words, regardless of how I frame the free will perspective, the question still remains: "is there anything metaphysically special about choices made by humans?"
If I am to address the above stance on free will with this in mind, my next question is: does this mean that determinism is false? In other words, if we cannot say for sure whether our choices are determined does that in any way imply that they do not have natural causes? Indeed, because the stance differentiated between causation and determination, it is entirely consistent with the particular framing of determinism that is something like: All events have one or more natural causes. This may be called "weak determinism" -- and, I would say, is an axiom of all experience, really, not something that is even questionable.
However, harking back to the issue of straw men, it is important to note that framing determinism like this doesn't prove anything. Therefore, I will note that "strong" determinism states: All events (including human cognition) are inevitable occurrences of and causally determined by one or more (antecedent) natural causes. Combining this with the argument at the start of this post, it seems to me that the primary question then becomes: are there any examples of human choices/intentions that cannot be fully explained (i.e., determined) by natural causes? If this is the primary question, then it seems evident that it is an empirical (in particular, psychological) question, and one that cannot be answered for certain without finding such an example. If one considers the area of psychology, and, in particular, statistical analysis, human actions can certainly be (and are) attributed to natural causes. However, assuming the proponent of free will finds this insufficient for "strong" determinism (i.e., because you cannot find all causes of every human action and have them hold true 100% of the time), then their theory becomes an unfalsifiable and, thus, practically useless empirical theory, kind of like Freudian psychology.
In conclusion: Is it true? I have no idea. Does it matter? Not in the least bit.
|
|
|
| The Future of Technology: Multimodal Input? |
[Apr. 17th, 2009|10:07 am] |
If the mobile phone is the future of computing and information access, that isn't very exciting. It's so difficult to find what you want on a mobile phone, right away when you want it. It isn't even very convenient to twitter on your phone, given how small the keys are (or non-tactile if you have an iPhone).
Multimodal input is one way to address this problem for a variety of devices, not just mobile. This includes, but may not be limited to, input via: speech, hand gestures, touch screens, internal accelerometers (for moving the device itself), eye tracking, sketching with a digital pen, face recognition, tangible interfaces, haptic feedback, and camera/computer vision-based input.
The interesting issue with multimodal input is how to fuse these types of input. Pointing with your finger can provide context for speech recognized as "put that there"; lip-reading can increase efficiency of speech recognition; voice and face recognition can both be used for identity confirmation; eye gaze can center a field-of-view for easier mouse clicking.
Multimodal input fits into a larger movement in both AI and HCI called "intelligent interfaces." What are intelligent interfaces? Well, there seems to be some disagreement. Some consider any interface for an "intelligent" system to be an intelligent interface. Others think that the interface has to behave in a manner that we normally assign to intelligent behavior, such as personalization. Of course, it's not clear whether or not these are the same thing (consider, e.g., Google's personalized search -- it's personalization, but is it really a part of the internal system or just the interface?). Then there are multimodal interfaces. They require "intelligence" in that there is a significant amount of pattern recognition and processing necessary to recognize, e.g., the user's speech.
|
|
|
| "The Chip" is manifested as mobile phones |
[Feb. 27th, 2009|10:17 am] |
The future is here. But where is my damn jet pack? my flying car? virtual reality?
Where is the The Chip?
A story featured in Wired outlines how the Wii, the iPhone and Guitar Hero exemplify "a blend of the digital and physical [reality]." In other words, for quite a while, society has considered what it would be like to live in a "virtual reality" -- the Matrix, the Wired (in the anime "Lain"), etc. -- but our virtual reality is actually melding with our "actual" reality. Technology that we foresee is manifested in a way that we did not expect, but it's still there. In the words of author Williams Gibson, author of Neuromancer, "The future is here. It's just not widely distributed yet."
In much the same way, though in 2003 I fantasized about the concept of a microchip inside ones brain that hooks up the Internet and so on, I think it is prudent to point out that what I was talking about (though I did not know it) is today's (and tomorrow's) mobile phone.
Consider the features I predicted: input with your brain, the ability to connect with other devices like an entertainment system, communicating and file sharing with other humans directly, accessing news and other information sources seamlessly, ubiquitous Internet access, having ones own personal web site. Almost all of these are available in one form or another on modern mobile phones:- Input with your brain -- Though this is not (yet?) available on today's mobile devices, many of them do have novel input. Consider: the multitouch technology on the iPhone that allows you to zoom in on maps by separating two fingers touching the device, the accelerometer on the HTC Touch Pro that will allow you to play racing games by moving the phone, and speech recognition on many modern phones, e.g., to dial your mother by saying her name
- The ability to connect to other devices -- Many phones have an infrared port that allows you to control external devices with a special application, though this feature does not seem to be widely used (given television watching has largely moved to YouTube
- Communicating and file sharing with other humans directly -- Though not as direct as sending something to someone's brain, mobile SMS, email and even telephony (voice over IP/VoIP) is widely used on today's phones
- Accessing news and other information sources seamlessly -- The killer feature of any "smart phone" or modern mobile phone is the ability to browse the Internet and, thus, to search for something in Google, look up a movie on IMDB, read stories on the Boston Globe, etc., all at the leisure of inputting such commands into your mobile phone (which, to be fair, is still somewhat poor usability, especially compared to using your brain)
- Uniquitous Internet access -- The ability to do just about any of the other things on mobile phones requires them to be connected to the Interent and, luckily, they are (and those that aren't are making progress toward that end). Sadly, the Internet access it not through hotspots or satellite as I predicted, but merely through the cellular networks themselves.
- Having ones own personal web site -- I think most will agree that this vision of mine ended up being manifested in the form of social networking web sites, such as Twitter and Facebook. So while each person does not have a web site on their phone, many do have access to these social networking sites on their phone and also have profiles (which are kind of like web sites) thereon.
Using the paper I wrote as a freshman in college as a guide, it seems clear that the single feature that The Chip had that modern mobile phones do not is a natural and intuitive interface. It is not obvious that mobile phones will ever (any time soon, at least) get the ability to read your thoughts, but that does not mean that they need to be painful to use (and that they need to be traffic hazards for people on them)!
|
|
|
| Approximations in AI |
[Nov. 18th, 2006|02:51 am] |
The Potential Uses of False Theories by Both Humans and Computers
Introduction
Humans throughout history, especially in fairly recent history have done their best to describe reality by any means available to them. We pay particular attention to getting close to the “truth” of matters, which is manifested by the fact that disciplines such as science and philosophy continuously go through paradigm revolutions and shifts on large and small scales, as new theories are devised.
Along this path, many theories have come and gone. A good theory has influence for a few years and then essentially disappears from discussion because it was found to be false or incomplete. Sometimes in truth-seeking disciplines a theory is built upon and expanded, emphasizing the inherent value in the original theory, but just as often, if not more so, a theory is forgotten and entirely abandoned.
A question not asked often enough is whether these aged theories could be used to approximate what they originally intended to exactly describe.( Read more... ) |
|
|
| Philosophy and AI |
[Jun. 28th, 2006|03:48 pm] |
I have been reading some book about Kant lately and I have recently finished the chapter on his theory of cognition: that is, how we use past experiences and knowledge, our power to reason, and present experiences to think (although "think" really doesn't do it justice, hence why they decided on "cognition").
In the past, I have strayed away from philosophy because it all seemed so pointless and useless (the pragmatist in me). Who cares how exactly humans are able to "know" anything? We do! Get over it!
Alas, this book is refueling an interest in philosophy for me.
Kant's theory of cognition hits on a few topics:
- The problem of mind-independent reality: is the outside world real (realism) or is it in our minds (idealism) or other?
- The problem of the origin of knowledge: does it ultimately come from experience (empiricism) or within us (rationalism) or other?
- The process of cognition: what goes into the cognitive process: memory, reasoning, perceptual sensations and application of abstract concepts to particular instances
- The process of concept forming: how are we able to understand ideal or abstract ideas while only experiencing particulars (e.g., the concept of triangularity when all we experience are things like triangles). In other words, how are we able to understand universals?
Basically, Kant hinges on the claim that our mind does a lot of the work that we often assign to the external world.
Why is this interesting to me? Artificial intelligence, of course. Anything that is about how humans do their thinking, their acting, deciding, and so on, can be useful to know for someone like myself interested in AI.
As a side note, anyone interested in transhumanism might also benefit from understanding of human cognition, if we are to improve it some day. Cool! |
|
|
| Jeff Hawkins and the HTM: he beat me to it! |
[Jun. 2nd, 2006|04:37 pm] |
In this article, Jeff Hawkins, the CTO of Palm Inc., discusses his model of the mammalian neocortex: the Hierarchical Temporal Memory (HTM), developed in his book On Intelligence
The HTM model is along the lines of what I have thought about when it comes to memory algorithmic formation, which could help to describe the dynamics of Quine's "web of belief" and the foundation of concepts in Kant's transcendental theory of cognition.
The difference between ideas I have previously posted about and his HTM is that the HTM is abstract enough that it can describe the formation of not only beliefs, but also concepts (including transcendental conditions for sensibility) and even ethical values and duties (as I previously posted about here).
The fact that so much philosophy and science can be put together to form such a cohesive picture of human cognition exemplifies the potential synergy between these fields.
To read about the details of the HTM, the primary white paper is located here. |
|
|
| Toward an ethical optimization learning algorithm |
[Apr. 30th, 2006|10:15 pm] |
Contents
* Introduction to naturalized philosophy
* The components of a complete ethical theory
* Meta-ethical proposal
* Ethical application proposal
* Possible criticisms
( Read more... )
It goes without saying that this is an idea for artificial intelligence ethics (or perhaps other types of learning, as well). |
|
|
| Evolution |
[Apr. 23rd, 2006|01:50 pm] |
Controlling Evolution
Throughout history, biological evolution has produced many solutions to technical problems of survival. When the lion needed food, it was given teeth, figuratively speaking. When the bird needed to flee preditors, it had wings. When the humanoid needed to do many of these things and did not have any of these features, they had tools to manipulate their environment.
We don't need wings because we can build planes. We don't need teeth because we have guns. And so on.
The tools that most allow us this power of environment manipulation are:
1. Hands, fingers (physical manipulation)
2. Reasoning, memory (cognitive manipulation)
3. Language (cooperation)
We are really the species that controls its own evolution, in this sense. How exciting! The way this evolution occurs is, then, through technology.
Artificial Intelligence
How does AI fit into the grand scheme of evolutions?
How will it help us evolve ourselves? Well... there are nanomachines with intelligent features that could do just about anything in the future while inside our bodies, there are "agent" technologies to do tasks for us to let us focus on other things in our lives, and then there is creating robots to completely replace humans in some areas.
Insofar as this means simulating many human characteristics, this is called "strong AI" - it's what you see in the movies. This is to be constrasted with the form of AI that I'm generally interested in - something like utilitarian/weak/whatever AI - the injecting of intelligence into everyday items to complete tasks that otherwise couldn't have been done, but not creating robots to talk to or anything.
Playing gods is the next step in our evolution, in the sense that we can create things to do tasks for us (not so much in the organism-creating sense). This is AI.
Is that megalomaniacal? Maybe. So sue me. |
|
|
| Object Construction |
[Apr. 16th, 2006|04:19 pm] |
At philosophy club, where many new ideas are found, Matthew and I discussed, among many other things, object construction.
What I mean by this is how we, as sensing humans, construct objects in our mind to understand the world.
In his philosophy of cognition, Immanuel Kant says that our experience of real-world objects is changed by the categories that we have inside us that we use to interpret them. In this sense, he is suggesting that we have some kind of innate categorization ability. Not only that, but the implication is that, to some extent, we have a framework already in us when we're born to use basic object skeletons in our experience.
A view similar to this is one called universal grammar, was proposed by Noam Chomsky (although the general ideas go back further). He says that all humans are born with an ability to understand language syntax (that is, the structure of human languages). That is why we can learn any human language when we're born and why we can translate among them. Things like similarities among languages (inflections, tenses, etc) are used as evidence for this theory.
What's interesting to me, though, is how these two theories are similar and address, really, the same issue - albeit from a different angle.
How are we able to categorize, abstract/generalize, and recognize patterns so well? Ignoring specific implementation details, such as neural nets, I'd like to suggest that the reason is that we have some kind of innate ability for these things.
Why this is interesting is that computers, generally, are not built with such an ability - and if my (and Kant and Chomsky's) assumption is correct, then we'll never be able to have efficient artificial intelligence without it. |
|
|
| Neural implants |
[Apr. 16th, 2006|01:15 am] |
I've been going through The Age of Spiritual Machines (recommended by a friend) lately, and some parts of it have made me think about the Chip.
First of all, Kurzweil says that neural implants are in the not-so-distant future. That is, electrical neurons (that are apparently 1 million times faster than biological neurons). These neurons use AI research knowledge known as neural nets for pattern detection and output, for each neuron, very simple results - but when put together in a extremely parallel network, produce exactly what our brain produces.
Somehow, the book says these neural implants could function in just about the same way as the Chip, but on a nano-scale. That, actually, seems like a lot better of an idea than a big old integrated circuit attached next to the brain: make it the brain itself.
Would these make a microchip in the brain obsolete? |
|
|
| The Chip: Remote login |
[Apr. 16th, 2006|01:14 am] |
I was thinking a while back about what would be fun with the Chip: remote login. You could login to another host (i.e., body) and use its resources - input and output devices.
That is, if you have the permissions, you could share the input devices: the five senses, basically. So, you could see things through another person.
Also, if you have permissions, you could use the output devices - that is, send signals to the person's arms and so on. So, with the input and output devices, you could control the person entirely - a la Ghost in the Shell, basically.
Of course, like computers nowadays, you would have to have security - both permissions and encryption on the login (e.g., ssh). |
|
|
| The Chip: Intrusion |
[Apr. 16th, 2006|01:14 am] |
Since this has been brought to my attention, I would like to make a post to discuss the issue of intrusion.
Who wants an operation to put something near their spinal cord? or their optic nerve? the operation for the cpu seems like the least of our worries, given that it's not in such a deep location in the brian as these other components. If it's not, though, then it must communicate with them through the brain - and if communication is wireless, that raises some radiation concerns.
Regardless, the issue I want to address here is primarily intrusion (although the others are important, too). What can we do about this?
I was talking to Jessica and she said that security was our primary concern, since body hacking is a problem. However, I noted that although that may be true, intrusion is chronologically more immediate, since the Chip could never be launched without it being solved - whereas security is an issue once it's launched. |
|
|
| On "the Chip" and Human Potential |
[Apr. 16th, 2006|01:11 am] |
Every decade, every year, and, in fact, every single day, what it means to be human is constantly redefined. Then, with the advent of the microprocessor and consequently the personal computer (PC), this redefining process took a sudden leap forward. The possibilities seemed endless once most people became acquainted with the Internet. For those who thought this, let me introduce something completely new and revolutionary to this assumption - the Chip. ( Read more... ) |
|
|
| Democratic meritocracy |
[Apr. 16th, 2006|01:09 am] |
We talked about Plato's Republic in intro to philosophy. I listened to the basic ideas and all, but what I mostly did was think. I do that sometimes.
The debate in the class was, basically, what's better? Democracy or Plato's rule by philosophers?
I thought... why not both?
Let's
say we take our current system (or, better yet, a European system, but
the differences are unimportant) and then introduce some standards.
Specifically, some degree requirements for the positions.
How
about we say all civil servants have to have at least a masters in
something. Then, candidates must have a phd - with a thesis in an idea
on how to make the government better.
So, when we vote for candidates, we'll be voting from a pool of phds.
It'll
be sweet. Just think. Civil servants/interns can get their masters and
then work on their phd while they're interning for an existing
candidate - much like how it works today (but with the degree
requirements added in, of course).
Once someone is elected
into office, their decisions must be logically proven and then reviewed
by a board of "senior" phds, which will also be elected from the phd
pool.
The legislature could debate over the validity or
soundness of arguments instead of pork barrels, interests, or any other
bullshit.
Maybe some will think this would be terribly
inefficient. I don't see a problem, though. Testing the validity of an
argument is a rather quick process for a philosopher (especially if
it's in symbolic notation!) and the time that would be saved over
bullshitting would be well worth it. |
|
|
| Intuition? |
[Apr. 16th, 2006|01:09 am] |
There are some people who would prefer to shrug off logic as some elitist method used to make others feel stupid. These people would prefer to use their feelings to verify the truth of something.
What is important to note, though, is that emotion - or, philosophically speaking, intuition - has its place in the thinking that we're doing here, modeling. Let's be truthful: when we come up with an idea, the idea just comes into our heads sometimes. We don't know why, but some connection was made such that we thought of this new idea.
The ideal of always having our brains use logic to come up with new ideas is not necessarily just unrealistic, it's probably dangerous, since creativity could be easily abolished if that was the case.
So, we can assume that intuition is necessary for the creation of new concepts and models. What about verification of them? Classically, logic is the tool used for verification of models - the consistency of the model being the chief property that is tested. Another popular method of verifying a model is how well its application works, given a set of instances of this being done. This is basically "experimenting" with the model and thus uses the logic of induction.
That's all and good, but what about models that cannot be experimented on (but are nonetheless internally consistent - i.e. have no self-contradictions)? A good example of this kind of model is a theory of consequentialist ethics. Since this kind of ethical system is, actually, impossible to implement (since no one is psychic), the only way to verify its "correctness" is intuition!
That is, consequentialism, at first glance, seems like a good idea. That's intuition talking.
The question, then, is whether or not intuition is a valid method for verification of models. I suggest that it is not, since any model or theory that requires it for verification is probably flawed (such as in the case of consequentialism). |
|
|
| The singularity |
[Apr. 16th, 2006|01:05 am] |
In The Age of Spiritual Machines, author Ray Kurzweil argues that at some point in the near future (a decade or two from now, more or less, at his guess), technological advance will increase to a point at which humanity will create intelligence surpassing himself and, since it's a "singularity", events will cease to be predictable in the ways they are today (an idea originally put forth by mathematician Vernor Vinge).
What will this mean for, say, physics? Well, if technology continues to advance at an exponential rate, then why is it absurd to suggest that mankind will affect larger bodies in the solar system as astrophysics is thought to? Could we build solar systems in the future?
Our effect on biology is already rather clear - technological evolution is an extension of biological, as far as we can tell.
What other areas will the "singularity" cause to cease to be predictable in classical senses? |
|
|
| Context |
[Apr. 16th, 2006|01:02 am] |
The very thing that defines human reasoning as opposed to anything that is easy to program is context
Context is defined by three things.
1. An environment - specifically, interacting to stimuli in the environment and having it change how you reason
2. Identity - an idea of who you are (something a computer generally doesn't know - instead, they often have "purpose")
3. Communication - in other words, other agents that you can talk to (which is like environment, but more specific)
Context in intelligence is essential for any model of human intelligence as well as any artificial intelligence approach that hopes to be successful - at least in a full/strong AI sense (more than amazon knowing what kinds of things you like). |
|
|
| Framework of artificial intelligence |
[Apr. 16th, 2006|01:02 am] |
If we are to model human intelligence or reasoning in a computer, we'd want to cover the two basic areas: memory and reasoning.
What this amounts to is that an intelligent machine must:
1. Be able to both learn and use knowledge ("memory")
2. Be able to reason with logic with or witout memory ("reasoning")
But wait, there's more:
3. Be able to connect on an emotional or social level
This last one is only for "robots" that we wish to model humans as a whole - not for AI applications that are utilitarian (e.g., using AI for database knowledge discovery).
So 1 and 2 are the essential aspects of AI - and, lo and behold, that is what research in AI is focusing on: machine learning (1), and reasoning (with uncertainy, causal reasoning, etc) (2). |
|
|
| navigation |
| [ |
viewing |
| |
most recent entries |
] |
| [ |
go |
| |
earlier |
] |
| |
|
|