Create AI from randomness, chaos and complexity? Sure, but it's just as likely to happen in a batch of cookie dough.

Well, AI arising out of random chance may not be so unlikely. Life arose out of chance - granted, a trillion years of chance, but we're creating an increasingly diverse electronic environment for that to happen.

Either way, I'm guessing we won't know when we have created an AI. Some people will think it is, some people will think it isn't, something weird will happen, and nobody will know what we did. Hopefully christian bale will still be around.

Human brains are nothing but highly parallel computers running software in the form of chemical reactions driven by proteins. As romantic as it seems, we are nothing but mechanical non-miracles. Machines will overtake us in terms of number of circuits, parallelism, and storage capacity. Sorry but it has to happen unless we destroy ourselves first.

Billions maybe. We're a wee bit short of a trillion yet.

(Just in case the kids are watching.

Jeez pepe, this is a lot of robo fearmongering going on.

I completely agree that people function just like computers, and that our existence plays out according to complex chemical reactions in the brain. I'm not trying to make some huge point about us being some divine creation.

I'm saying that we are self aware, and computers are not. The discussion is whether computers will ever be able to be conscious and thus, truly thinking. I think that is a required element if we're going to create a competing machine based life form.

Since we have radically expanded computer's thinking power over the years without any detection of even the mildest self-awarence, I think its safe to conclude that neither software complexity nor raw computing power is the engine behind consciousness. Until that starts developing in machines, I'm not going to be concerned, as I don't think they could be an uncontrollable threat otherwise.

(This is, of course, a different scenario than one in which we ourselves accidentally program robots to destroy ourselves. Thats much more plausible in my mind.)

How exactly are computers programs static? If you examine the consequences of the theory of Universal Computation, computers exhibit a quintessential form of chaos. There is as much chaos in a one dimensional cellular automata as there is in the physical mechanisms of the brain.

Self awareness? What's that? Are you aware of the complete totality of your being? What then is self awareness? Haven't the last couple of centuries of psychology taught us that most people harbor some pretty staggering delusions about who they are? How many IQ points do you gain by being really really self-aware?

What is it that humans do that is so special? Most of your interactions and decisions can be described by words. Anything that can be described by words can be parsed by a computer. What disciplines can't be emulated by an expert system?

I can see robots being programmed on purpose to kill Luddites...

What is it that humans do that is so special? Most of your interactions and decisions can be described by words. Anything that can be described by words can be parsed by a computer. What disciplines can't be emulated by an expert system?

Experience pain and colors. See my post above where I get into this a little bit. The whole category of perceptions, sensations and desires is a little hard to square with strictly logical programming.

For instance, if you believe that human color perception is simply a matter of fooling the human brain into believing it sees something called 'color', then you have to conclude that human brains simply aren't very smart or logical. A perfectly logical robot could not see color, it would simply see a field of numbers.

Now it's true that a robot could take those numbers and compute 'colors' correctly; it could correctly say that apples are 'red' and blueberries are 'dark purplish blue,' but it doesn't have to experience 'perception' in any way to do that. In fact our current computers are perfectly able to do this just as well as any robot could, but there's no question that they do this without experiencing anything like 'perception.'

In fact it's perception that's mysterious, not intelligence. If human brains are just logical machines, then what is it that these logical machines are experiencing when they see colors such as red, blue, yellow or green? Or when they 'feel' pain or have desires? In fact how can logical machines experience anything at all? What's logical about perception? What kind of logical idea is the color 'red'? No one has ever been able to put the color red in words, so here's something that can't be described, yet we can see it.

I believe these are questions that have to be answered if a real artificial intelligence is ever to be created.

What is it that humans do that is so special? Most of your interactions and decisions can be described by words. Anything that can be described by words can be parsed by a computer. What disciplines can't be emulated by an expert system?

We find random connections where there are none. We find/create patterns where there are none. To steal a bit from Asimov, we can come to correct conclusions based on incomplete instructions and information . Emulating intelligence isn't the same as creating a real intelligence.

Computers think linearly. No matter how many processors you through at a problem, all the computer systems we have now start at line 1, go to line 2, end on line 3. There may be loops, recursions, etc but in the end its still a linear process.

Humans don't really think linearlly. We can force ourselves into the thought processes of formal logic and mathematics, but our brains work on many levels at once. This is what gives us the ability to look at a random blot of ink and see faces and designs in it. It allows us to think in metaphors.

This is what I was getting at with my first post about an AI arising from the chaos of current systems. Only from such chaos can I see an AI be able to do more than "emulate" human intelligence. I'm not sure I myself buy the idea that it could happen, but...

Sentience and consciousness are emergent properties of the brain. There is far more to the whole than can be described by looking at the biology. Its like trying to describe the contents of a Word Document as being nothing but 1s and 0s, or describing a piece of music in terms of the physical properties of the sound waves- frequency, amplitude, etc. In both cases there is useful information in such a description, but the central essence is lost. I'm not a religious fundamentalist looking for some mystical solution to the issue like a soul. I'm only pointing out that the current vogue in brain science is overly reductionist and we are far from understanding the workings of the human mind and consciousness. Or the animal mind for that matter.

I agree with a lot of the current thrust of posts, but the idea that we'll have to figure out what makes sentience tick before we can design a computer to do so seems to me to be a false standard.

For one thing, one of the primary reasons we can't figure out the source of sentience is that we can't agree on a definition of sentience. Self-recognition in an advanced sense is reserved to humans and seemingly, from recent research, possibly a couple of other primates, but of course, self-awareness in terms of awareness of your physical being is something all animals have to some extent, and that bacteria have in a certain manner of speaking (that is, they are able to react to environmental stimuli in determined ways to move towards food, for example, or extend a psuedopod in some direction; these aren't "choices" in a human sense and are very much similar to running a biological program, yet is also qualitatively different in that the types of sensory inputs its subject to are quite different from any inputs your typical computer would recognize). There are of course many unresolved arguments as to the degree humans "choose" anything, not from a genetic determinism viewpoint but rather from the quite reasonable viewpoint that our actions are controlled by the physiochemistry of our brains. We can "self-stimulate" various processes within our brains, but that is still one set of chemical reaction causing others, even if it's all very complex. While avoiding the full free will/determinism debate (or trying to), it's fair to say that human thought is the process of various physiochemical reactions; one can argue that the complexity has crossed some threshhold into "self-determination", that there is some super-natural "soul" from which the chemical processes proceed (which is made ever more unlikely, empirically speaking, as we understand more about how we're subject to the physicality of the brain), or that our thoughts are indeed simply and only a number of biochemical "programs" running in parallel, responding to biochemical stimuli.

All that being said -- to argue that we have to understand, and seemingly therefore resolve these questions before we could program computers to really come close to emulating the same is like saying we have to completely understand genetics before we can make artificial biological life -- it's not so. (The advisability of creating novelties from a system you don't understand fully is a different matter.) We have an advanced understanding of certain parts of genetics, and can make simple organisms or quasi-organisms (viruses), and alter more complex ones (animals and plants). The ability to make a computer program you don't understand is elementary -- a colleague once said of a program he made to emulate ecological processes, "I started up with a phenomenon I didn't understand, and ended up with a program I don't understand."

Designing a computer program capable of re-writing and re-programming itself, and complex enough to respond to a number of outside cues and write new programming to reflect this, is complicated but it seems to me imminently doable in the near future. Will that be enough for sentience? Of course not. But I bet we could do it, and not understand it -- and we'd be on a road to making computers that could potentially do more, without our understanding; a robot that can winnow and strengthen aspects of its own program in response to external stimuli just doesn't seem technically beyond our grasp; once you have that, it would seem you could rapidly simulate some kind of evolution that would undoubtedly generate something you couldn't understand with novel behavior (just as creating a bacterium or virus and subjecting it evolutionary pressures would create responses that would likely not be predictable beyond a limited time scale).

As far as the "gray goo" goes, I'm not so worried. Most fictional depictions of it show speeds of replication and action that, frankly, violate fundamental laws of thermodynamics (that energy has to come from somewhere, and waste heat has to be disposed of somehow). In my personal opinion, we will be unable to build nanomachines that are substantially more efficient, active, or dangerous than naturally occurring nanomachines like bacteria.

The reason I keep coming back to sentience isn't to say that we need to "find" it, or come up with some sentience program for machines, but to point out that its gradual emergence in organic life has, thus far, been entirely absent from computer evolution.

As to what sentience IS, yeah thats a big debate. For the sake of this discussion, I see sentience as being a necessary component to make independent decisions free of routinized programming.

The fact that sentience is absent from computers but present in living organisms IS significant. I'm not saying we're divine creations or anything, but there is a fine red line that separates sentient life from non-sentient activity, and it creates a huge difference. Without a way to understand the world around them, and AI will be more artificial than intelligent.

What is it that makes your form of perception more real than how a how a computer perceives something? You are aware that the world your perceive isn't the world that exists objectively? Red is only red because some tasty berries in the jungle were red and there just happens to be the right chemical for transducing a very specific frequency of light into a digital signal.

The logical idea of the color red? A frequency of electromagnetic radiation between X nanometers and Y nanometers.

From there, your brain says "related to the leaves changing colors in the fall, ripe strawberries" etc...

Actually for a computer the logical idea of the color red is a set of three numbers.

We can write computer programs that deal with color in any way we want. See Photoshop or MS Paint, just to name two programs. However in none of these programs is there any possiblity of the computer somehow having or needing the 'perception' of a color. The computer just manipulates symbols and to a computer a color is just another symbol. To say a computer can have perceptions is to anthropomorphicize the computer.

Human beings have perceptions, not only of color but of pains and feelings and urges and wants. And I think that's a really interesting difference between humans and computers. It seems to hint at fundamental differences, which I don't believe can be explained away by hand waving.

As you say, we don't perceive the world directly. Our brains apparently handle color information in a way that must be similar to computers. And then we somehow 'perceive' that information as colors. Now when we have those perceptions of colors, we can't really do much of anything with those perceptions. We can't manipulate our perceptions in anything like the way that Photoshop can manipulate colors. We can't store specific colors in memory, for instance. We can't change colors or replace them or adjust them, certainly not at will. For supposedly intelligent biomachines, it seems we're pretty limited in what we can do!

If we were truly intelligent, it would seem that we would have access to the computing parts of our brains. It would seem that would be an extremely useful and logical way of going about things. We wouldn't have any trouble with math, for instance. We would be able to precisely remember and catalog individual people, lions, and plants. But our consciousness seems to be stuck permanently in a vague and fluid world of 'perception.' Why is that, and what is this perception thing all about? And what good is it?

We all experience a vivid sensation of the color red and other colors (if we're not color blind). Well what sort of idea is that? If our brains are sophisticated computing machines, then why is it that some people (me) have trouble adding 2 plus 2? If we ever create thinking robots, will these robots have trouble with math, too? And if that's the case, what kind of software program is that, which can't even do math?

I just think that 'intelligence' is really a solved problem. Our computers are already as intelligent as we could ever want them to be. When we want thinking robots, I believe what we want are robots that can perceive things, and I just don't think perception is a solved problem at all.

Nerdnam, I completely agree.

Not only is our perception unique from computers, computers have a unique numerical processing ability that make the two of us even more different. A pocket calculator can do mathematical equations in a flash that no human can do in their head. But it isn't the slightest bit aware of what its doing. I, on the other hand, am possibly a bit too aware of what I'm doing, and I forgot my division tables a few years back. I still win in the "intelligence" game.

All this speculation about what computer's perception actually is, well, thats just wishful thinking. There is zero, nada, zip perception inside even the most complicated computer today with the most sophisticated AI program. Until a machine is capable of understanding that it exists and there is an objective world that it is perceiving and interacting with, then it has no way to evaluate that outside world in intelligent terms and make sophisticated decisions of how to accomplish goals in the real world.

Thats why a computer can't learn or apply anything new without being programmed to.