Skepticism and the scientific method

As I have mentioned before, I am something of a Pyrrhonian skeptic. Today, I would like to discuss the differences between pyrrhonism and the scientific method, as well the effects of this difference.

In many ways Pyrrhonian skepticism and the scientific method are quite similar. Both teach that human beings are small, Lilliputian creatures in a big and mysterious universe. Both embrace the unknown, asking people to search for answers amid uncertainty.

But beyond that, the two philosophies are radically different. The first main difference has to do with absolute truth. Pyrrhonism rejects the idea of absolute truth, stating that nothing can be known with absolute certainty. The scientific method, on the other hand, holds that absolute truth can be uncovered with experimentation. Science claims that the universe is a logical one, and certain truths can be uncovered by logic and reasoning (skepticism rejects this analysis, since the universe is not necessarily logical).

The other central difference between skepticism and the scientific method has to do with one of the later additions to the list of key principles of the scientific method, Occam’s Razor. This is the idea that if one is presented with multiple hypothesis, the simplest one is the one that should be considered the most credible. Occam’s Razor has its roots some of the writings of Thomas Aquinas. In one of his works, he states “If a thing can be done adequately by means of one, it is superfluous to do it by means of several; for we observe that nature does not employ two instruments where one suffices.” Pure skepticism rejects this analysis, since the complexity of a hypothesis is not necessarily related to its validity. Also, Occam’s Razor includes the elimination of hypotheses that are contrary to common sense. Skepticism rejects this analysis as well, since “common sense” is subjective and the popularity of a hypothesis may or may not be related to whether it is true. On a related note, the scientific method holds that an experiment should be able to replicated and any other scientist should be able to achieve the same result. (Skepticism rejects this claim as well.) 

So, are the scientific method’s divergences from skepticism strengths or weaknesses? Are they what has carried science this far, or are they foibles that have prevented science from reaching even further? I suspect that the answer is a little of both. Pyrrhonian skepticism’s rejection of absolute truth would make science grind to a halt if it were a part of the scientific method. This rejection absolute truth would also contradict the scientific principle that experiments can always be replicated, which would make experimentation pointless. Science needs some level of common sense, since it assumes philosophical Realism and Infallibilism (otherwise there would be no point to truth-seeking).

But Occam’s Razor is another matter—I believe that this principle has actually slowed science down. The main reason for this is that the definitions of “common sense” and even “simplest hypothesis” are dependent on cultural factors and religious beliefs. For example, in Galileo’s time the “simplest hypothesis” was that the Sun rotated around the Earth. Today, the scientific community often uses Occam’s Razor to shun hypotheses that seem unlikely or unpleasant to them, even if such hypotheses are plausible. Though science needs some level of common sense, Occam's Razor is going too far.  

Beyond Turing-completeness

Today I read an article in the science magazine “Discover” about a new kind of computer called Darwin 7. The article was in the form of an interview with computer scientist Gerald Edelman. Edelman explains that in the biological world, there are obvious advantages to consciousness, particularly the higher-order consciousness human beings have. Conscious beings are able to adapt to different scenarios and learn, making them more adaptable. Therefore, says Eldelman, it would be advantageous to create computers that are based around a model of a brain rather than being programmed. Edelman and his colleagues have done just that: The device is called Darwin 7, and it is, as Edelman says, a computerized brain.

Before I discuss the implications of this it is important to explain the difference between a computer program and a BBD, or brain-based device (like Darwin 7). A computer program consists of a series of instructions typed in computer code. In my post on the Chinese Room, I explored John Searle’s proof that a computer program cannot truly have understanding because it has syntax but no semantics. But a BBD is very different—its “brain” is not encoded but instead is an physical object. It simulates the neurons of an organic brain in order to “think.” In other words, a BBD is not Turing-complete; it is something different entirely. Searle’s proof does not apply; a BBD is capable of true understanding and learning.

The latter has already been tested, says Edelman: Versions of Darwin 7 have been taught to perform various tasks, and the advantages of a machine that can learn are very clear. In one test, robots controlled by a BBD played soccer against robots controlled by an AI program. The BBDs won 5 games out of 5, since they were able to adapt to every situation, while the AI-controlled robots did not have conditionals for every scenario.

Furthermore, Edelman says that the future of BBDs is bright. Edelman and a colleague have created a BBD that is about as complex as a cat brain, and it is very close to what he calls a “conscious artifact.” This BBD is so complex that it runs continuously like a real brain (simpler BBDs only react when they receive input), but it lapses into a “rest state,” similar to the state people’s brains are in when they are not thinking of anything. The point it, sooner or later BBDs are going to surpass AI because of their ability to learn. 

In my mind, the concept of super-smart a BBD re-raises the question of computers in relation to the future of humanity. Science fiction sends us conflicting messages about how computers will affect our chances for survival as a race—some SF preaches an optimistic message, while other works warn us that computers will be our downfall. Today, though this fear is somewhat present in our culture, most computer scientists hold that there is nothing to fear from AI since they are simply a collection of conditionals. They still follow Searle’s Chinese Room, so they are incapable of though are therefore cannot consciously act to destroy humanity. But BBDs are different. Though I know little about the subject, it appears that BBDs are far more likely to “betray” humanity than a lifeless computer program. The more complex a BBD, it seems, the more intelligent it is and the more likely it is be irrational. Additionally, remember that BBDs are modeled after human brains—and human brains are not exactly the most efficient or rational thinking machines in existence; far from it. So, while I have few qualms with letting a computer program run the world, a BBD is a different matter. Before we use these new devices, we need a far better understanding of them. Hopefully BBDs will facilitate the study the brain, which in turn will allows to create better, more stable BBDs. For now, though, all they do is play soccer, so I am not worried just yet.

On a different note: I would like to end with an amusing hypothetical situation involving Turing-completeness. Recall that even analog computers are Turing-complete, since they can technically be programmed for every task. In a comic strip (link), Randal Monroe envisions a new kind of computer, which is technically Turing-complete. I find his idea both hilarious and fascinating, if somewhat impractical. Even funnier, it is a philosophical stance that technically cannot be disproven. So I guess we could be just a bunch of rocks. 

Drake’s Equation and the Great Filter

Astronomer and scientist Frank Drake devised what is now known as the Drake Equation to answer the question of how much extraterrestrial life exists in the universe. In this equation, Drake included a variety of factors that determine if detectable intelligent life can exist (click on image at right for exact explanation of variables). In 1961, Drake and his colleagues estimated the values of these factors, and their result was that there should be approximately ten planets in the galaxy with intelligent life. This gave much encouragement and support to programs such as SETI, but so far NASA has been unable to find either an Earth-like planet or life beyond Earth. Today, the values in his equation have been revised, yielding a new result of 2. Even so, the Drake Equation is mostly based on conjecture, and though it is the best attempt to explain the apparent absence of extraterrestrial life it is far from reliable. Also, Drake’s equation only applies to detectable extraterrestrial life—for all we know, there could be plenty of life out there: as long as it isn’t constantly broadcasting on the handful of frequencies we monitor, we would have no way of finding it.

Regardless, the question remains: Is there extraterrestrial life? And if there is, why haven’t we seen it? Futurist Robin Hanson suggests that there is a “Great Filter,” or a factor preventing extraterrestrial life from flourishing. Hanson examined the evolution of life on Earth, and his conclusion (which is mostly accepted in the scientific community) is that the Great Filter exists in one of the following eight steps toward complex life:

1) The right star system (including organics)

2) Reproductive something (e.g. RNA)

3) Simple (prokaryotic) single-cell life

4) Complex (archaeatic & eukaryotic) single-cell life

5) Sexual reproduction

6) Multi-cell life

7) Tool-using animals with big brains

8) Colonization explosion

Today, we are still unsure of which one of these is the Great Filter. However, many scientists and philosophers have examined the Great Filter, and various different conclusions have been drawn. My favorite analysis of the subject is the futurist Nick Bostrom’s (the link to his paper can be found here). I would like to take a moment to summarize Bostrom’s analysis and conclusion and why I find it to be the most satisfactory.

Bostrom explains that human beings are between steps 7 and 8. Therefore, the Great Filter could be either behind us (in steps 1-7) or ahead of us (step 8). However, most of science tells us that steps 2-7 follow naturally from each other, since once simple life begins (at least according to biologists) it will eventually evolve into complex life. This leaves us with two options: either the Great Filter is the initial synthesis of life, or it is space colonization and expansion. There is some evidence that suggests it is the former: though organic compounds have been spontaneously created in a simulated “natural” environment, no cellular life has ever been synthesized. If it is the latter, though, our future looks quite bleak. This would mean that some factor that we will encounter in our near future will either cripple us or destroy us entirely (Nuclear war? Overpopulation? Epidemic?) For this reason, says Bostrom, we should rejoice at not seeing any simple extraterrestrial life, since that would imply that the Great Filter is ahead of us.

Looking at the problem from a purely scientific point of view, it seems to be common sense that some extraterrestrial life must exist. The galaxy is not the universe—there are billions of galaxies, and so many stars that we can never search them all. Though the Drake Equation says life is unlikely, the universe is so big it is bound to be somewhere. For now, though, we are going to have to be content with pointing radio antennas at the sky, since we have no other way of searching for life. Perhaps this is for the best—I doubt human beings would react well to finding extraterrestrial life. But that’s a discussion for another day. 

Musings on the Turing Test (part 2)

Happy New Year to all! I hope that 2009 is a better year for all than 2008. Sadly, the Gaza crisis is still going on, but I hope that it will soon be resolved. However, I will not be addressing that particular issue today, as I would like to talk more about the Turing test.

The Loebner Prize is a contest held every year, in which contestants try to program a computer to pass the current year’s version of the Turing test. The winner is the program that manages to appear “human” to the greatest number of examiners. In 1991, there was some controversy over the winner; the computer program considered “most convincing” fooled many examiners because it was programmed to make typing errors. Since then, the Loebner Prize has focused on “chatterbots”—computer programs that simulate a conversation (typed of course).

This brings me to the point I would like to discuss today: the so-called concept of “artificial stupidity.” This is the idea that computer programs must be made to make errors in order to appear human. This idea is not something new; even Alan Turing in 1948 realized that a computer that appears perfect cannot pass as human:

“It is claimed that the interrogator could distinguish the machine from the man simply by setting them a number of problems in arithmetic. The machine would be unmasked because of its deadly accuracy.”

Turing’s point is clear: to appear human, machines cannot be perfect.  This is evident in the Loebner Prize winners of both 1991 and 2008. In fact, after looking at the 2008 transcripts, I realized that all of the top 5 programs committed errors on purpose. (These transcripts are available here). Also, many of the more successful ones delayed their responses by an amount of time proportional to the number of words they “typed,” as an instantaneous response is suspicious. This, too, is a form of “artificial stupidity.”

What does this mean for us? For the layman, very little. For the computer scientist, though, it more clearly defines the challenge of making computers seem human. This challenge no longer consists of simply making computers smarter, as it did 30 or 40 years ago; now, it consists of making computers imitate all the nuances of human beings. This is probably a much harder task, but I have no doubt that computers will eventually get there. When we reach that point, we are going to have to ask ourselves some serious questions about our humanity. Until then, all we can do is wait.

**As a side note: Last year’s winner of the Loebner Prize, a program called Elbot, can be “talked to” on the creator’s website. I conducted a half-hour conversation with it, and what I found was startling—the program is able to have a perfectly normal-sounding dialogue on almost every subject. I strongly recommend trying it out for yourself; the link can be found here.  

Exploring the Chinese Room

Monday, in my post about the Turing test, I briefly explored John Searle’s thought-experiment “The Chinese Room.” Today, I would like to delve further into this interesting topic.

First, I would like to better explain the argument itself—I feel did something of a shoddy job of doing so in Monday’s post.  Rather than explain it myself, I will quote Searle’s description of the thought-experiment from his paper, “Minds, Brains, and Programs.” Unfortunately his description is a bit lengthy:

“Suppose that I'm locked in a room and given a large batch of Chinese writing. Suppose furthermore (as is indeed the case) that I know no Chinese, either written or spoken, and that I'm not even confident that I could recognize Chinese writing as Chinese writing distinct from, say, Japanese writing or meaningless squiggles. To me, Chinese writing is just so many meaningless squiggles.

Now suppose further that after this first batch of Chinese writing I am given a second batch of Chinese script together with a set of rules for correlating the second batch with the first batch. The rules are in English, and I understand these rules as well as any other native speaker of English. They enable me to correlate one set of formal symbols with another set of formal symbols, and all that 'formal' means here is that I can identify the symbols entirely by their shapes. Now suppose also that I am given a third batch of Chinese symbols together with some instructions, again in English, that enable me to correlate elements of this third batch with the first two batches, and these rules instruct me how to give back certain Chinese symbols with certain sorts of shapes in response to certain sorts of shapes given me in the third batch. Unknown to me, the people who are giving me all of these symbols call the first batch "a script," they call the second batch a "story. ' and they call the third batch "questions." Furthermore, they call the symbols I give them back in response to the third batch "answers to the questions." and the set of rules in English that they gave me, they call "the program."

Now just to complicate the story a little, imagine that these people also give me stories in English, which I understand, and they then ask me questions in English about these stories, and I give them back answers in English. Suppose also that after a while I get so good at following the instructions for manipulating the Chinese symbols and the programmers get so good at writing the programs that from the external point of view that is, from the point of view of somebody outside the room in which I am locked -- my answers to the questions are absolutely indistinguishable from those of native Chinese speakers. Nobody just looking at my answers can tell that I don't speak a word of Chinese.

Let us also suppose that my answers to the English questions are, as they no doubt would be, indistinguishable from those of other native English speakers, for the simple reason that I am a native English speaker. From the external point of view -- from the point of view of someone reading my "answers" -- the answers to the Chinese questions and the English questions are equally good. But in the Chinese case, unlike the English case, I produce the answers by manipulating uninterpreted formal symbols. As far as the Chinese is concerned, I simply behave like a computer; I perform computational operations on formally specified elements. For the purposes of the Chinese, I am simply an instantiation of the computer program.”

Searle’s point is obvious: In the proof, he is manipulating Chinese symbols without true semantic understanding of what they mean. This, he argues, is what computers do: they simply carry out “the program” without having true understanding of what they are doing. It is important to note that Searle is not a dualist—he does not believe the human mind has any kind of non-physical component. He concedes that the human brain is simply a biological “machine,” and that an artificial mind could hypothetically be constructed. Searle is trying to prove that a computer program can never create a true “mind” because computer programs are in scripts that have syntax but no semantics. Essentially, Searle is challenging the computational theory of the mind: the idea that human beings cannot be explained in terms of input/output (note how similar this is to philosophical determinism).  

Also, I should mention that though I had never heard of the Chinese Room argument until the other day, it is one of the most important issues in cognitive science and philosophy today. In fact, the influential computer scientist Patrick Hays even joked that cognitive science should be renamed “the ongoing research program of showing Searle's Chinese Room Argument to be false.” There are an enormous number of responses to the argument, and unfortunately I do not have time to cover them all today. However, I would like to look at the implications of Searle’s argument and at some of the more convincing responses. 

Many philosophers and scientists have looked at what the Chinese Room thought-experiment implies, including John Searle himself. Searle created the following proof from his thought-experiment:

Axiom 1: Computer programs are formal and syntactic.

Axiom 2: Minds have mental, semantic contents.

Axiom 3: Syntax is not enough to create a semantic mind.

Conclusion: Programs are “neither constitutive of nor sufficient for minds.”

Searle’s conclusion is intuitive enough, given the data he is starting with. Axioms one and two are pretty obvious—1 states that computers have no true understanding of things, and 2 states that human minds do. Axiom 3 is what the Chinese Room proves—the fact that a computer can pass a Turing test without true understanding (at least, according to Searle). However, as I mentioned, the Chinese Room has attracted thousands of intellectual critics, and there are a multitude of responses to the proof from various areas of science. These responses attack Searle’s axioms, his conclusion, and the validity of the thought-experiment itself. I would like to take a few moments to explore some of these claims. 

The first is the “systems” response. This states that even though the man in the room does not understand Chinese, the man, the room, and the program as a system do. However, Searle’s reply is that it is possible for the man to memorize the program, making him the entire system even though he still has no understanding of Chinese characters. The “systems” reply is that the mind is virtual mind, which has a variable physical component. (For example, the software of a computer is a virtual machine) Thus, there is an “implementation independent” virtual mind at work. Searle, however, would maintain that such a virtual mind is still a syntactic simulation incapable of cognitive understanding.

Other responses are related to so-called appeals to reason. For example, a “program” to do what Searle is suggesting would be enormously complex, and it may require a whole new kind of programming. However, I will not even address these because they are insignificant—the Chinese room is a hypothetical case, after all. 

So, what is the final verdict on Searle’s Chinese Room? I don’t have one. Searle’s proof seems legitimate, but several of its aspects remain unproven, as many of the responses show. I promise to revisit the Chinese Room soon, since it is such an important and influential argument. For now, all I can say is that since the Chinese Room resides in the grey area between science and philosophy, someday either experimentation or logic may yield the answer. 

Musings on Daisyworld

James Lovelock and Andrew Watson’s classic computer simulation Dasiyworld is a major proof for the Gaia (“Earth as one organism”) hypothesis. In the simulation, a planet has only two species, white daisies and black daisies. The black daisies warm themselves; the white ones cool themselves. According to the simulation, when the planet’s temperature is cooler there is a predominance of black daisies, and when it warms the white daisies increase in number as the black ones decrease. The idea is that the combination of the white and black flowers serves to regulate the planet’s temperature to a certain extent. The Daisyworld planet’s temperature is more moderate than that of an empty planet because of this balancing effect. This has been used to “prove” the Gaia hypothesis because it highlights that fact that organisms can unwittingly work together to function as one large organism.

However, I feel I must deprecate this idea. Though I cannot refute the fact that organisms often work together as a single organism, there are too many random factors involved for this hypothesis to be true. Evolution itself is a random factor, brought about by the mutation of genes. A random mutation could easily create a new species that does not fit into “Gaia.” I suspect humans are the perfect example of this—how is it that Gaia would welcome a species that is psychologically unstable but can build nuclear weapons? Also, environmental or other external factors can clearly unbalance this cycle.

Because of all this, it is absurd to think that this kind of a relationship between species exist for more than a short period of time. Though this kind of symbiosis can occasionally occur, the idea that the whole planet follows this pattern seems a bit of a stretch. 

What computers can teach us

Philosophers and even computer scientists will often argue that computers can never achieve our “thinking power,” and that they definitely cannot surpass us in terms of cognitive ability. However, I strongly disagree with this, for both a philosophical reason and a practical one.

First, the “practical” reason: Simply, computer science in increasing at a terrifying rate (an exponential rate, if you believe in the technological singularity). Twenty years ago, people believed computers would never be able to solve the “salesman problem” for only 50 or so cities—today, we have solved it for thousands, according to this New York Times article.             

The fact is, unpredictable advances in the way we understand computers will probably allow them to do some serious number-crunching, making them far better and faster at mathematics and science than we are. Furthermore, computers will soon have the ability to improve themselves, thus increasing the rate that artificial intelligence, or AI, increases.

Next, the “moral” argument I mentioned. The point that skeptics of AI often mention is that AI will never be truly human, since a computer is simply a complex series of pre-programmed responses. However, I have to ask: what, then, are we? Isn’t a human being very similar—just a very complex series of responses programmed into our brain? Granted, we are affected by our environment, but computers can also learn from experience—the ability to learn is nothing special. As any computer programmer knows, with enough “if” “else” and “while” statements, computers can learn and adapt to their environment no matter what the circumstances are. Thus, human beings as computers are not so different after all, and artificial intelligence might not be as impossible as some people like to think. Though this sounds deterministic, this is the most salient argument for AI.

The fact is, computers are already on of the most important inventions of the modern world, and the development of true AI is implacable. The important thing, though, is how we deal with the fact that computers will still be far smarter than human beings. I do not agree with those who would ban computers or put a moratorium on computer research—the answer to the problem of computers must be one of wisdom, not deliberate ignorance. We must not simply succumb to the inevitable—we must ride the crest of this wave, using the enormous potential of AI to help us in a way that will not infringe on our humanity. 

On “The Last Question”

Inspired by Asimov’s The Last Answer, I have decided to talk about another of his famous short stories, The Last Question. In this story, a computer called AC ponders whether entropy can be reversed. However, it does not come up with the answer until after the universe undergoes heat death. At the end, the computer discovers the answer and decides to demonstrate it:

“And AC said, ‘LET THERE BE LIGHT!’ And there was light—”

Of course, this brings up some interesting questions. First of all, Asimov is referencing the concept of the technological singularity, the idea that technology increases at an exponential rate, and that it will eventually increase infinitely fast so that it encompasses everything. However, I have some doubts about this. If the world were to have a nuclear war or fall into a worldwide depression (both somewhat realistic!), wouldn’t the rate technology is being invented at decrease? After all, technology is probably dependent on the economy (and certainly on the world population). If such an event were to occur, we may never reach this technological singularity.

Asimov is also making an amusing statement about the creation story. (A devout atheist, he often derides religion in his writing.) In The Last Question, he is toying with the definition of God by exploring how an omniscient computer could possible “become” God. If the concept of the technological singularity is true, this is possible—an intriguing yet frightening thought.

But don’t look for this in your lifetime—one thing Asimov hammers home in this story is the stretched-out timeline. If it happens at all, the technological singularity will not happen for billions of years, and our moral, ephemeral bodies simply won’t last that long.