Uncanny valley chart on 30 Rock

The uncanny valley graph made an appearance on 30 Rock last week where the show’s writers upped the chart’s silliness by explaining it “in Star Wars.”

What does the Turing Test test?

I’ve been thinking about the Turing Test this week after reading this Wired article profiling futurist and inventor Ray Kurzweil. The following paragraph describes part of the plot from an upcoming movie featuring Kurzweil:

Ramona [a sentient AI] is on a quest to attain full legal rights as a person. She agrees to take a Turing test, the classic proof of artificial intelligence, but although Ramona does her best to masquerade as human, she falls victim to one of the test's subtle flaws: Humans have limited intelligence. A computer that appears too smart will fail just as definitively as one that seems too dumb. “She loses because she is too clever!” Kurzweil says.

So what is actually measured by the test? Here, Kurzweil claims that AI will have problems with the Turing Test because, almost by definition, AI will be more “clever” than humans. However, in the next paragraph Kurzweil suggests that the Test is actually meant to measure “human emotion,” rather than intelligence.

Whether or not the test is meant to measure intelligence or the presence of human emotions, both standards are problematical in that they both depend on language. Consider the example of today’s Dilbert:


In the cartoon, Scott Adams suggests that certain patterns of human behavior, in this case, business jargon, would lead to a failure to pass the test (apparently because that jargon is designed to have the appearance of intelligence, rather than actual intelligence). Which is a great example of the way in which the Turing Test isn’t an absolute method of classifying intelligence, no more than an I.Q. test is. Instead, it is a culturally-situated artifact that attempts to account for a particular conception of intelligence that is accepted at a particular point in history.

Humans v. algorithms: Search edition

ReadWrite Web has posted a rough transcript of a conversation between the advocates of rival search-engine technologies: Wikio founder and Wikipedia co-founder Jimmy Wales and Mahalo’s Jason Calacanis. Wikio is an attempt by Wales to create an open, user-generated search algorithm, while Mahalo uses people to power their search results. It’s a pretty interesting conversation about two of the most interesting projects trying to take down Google.

Update: Wikia is the search engine; Wikio is something else (thanks, Jim).

’Nother update: Apparently, Wikio, which was originally attributed to Wales in the story I link to above, is a startup by former Netvibes CEO Pierre Chappaz. ReadWrite Web has corrected their post.

The Matrix Fallacy: Know-what ≠ know-how

I remember an episode of Alvin and the Chipmunks from when I was a kid, where, for some reason, the ‘munks were involved in a baseball game. At a crucial plot moment, Simon was called up to bat, and the drama of the scene came from the fact that Simon wasn’t athletic (for those of you not familiar with the characters, Simon, on the far left, was the nerdy one). However, before Simon stepped up to the plate, he took a quick moment to work out the physics of the ball’s trajectory (I can’t remember the details, but if you want to optimize the distance of a projectile, you should launch it at a 45º angle; to go a specific distance—say, over the outfield wall of a baseball field—you just need to know how much force to put behind it) and he then promptly stepped up to the plate and knocked the ball out of the park.

Even though I couldn’t express why at the time, I knew that wasn’t right. Simon’s baseball heroics reminded me of those scenes in The Matrix where the characters have skills—karate, flying a helicopter—imported directly into their brains through the data ports in their heads, the implication being that the mere fact that they have received this information makes them able to physically perform specific tasks. However, know-what does not equal know-how.

I know there are examples where people have brought physical processes into coordination with abstract information; one example that comes immediately to mind is musicians with perfect pitch. With the right kind of training, a person can be taught to distinguish individual notes purely by sound or hit a ball with so many pounds of force or at a particular angle. However, the key word here is “training.” The knowledge wouldn’t be sufficient to the skill, for the skill could only be acquired through bodily training.

It seems far more likely that physical processes like hitting a baseball or kicking Agent Smith’s ass are dependent on embodiment; that is, the process is learned through the combination of the mental and physical systems of the body (see Varela, Thompson, and Rosch’s The Embodied Mind). For that reason, merely having information about something does not necessarily translate into having the skill—or know-how—to accomplish a task.

I started thinking about this recently because of Bionic Woman. On an episode from a few weeks ago, Jamie has to stop a whirling fan so Dr. Burke can practice his karate moves on terrorists. As she prepares to grab the spinning blades, we get a shot of her bionic vision, and there is a readout showing the fan’s rotation speed in m/s*s.

This struck me as another example of the Matrix Fallacy. Knowing how fast a fan is spinning doesn’t necessarily translate into knowing when to reach in and grab the blade.

I suppose you could argue that Sommers’ bionics have solved this problem for her by interfacing between her physical systems and abstract ideas like m/s*s. However, this kind of abstract processing has been the goal of AI since its inception, but has so far seemed impossible. At any rate, it would be more interesting if the show illustrated the bionic woman’s skill in ways that made more sense in light of embodiment.

Predicting the future

A few weeks ago, Richard MacManus at Read/Write Web posted a list of ten future web trends. At the time, I didn’t take much notice because the list was pretty standard and not very interesting to me. However, yesterday MacManus posted a follow-up article where he listed trends that users had mentioned in the comments to his previous post. It’s interesting to compare the two lists; for instance, MacManus lists boring web trends that will never happen (the semantic web), while his readers list pretty interesting web trends that will never happen (intelligent agents).

All jokes aside, both lists are also intriguing for their attempts at prognosticating the development of future technologies. I’m always fascinated by this behavior: on the one hand, the prognostications are always somewhat off, but, on the other, futurism has an effect on what kinds of technologies are developed. MacManus mentions the history of AI in his first post, noting that it has been a goal of computing since the 1950s. However, there is been practically no progress in the field. Most of the examples MacManus lists are actually human intelligence being connected through computers, as with Amazon’s Mechanical Turk. The only true attempt at AI that is mentioned is Numenta, which is attempting to build computers based on (what sounds like) a connectionist, neural network model. I don’t think that this model is much of an improvement (at least as far as AI is concerned) on the cognitivism model which computers are based on now, so I would be surprised if it led to true AI, but it is an attempt at something new. The point is, the holy grail of AI has been desired for years, even though it hasn’t been practical to apply, because of the effect of the kind of future prognostication that is represented by these posts.

Using this lens, the two articles represent lists of what people desire the future to be. Fortunately, both are a bit more positive than Richard Watson’s chapter on the future from his forthcoming book Future Files: A History of the Next 50 Years, where he sees a future of depressing loneliness and disconnection. MacManus and his readers are much more positive, seeing a future where the web will deliver new systems to improve people’s lives. Simply by virtue of the fact that both versions of the future coexist right now, the new web that emerges will likely contain elements of both.

Complexity and AI

As these two recent articles—“AI Reaches the Golden Years” and “Brainy Robots Start Stepping Into Daily Life”—suggest, there is currently quite a bit of interest in the development of artificial intelligence. How to implement true machine intelligence, though, is still an open question.

The Wired article points out a problem AI has had since its inception: how to deal with ‘common sense.’ Even though computers like Deep Blue rock at chess, because of the complexity involved in modeling this kind of fuzzy knowledge, programming machines to do what are considered to be normal, everyday tasks is extremely daunting.

Deep Blue: “Bring it on”

In his 1994 book Complexification: Explaining a Paradoxical World Through a Science of Surprise, John L. Casti points out this difficulty with what he calls “top-down” AI models; that is, models that attempt to program in all the environmental factors that will affect the AI. This is the method that has a hard time with mundane tasks. Another, more profitable, method appears to be modeling systems from the “bottom-up,” that is, mimicking the process of the brain and allowing complex behavior to originate from those interactions.

This method, too, has its difficulties. Casti notes, and plieb has pointed out, that there is good reason to believe that to actually produce brain-like activity, a device “must also share the size, connective structure and initial configuration of the brain” (160). Though such a thing may be possible, Gödel has suggested that if it were to evolve, it would be too complex for us to understand, much as our own brains’ functions remain a mystery to us (Casti 167).



Gödel’s solution is supported by attempts at creating A-life via cellular automata (CA) like Conway’s Game of Life (an example of a “glider gun” CA from the Game of Life is pictured above_. Following Steen Rasmussen’s rules for what A-life must look like:

Postulate 1: A universal Turing machine can simulate any physical process.

Postulate 2: Life is a physical process.

Postulate 3: There are criteria by which we can distinguish between living and nonliving systems.

Postulate 4: An artificial organism must perceive a reality R*, which for it is just as real as the “real” reality R is for us.

Postulate 5: The realities R* and R have the same ontological status.

Postulate 6: We can learn about the fundamental properties of our reality R by studying the details of different R*s. (Casti 168-69)

A Game of Life board that could execute such a CA would be roughly 3 square kilometers in size (Casti 228).

These results to not mean that artificial life is out of the question. Casti suggests that the correct response to Gödel’s statement would not be to build AI, but to “grow” it using a bottom-up approach. This would allow for simple processes to form an aggregate—the final form of which, as noted above, would be too complex for us to completely understand—that would exhibit life-like behavior. This system would exhibit the properties noted by Stuart Kauffman in cellular interactions; that the “unimaginably complicated network of interactions” occurring in the cell don’t “lead to utter chaos, but rather results in the cell organizing itself into stable patterns of activity appropriate for its particular function in the organism” (Casti 267). Taking advantage of this spontaneous order is what the NYT article calls “cognitive computing” and falls under the heading of complexity. If such life is ever achieved, perhaps it will be Gödel’s standard—that we can’t understand it—that will validate it as being “alive,” rather than any other arbitrary list of behaviors.