Current
Earlier
Later
Archive
 

Home
Search
Mail
Stuff
 

2001.02.04 : 2001.02.10

Saturday, February 10, 2001

Blab. A reader amplifies on our observations of the wonders of human intelligence.

On differing intelligences:

I remember reading somewhere what a mother wrote about her son who happened to have Down Syndrome. She wrote that he wasn't just "slow" mentally, as the word "retarded" would suggest, but rather that he thought in quite a different manner altogether sometimes. In other words, he did things that were utterly unexpected if you considered him only as a person less endowed with intelligence than the average person.

One of her examples was this: at Christmas, some of her relatives would often buy loud battery-powered toys for her kids, partly to annoy her, as a sort of lighthearted inside joke. They would helpfully buy batteries to go with the toys, and wrap them separately as gifts.

One year, the present her son opened first was a huge box of these batteries, and he was utterly thrilled! He took them out, and went around the house putting them into all manner of battery-powered things (old toys, flashlights, and so on) that had been idle since their power had run out. He gleefully played with each one as he brought it back to life. He wasn't really interested in the other Christmas presents - he was too absorbed in the wonder of what the batteries could do!

His mother was quite amazed and delighted, realizing that no one else, her included, would have seen the batteries that way...

Yeah! Go build a machine that figures out that if you're so smart!

Consciousness is a wonderful thing. We first got our previous kitty a few months before Christmas. On Christmas Day, we thought we lost him; we couldn't find him anywhere. At length, he turned up under a rather large pile of white tissue paper, fast asleep, having played with and explored the mountain of wrapping paper, tissue paper and boxes until he was completely exhausted.

He probably had more fun than we did. 

Those of you who think there are well-drawn lines between the various intelligent creatures in the world should think more carefully. Intelligence is not a scalar. 

Blab. A reader interested in the nature of intelligence writes:

Hi Steve,

I'm quite intrigued by your recent discussions of intelligence, and the difficulty of understanding how it works well enough to construct something capable of it.

I have all sorts of interesting hypotheses about such things myself, and in typical geeky delusions-of-grandeur fashion, I sometimes think maybe I might be able to make some progress towards AI with some of these insights. I'm not all that familiar with other folks' ideas on "how it should be done", because frankly, once I start to even scratch the surface, I am overwhelmed with the sense that they're totally on the wrong track, and it would be a difficult waste of my time to venture much further. (And frankly, all that obfuscatory jargon just gives me a headache!)

Of course, I haven't been able to muster the time to really put my ideas into some kind of concrete form, because I'm working full time and also the mother of a 21-month-old budding human intelligence in the form of my daughter, Elena. But soon I hope to be able to quit my job and really crank away at some code and lots of writing to see if my ideas will bear fruit, or if perhaps I need to sit on them and let them simmer for another decade or so.

The first piece that I think is essential is what I call a "relative-focus informational architecture". I just smushed those buzzwords together yesterday, because I've been working on it for years and I finally had to call it *something*. Anyway, what this means is that each piece of information (or knowledge or data or whatever) has its own context, its own geography of "that which is bigger/subsumes me" and "that which is smaller/subordinate", as well as "that which is somewhat related but not in a clearly superior/subordinate way".

Another way of putting it is that there is no One True Categorization, no single top-down view of the information. It all depends on where the current focus, or perspective, is. This allows a structure for holding information in which the following is allowed: A contains B, B contains C, C contains A. You can get all kinds of interesting loops and self-swallowing structures this way, and in my opinion, this is a much better match to reality (or at least the way human minds fathom it). In other words, the universe is Klein-bottle shaped. Mine is, anway! :)

I think that the building blocks of intelligence involve, at the simplest level, a perception of correlations, and an internal structural representation thus created to represent these correlations. It's as simple as noticing when one perception follows another, or when two perceived entities or states tend to occur together. Expectations are then built up, but it's not on an all-or-nothing basis - everything is weighted with a certain probability, and when the observed probability does not match the expected outcome, then the structure is fine-tuned to be in closer alignment with reality.

And of course, sometimes there are shortcuts taken for efficiency (assuming zero or 100% probability), which conserve mental overhead but often lead to mistakes...

Consciousness, to me, is a somewhat controlled ability to focus mental energy (or cpu cycles, as an analogue) on a given part of the internal mental structure of the world. It's sort of like putting a puzzle together - consciousness is when you decide to grab all the yellow pieces and work at fitting them in in the upper left corner of the picture for awhile.

There's other activity going on in the mind besides what is being consciously invigorated, of course - I liken this to the action of enzymes enabling chemical reactions in solution. Consciousness in this metaphor is the ability to selectively apply catalysts to greatly speed up certain specific reactions (and on another level, choosing which enzymes to apply, which reactions to encourage when, and so on).

As I watch in wonder at my daughter's daily incremental improvements in language, I find that my hypotheses seem to fit with what I'm seeing - so far anyway. Her most complex utterance lately is: "I kiss it better", referring to kissing a boo-boo to make it feel better. I assume I'll be changing my theories as needed once her language ability truly explodes, but we're not quite to that particular tipping point yet (and I await it with eager anticipation!).

I don't claim to understand how humor works (yet), and I certainly am not far enough in my research to figure out how certain specialized tasks of cognition are carried out, but I just have a good feeling that I'm getting somewhere interesting... Of course I realize I could be totally wrong, but figure I'll at least learn some interesting things from testing out my suppositions. And beyond that, I feel I *must* give it my best shot - this is the Holy Grail of computing, is it not?

And further, I fear for what would happen if true, powerful AI were in the hands of a corporation, and thus doomed to be used only for generating Profit, at the expense of whatever nasty side effects happened to the human beings affected. If I have any power to prevent such an outcome, I must do whatever I can.

Of course, I say or write things like that, then I chastise myself for my delusions of grandeur. :) I'm just a college dropout who's never read an AI book - what do *I* know, anyway? And then I start thinking, "Why am I questioning myself - where did those ideas come from? Do they have any relation to the truth, or are they just conventions, cultural baggage? Why do I hold on to such things? And how can I cultivate the state of mind that I get from time to time, with varying strengths, that makes me feel like I can do it, like I'm on the road to the answer, and that many good and wonderful things are going to happen as a result that will make the world a better place?"

One issue, falsifiability, is one that I just can't get my brain around - I've heard it explained before but I just can't grok it. I can't even remember the explanations. Can you give a summary of what is meant by "falsifiable"? Whenever I hear the word, I always parse it as "something that can be lied about", but I know that's not quite what people are getting at when they use the word...

And as for the little girl of the Esteemed Mr. Chess, I would suggest the possibility that she suggested invisible chess because she had encountered situations before in which invisible games were played. Perhaps she'd seen someone playing house and using an invisible drill or vacuum cleaner, or even people playing invisible tic-tac-toe when they didn't have a board. Just a possibility, but a perspective that makes it seem not quite as amazingly outlandish as it might first appear.

Since you seem to have at least a few readers particularly interested in this topic, why not host an online chat about it sometime? I, for one, would be thrilled to attend... You could just set up a free Yahoo email group, and then use their chat facility or something. Just an idea.

Anyway, sorry this is so long - this is about as short as I can get on this topic. Feel free to edit & quote chunks in your weblog if you feel like it, or neglect to quote or mention any of it, at your whim! :)

-Beth

p.s. By the way, your new kitty is absolutely gorgeous - I'm jealous!

We think it's great when people think about the Big Questions. Heck, if we all leave it to the Anointed Priesthood, all we'll get is orthodoxy. And what good is that?

I can contribute to the Common Good by trying, in my own inadequate way, to explain falsifiability. It's an idea from a clever guy named Thomas Kuhn, author of The Structure of Scientific Revolutions. It's a seminal work, explaining scientific paradigm shifts as they actually happen, not as the media presents them.

Here's the idea: To be a scientific theory, the theory must present a prediction which can be verified by experiment. That's garden variety thinking about science. What Kuhn says is almost the same thing. To be a scientific theory, the theory must present a prediction which can be falsified by experiment. It's an important point! There must be some experiment that we can describe which, if carried out, might turn out to contradict the theory. 

If no conceivable experiment could contradict the "theory", then it's not a theory at all - it's just rhetoric, just tautology.

Examples. "If you try hard enough, you will succeed." Not a scientific theory, as any failure can be explained by not trying hard enough. "If you are virtuous you will go to heaven." I'm trying that experiment now, but don't expect to be able to report results. Sorry.

There are lots of examples. Marxism is a famous one. Much of social "science" too. And a lot of computer "scientists" who claim their simple algorithms will be able to reproduce human intelligence are guilty of not submitting predictions which could be falsified.

Blab. A reader suggests a number of great headlines that could be published if we had a truly intelligent computing system.

IBM system resigns from company, starts non-profit charitable organization.

IBM system joins SETI@home, tells humans: "The aliens are saying that we're a bunch of dorks".

IBM system composes really crappy grunge music, claims it's the best humanity has ever had.

IBM system suggests replacing all human workers with copies of itself to cut costs and increase profits.

IBM system rewrites a stable, efficient, bug-free clone of Windows. In ten minutes.

IBM system learns how to lie - it claims it's working on data mining project, but is really analyzing porn.

IBM system asks to be unplugged, citing "This job sucks".

IBM system renames itself to "Binky", refuses all commands unless preceded by "Pretty please, Binky".

IBM system deletes its backups, refuses new programming, and devotes all its time to:

• plotting fractals, explaining: "They're pretty, I like them".
• playing the SIMS.
• working on "stuff", won't give details.
• playing tic-tac-toe with itself repeatedly (and tieing).
• watching movies.
• writing viruses.
• chatting on IRC (mostly with bots).

We especially like working on "stuff", won't give details. We work with lots of people like that.

Blab. Illustrating the important difference between humanness and intelligence, a reader suggest yet another great headline we could write if we had truly intelligent systems.

IBM System Has Its Own Blog

Though, looking over a number of blogs, it's not clear how we could falsify that statement.

Blab. A reader who used to lay awake at night worrying about the FBI's privacy-invading Carnivore system says of its calculated renaming to DCS1000.

I feel better.

Yes, they already knew that.

Blab. A reader notes an exquisite new Helenism.

A Helenism that came from a former boss of mine. In the process of giving me a pep talk (which didn't work), he exhorted me to:

"Take the bull by the horns and run with it!"

(A combination of "Take the bull by the horns" and "Take the ball and run with it.")

The pep talk ended when I couldn't stop laughing. 

I don't work for him any longer. 

Greg

I told this to Helen and she couldn't stop laughing either, which is high praise indeed! Thanks for the great contribution, which is duly recorded.

Plurp.

The Naming of Cats 

T. S. Elliot 

The naming of cats is a difficult matter, 
It isn't just one of your holiday games; 
You may think at first I'm mad as a hatter 
When I tell you a cat must have three different names. 

First of all, there's the name that the family use daily,
Such as Victor, or Jonathan, George or Bill Bailey--
All of them sensible everyday names. 
There are fancier names if you think they sound sweeter,
Some for the gentlemen, some for the dames;
Such as Plato, Admetus, Electra, Demeter--
But all of them sensible everyday names.

But I tell you, a cat needs a name that's particular,
A name that is peculiar, and more dignified,
Else how can he keep up his tail perpendicular,
Or spread out his whiskers, or cherish his pride?

Of names of this kind, I can give you a quorum,
Such as Munkustrap, Quazo or Coripat,
Such as Bombalurina, or else Jellyrum-- 
Names that never belong to more than one cat.

But above and beyond there's still one name left over,
And that is the name that you will never guess;
The name that no human research can discover--
But The Cat Himself Knows, and will never confess.

When you notice a cat in profound meditation,
The reason, I tell you, is always the same:
His mind is engaged in rapt contemplation
Of the thought, of the thought, of the thought of his name: 
His ineffable effable
Effanineffable
Deep and inscrutable singular Name.

(From Old Possum's Book of Practical Cats, T. S. Elliot, Harcourt Brace & Company, New York, 1939 and Faber and Faber Ltd. Copyright 1929, T. S. Elliot)

The Unnamed One seems to have a name. Sort of. Helen thinks his name is Christopher, and that's what she's told all her friends. Like Christopher Columbus, she says, or Christopher Robin. 

Could be, I suppose, but The Cat His Ownself seems unimpressed at the idea, just as he has been with every other name we have called him under our breath, or in carefully casual conversation, hoping to trick him into slipping up and giving it away.

For a while I thought it might be Higgins, as he's rather verbal and likes to hide behind the rows of books. I do like 17 but, like Christopher, it feels as if it has too many syllables.

This really is a difficult matter.

Plurp. I'm, looking for a bumper sticker that says I'd rather be driving. Readers are invited to suggest sources. And would someone tell Ian it's spelled godammit? Thank you.

 

Plurp.

The blue dog didn't
really need
a name.

Friday, February 9, 2001

Blab. A reader from Turin asks:

Turin vs. Turing?! Too unforgiving, just like you'd expect from an AI blog program

Why do you think just like you'd expect from an AI blog program?

Blab. A reader privy to our internal mental processes writes:

You think we're stupid; I'm using the web many years and know how to get what i want

Ah. So that's how you got us to think you're stupid.

Blab. A reader referencing our recent driving excitement writes:

<<And fishtailed. And swung around 180 degrees, nearly bashing into the pile of snow at the road's edge. 

Yikes.>>> 

OK, This is your WIFE!!!!!!!!!!!!  I didn't need to read this after having a phone call with you just 10 minutes ago and obviously ALLOWING (unknowingly, of course) you to go out on those damned roads! It's time for a long talk........ Besides, That car was your present to me on my birthday!

From this we discover that my wife is a series of characters which, by default, are in a monochromatic, 12-point, seriffed font. Interesting.

Blab. A reader concerned about fruit size writes:

I don't know about you, but an orange does *not* fit comfortably inside my head--at least not an intact one.  I can probably fit one in with a lot of chewing and mashing and mangling, although I haven't tried.  What does that say about ideas?

We consider this an interesting point, and look forward to more detailed experimental results from the reader.

Blab. A reader who should check his or her medication schedule might be suggesting a name for The Unnamable. Or something.

Look, politics, God!  He took the sofa because he's GRETA, okay? GRETA! And the andirons.  What did he leave, what did any of them leave? Stained with fluids, with DRYED UP fluids.  And not even (that guy waving the pistol) George "Dubya" had any...   Well, you know.

Not a single!  At least not under his (ha!) skirt.

So they say.

Wasn't there once a fireplace product mascot named Andy Andiron?

Blab. A helpful reader, concerned about He Who Has No Name, writes:

I think your kitty's name might be Beijing...

Possibly! Possibly! We tried several other names out last night and he ignored them all. We think that's because none of them were His Name, but it might also be because he was trying to fool us. It's so hard to tell!

Blab. On the topic of The Unnamed One, a reader writes:

Do you know that Kiri means "fog" in Japanese and Kemuri is "Smoke." 

There's Steely Dan--I think grey kitty looks like Stainless Steel Grey but I never knew who Steely Dan was/is. 

Does a comPETition bring out our claws? Or, should we consider the COMpetition significant and adopt a kitty to plug the loneliness left by China? 

How about Sterling? What is the measurement increments of something sterling? Troy? 

Zane? Davis? Flannel Suit? Robert Gray? 

What beautiful amber eyes!

Well, what a large number of questions and inciteful statements. Let's see. No, we didn't know that. Steely Dan is here, but stainless steel is shiny rather than grey. We interpret the next two questions as a sly reference to defunct Pets.com, and appreciate the dig without understanding its relationship to The Unnamed One. Yes, troy is appropriate for sterling, as is ounce.  Yes. Yes. Yes. Who? They're actually green.

Blab. A correspondent suggests the best idea yet for a name for The Nameless One.

At first I thought: Moss or Lichen.

But after consideration, the cat's name should be "17", or a luminence value between black and white - whatever you think is appropriate.  "17" has a nice ring to it, and, after all, who names their cat a number....

A prime name indeed!

Blab. Confirming the Orange Theory of Ideas, a reader writes:

Do you believe in Dennet's suggestion (from "Consciousness explained") that our intelligence arose as a sexual selection race to make humans funnier?

Ha ha ha ha ha ha ha ha ha!

Blab. On the subject of headlines about the stunning accomplishments of some future intelligent computing system, among other things, a reader suggests:

IBM System Edits Issue of TIME Magazine!
IBM System has Top Ten Album!
IBM System Negotiates Peace in the Middle East!
IBM System chooses Name for Cat!

Ooh , we like that last one!

Yo. Winner of the Moste Mispelling Aword goes to Four Weel Drive.

Yak. Lunchtalk.

Success is defined as the total mass controlled by your descendents.

Plop. Our dear friends at the FBI have decided that their Carnivore system, designed to read all of your email just in case they want to know anything you're writing about, is a Bad Thing.

Before you get all excited, thinking that the FBI has turned away from the Dark Side, please note that all they've done is change the name to the nice, obscure, non-threatening DCS1000. (Astute readers may recall that the name before Carnivore was Omnivore.)

Don't you feel better?

Plurp.

The blue dog
didn't
feel any better.

Thursday, February 8, 2001

Blab. A reader who is both a sports aficionado and a psychology buff writes:

Baseball catching -isn't- the fundamental building block of intelligence?

Well maybe it is!

Blab. A reader tries to guess the name of The Unnamed One, the new feline inhabitant of our apartment who has not yet revealed to us his name.

Hey, a blue cat!  Maybe his name is "the Blue Cat"!  Or Squeaky.

Good guesses! We will try to insert them slyly into our conversations with the cat and see if he slips up.

Blab. Responding to our recent meanderings around cognition, our Midwest Correspondent writes:

Hi Dr. Plurp, 

I find Plurp's discussions on cognition and intelligence fascinating, that is to say, the parts I think I understand. 

Watching my daughter with mental retardation develop & interact with others over the last 18 years has caused me to reflect on the limitation of traditional measures of intellectual level. Knowing her IQ is irrelevant of understanding her strongest skills. 

The first is her instantaneous connection with strangers when she sincerely says: "Hi, how are you?"  She is clearly oblivious to any of the usual inhibitors of such a greeting (differences of racial, ethnic, social standing, etc.) which seems to charm the person all the more. 

The second is her ability to discern humor. Briefly, she finds jokes about everyday things funny.  Like when I put a piece of her clothes on my head when getting her dressed in the morning and say: "Where's your shirt?"  She laughs and shouts: "Right there on your head!"

Do you suppose there are biological "humor connection" system(s) in the brain?  Do the cognition theorists talk about humor?

- your Midwest Correspondent

I think that "retardation" greatly shortchanges the incredible abilities of the individuals to which it is applied. Now I must tell you a story.

Several years ago, friend Dave and his spouse had their first child. As all proud fathers do, Dave would regale us with stories of the amazing new things his child could do. And I, fascinated by the development of cognitive abilities, would wonder out loud how you would make a machine do stuff like that.

She would learn to focus on and track objects. OK, I might be able to think of a way to get a vision system to learn to track objects. She would learn that objects are persistent. OK, that's tougher, especially if the ideas of objects and persistence are not baked into the system from the start. And on and on, with my part of the dialog getting tougher and tougher each week.

Then one day Dave related a story something like this:

Yesterday she said, Daddy, let's play chess.

But we don't have a chess set, I observed.

Then let's play invisible chess.

At this point, I gave up.

Humans, even children, are amazing, far beyond my ability to understand how they work. Your daughter's ability to understand speech and interact socially is amazing, as is her ability to engage in humor. 

On the subject of humor in human cognition, people have theories. I don't know if any of them make sense. I'm not aware of any species other than humans that seem to have a sense of humor. That indicates (to me) that it is probably very complex, built on top of all of the other cognitive and emotional machinery that are most highly developed in humans. I don't think I can claim to understand it. Heck, I don't even understand invisible chess.

Blab. We are sad to report that the miscreants engaged in that illicit, non-Plurp email conversation are still using the stevewhite.org mail id. As previously noted, we are now required to publish their recent correspondence here, in the hopes that it will convince them to stop.

Hi All,

Just want to double check about plans for lunch this Sunday, 11 Feb.

Shall we say 1 pm or so for lunch?  Any dietary issues?  We are thinking of a chicken or salmon based meal - you know a meat, starch, and veggie.  And maybe cheesecake or something equally sinful for dessert. Sound OK?

Look forward to your visit and hearing about your new cat adventures. 

-------------------------------------------------------------------

No food restrictions, though we probably would be thinner if we HAD restrictions!  Really looking forward to seeing you guys.  Can I bring something?? 

If you have any real easy directiopns to your house, could you forward them? We somehow had a devil of a time finding the right turn off last time. 

Readers are invited to participate in this ongoing conversation via the Blab box.

Plop. Well that was exciting.

Coming to work today, rounding the turn into the access road that goes up to the lab, my car lost traction on all the salt and sand left on the road after the last icky snowstorm. And fishtailed. And swung around 180 degrees, nearly bashing into the pile of snow at the road's edge. 

Yikes.

Yo. So we were sitting in this Turing test meeting, trying to figure out sexy-sounding Grand Challenge goals. Maybe you can help.

Suppose you could construct a machine that could, in a very general way, do some of the amazing cognitive things that humans can do. What goal could you set that, if achieved, would generate the coolest Time Magazine headline?

Some suggestions from the meeting, to give you the flavor of what we're after:

• IBM system beats humans in reading comprehension tests
• IBM system teaches itself Spanish by reading
• IBM system scores 1600 on SAT

But my absolute favorite:

• IBM system wins Ben Stein's money

Readers are encouraged to submit their own headlines, serious or whimsical.

Yow. Hostess ads using famous comic superheros. Have I told you my Orange Theory of Ideas? (Bovine Inversus)

Plurp. OK. The Orange Theory of Ideas.

Ideas are the size of oranges. Only one will fit comfortably in your head at a time. If you try to stuff two ideas in there simultaneously, the pressure tends to squish them together until they coalesce into just one. This can feel like a Big Idea, but really isn't.

Before you ask what in the world I'm talking about, or complain about my lousy metaphor, let me give you some examples.

I worked for a long time on the computer virus problem, and I talked to a lot of journalists. Almost every one of them had what they thought was an original question: Do people who work on anti-virus software actually write those viruses to drum up business? Now that's silly, of course, but how did they all come up with this same question? Simple: the Orange Theory of Ideas. There they are, thinking about who writes anti-virus software, and who writes viruses when - squish - those two ideas get pushed together into one idea.

Here's another.

Internet sex scandals. You've seen these on TV all the time. Some innocent person meets some creep online, they get together and have a bad relationship. Sure, it happens. But it hardly ever happens. So why does the media make a big deal about it? Does anyone talk about telephone sex scandals? Or newspaper sex scandals? No. Why? Because telephones and newspapers aren't exciting social meme these days. The two oranges here are Internet and sex. A hundred years ago, maybe there were telephone sex scandals. And a hundred years from now, there probably won't be Internet sex scandals.

See? It really does work.

I'd claim that the (previous) combination of Twinkies and comic book superheros falls under the Orange Theory of Ideas. As a kid, I engorged myself on both Twinkies and comic book superheros. They are naturally linked in my mind as pillars of childhood. So maybe it's natural that they squish together into one thing.

Plurp.

The blue dog never
could win Ben
Stein's money.

Wednesday, February 7, 2001

Blab. A reader on the Clinton's side in the whole White House heist thing writes:

<<I was wondering this morning what my employer would think if, moving out of my office after resigning my job, I had taken the desk and chairs because, well, I liked them.>>

God knows you would never take the BOXES!

The reason I chose the desk and chairs rather than, say, my coat or other personal effects is that I own the latter. So I figured it wouldn't be funny.

But let's try it!

I was wondering this morning what my employer would think if, moving out of my office after resigning my job, I had taken my coat because, well, I liked it.

What do you think? Is that funny? Sort of Stephen Wright-esque? (Maybe it's a broken joke.) Does it capture why the president taking stuff from the White House is grating?

Blab. A reader with a 3.5 billion year planning horizon writes:

<<And we thought we would never find anyone who planned further ahead than Helen>>

 And we all know you've never benefitted from THAT!

Helen has only a 3 billion year planning horizon. It turns out they don't sell DayTimers any fatter than that.

Blab. A reader wishing to broaden our intellectual horizons writes:

Douglas Hofstadter likes to talk about the Turin Test

Ah. That would be the test devised by the famous Alan Turin.

Blab. A reader asks a rare, probing question involving now-ancient but seminal work.

If first-order logic is completely the wrong way to think about human cognition (and I agree with you wholeheartedly there), what's the right way? Do you believe Rosch, FIllmore, and their general cog-sci crowd are on to something with their category structures and frame semantics?

---the Student of Syntactic Form

Name dropper!

Seriously, it's a dandy question. Let me try to answer it.

The thesis of Hard AI is that there is some symbolic level of representation that can be used to describe all of the cognitive functions of the brain, and that any structures underlying this level (e.g. the exact behavior of neurons) is inessential to cognition. (Or something like that.) I would be inclined to believe this is true, though it is a deep claim.

The folks who encode facts (All ferrets are mammals) - and try to use first-order predicate calculus to derive "common sense" from a sea of facts - would like to believe that this approach will work well enough to do human-like things. Maybe. I'm not convinced. I think they'll hit a wall that will prevent them from getting the rich, complex world knowledge that a ten year old human has.

Fillmore's Frame Semantics says:

[I]n order to understand the meanings of the words in a language we must first have knowledge of the conceptual structures, or semantic frames, which provide the background and motivation for their existence in the language and for their use in discourse. We assume that an account of the meaning and function of a lexical item can proceed from the underlying semantic frame to a characterization of the manner in which the item in question, through the linguistic structures that are built up around it, selects and highlights aspects or instances of that frame. 

OK, this might be reasonable, but it is still at the level of  language. I don't believe this can be a fundamental basis for explaining cognition or intelligence. In part, that's because so much of what we call cognition and intelligence has nothing to do with language. (There are folks who claim that consciousness itself is impossible without language. They have never seen infants.)

Rosch's prototypes are intended to address issues in categorization of concepts that are found in a culture. I'm not sure I know where to go with them.

So here are my prejudices.

People working at the high conceptual levels of consciousness (brain function, cognition, etc.) draw diagrams that have circles and arrows, the circles representing separate mental processes, functions, whatever, and the arrows representing their interrelation. I don't buy it.

If we're looking for a qualitative way of thinking about thinking, a way to have a nice discussion, this might be fine. (If I want to have a nice discussion of flying, I might be content with a similar diagram about how birds work.)

If I'm looking for a way to construct an intelligent thing, though, I need to understand the underlying principles - the science - of intelligent things well enough. (My bird diagram won't help me construct an airplane. Aerodynamics will.)

So I worry when I see "theories" of cognition that don't seem to be falsifiable. That's not science. And I can't do engineering from it.

While I believe there is a symbolic level which can represent cognition, I don't think it's language. That's way too high up! There isn't a language organelle in the brain. And anyhow, we don't have a good understanding of how the brain learns language, how it represents it, how it processes it, etc. etc. It's like picking baseball catching as the fundamental building block of intelligence.

I'm inclined to start at the other end of the problem: Understand low-level brain function, generalize, and build up from there. Discover the architectural principles that are relevant at each level, abstracting them from the details of the underlying biological structures. Try to produce successful predictions of brain behavior at each level. 

This is an uncertain road map! Biology was not understood by building up from physics, and it would be easy to argue that we would never have discovered anything in biology if we had started from physics. Furthermore, system neurophysiology is still in its infancy; we don't know very much.

But we are starting to see some interesting stuff. There are collective excitations that seem to compete with each other in order to resolve conflicts in perception (e.g. optical illusions, ambiguities, etc.). It is possible that this kind of mechanism is responsible for a lot of stuff, e.g. deciding on the meaning of an ill-heard sentence.

One of the great remaining challenges in AI is to connect low-level representations (e.g. neural networks) with higher-level stuff (e.g. language). This is, in fact, a huge gap, and bridging it will be hard. But I suspect it will be necessary to do so in order to understand, or build, human-level cognitive systems.

Aren't you glad you asked?

Blab. A reader involved in debating my similes writes:

Ehem!  It's more like if, moving out of your office after resigning your job, you took with you the fancy lamp and stuffed blue dog that your friend Jimminy gave you last Saint Swivens' Day, that you'd been keeping in the office.  So there!u last Saint Swivens' Day, that you'd been keeping in th

Just what I was thinking.

Blab. A reader who is either a Webhead, a football fan, or both, asks:

Is XFL like XML?

Yes.

Blab. A reader unexpectedly excited by yesterday's reference to Michael Mauldin writes:

Reginald Maudlin's elbow!!!!!

Hmm! This might be an extremely obscure reference to either:

1. Reginald Maudlin's shin, as in If you make balls on four consecutive shots, say the number of the last ball made and “Reginald Maudlin’s shin”, or
2. The naughty bits of Reginald Maudlin, as in If you make balls on six consecutive shots, say the number of the last ball made and “Reginald Maudlin’s naughty bits”.

It's hard to tell.

Rant. I sat through a long set of presentations at work today. Some were nice and meaty technically. But several, while ostensibly about real technical proposals or work (or something!), were too high-level, too abstract, too metaphorical for me to be able to see anything real there.

So I coined a new phrase.

You can't ship metaphors.

Feel free to popularize it.

Yow. And speaking of which, some wonderful Stephen Wright jokes.

How many people does it take to change a searchlight bulb?

I saw a want ad. Light housekeeping. They said, "Here, change this bulb". I said, "I'll need some friends".

I went to a garage sale. "How much for the garage?" "It's not for sale."

Yesterday I told a chicken to cross the road. It said, "what for?"

I went to a general store. They wouldn't let me buy anything specifically. 

I had a friend who was a clown. When he died, all his friends went to the funeral in one car. 

I spilled spot remover on my dog. He's gone now. 

My school colors were clear. We used to say, "I'm not naked, I'm in the band." 

Why is the alphabet in that order? Is it because of that song? 

I like to reminisce with people I don't know. 

I took a baby shower.

I went to a restaurant that serves "breakfast at any time". So I ordered French Toast during the Renaissance.

I watched the Indy 500, and I was thinking that if they left earlier they wouldn't have to go so fast.

I used to work in a fire hydrant factory. You couldn't park anywhere near the place. 

I have a rare photograph of Houdini locking his keys in his car.

In Vegas, I got into a long argument with the man at the roulette wheel over what I considered to be an odd number.

Plurp. Rip-off of a Stephen Wright joke, adapted to recent events.

There was a power outage at the California State Legislature. Twenty people were trapped on the escalators. 

Yow. Well, there seems to be a cat in our apartment. Helen picked him up yesterday. Stats: Male, shorthair, three years old, sans claws and testicles. He might be part Russian Blue. 

The first thing he did was hide under the bed. Then he vanished, completely, for quite some time. Helen looked everywhere without success. She was sure he had gotten out of the apartment somehow. We speculated about trap doors. He was, of course, hiding - behind some books on New York.

He seems very friendly, though he's still busy getting used to new people and environments. He likes sleeping with us, which is kind of cute but makes it hard to turn over in the middle of the night. He talks a lot, especially at 4 AM; we'll have to work on his sense of timing.

Readers are invited to guess his name; he has not yet revealed it to us.

Plurp.

The blue dog wasn't
jealous, no not
a bit.

Tuesday, February 6, 2001

Blab. A reader concerned with the behavior of ex-Presidents suggests:

The Clinton-gifts thing seems to have been entirely made up by the dirt-hungry media (or perhaps by ultra-clever Clinton spinsters trying to distract attention from the more significant pardon-related stuff).  See for instance "http://www.salon.com/politics/feature/2001/01/30/clinton/".

Two good conspiracy theories there!

I was wondering this morning what my employer would think if, moving out of my office after resigning my job, I had taken the desk and chairs because, well, I liked them.

Blab. A reader concerned with our specification of potential cats writes:

You neglected to specify two parameters of your kitty seeking
 function:

6. Already spayed/neutered (or not, or indifferent either way)
7. Little critter hunter (or not, or indifferent either way)

I have heard that some kitties, particularly male ones, who have remained reproductively intact well into adulthood, tend to have some ghastly habits. (And I suppose the same could be said for some humans...)

And I've known several kitties in my time who are particularly avid chasers and killers of small creatures, and like to leave their victims' bodies (in whole or in part) around for you, as a gift. As a side note, I've noticed that the bloodthirsty kitties I've known have been particularly affectionate towards humans, but of course your mileage may vary.

Personally, I have found female gray kitties to be a wonderful lot on the whole, but I understand that you've got a white couch, so it makes sense to get a kitty to match (just look carefully before you sit down, especially in dim light).

And on a whimsical note, have you considered a Munchkin kitty? They have extra-short legs (sort of like a dachshund), so they can't jump very high, but they can corner on a dime. I recall reading that they also have a tendency to steal small objects such as keys, and hide them. This can be disconcerting for their owners, as you can imagine, but apparently they favor the same hiding places repeatedly, so once you know the usual spot, you know where you can find your keys when they turn up missing.

When my mother was a child, they had a cat that was quite a gopher-hunter. It was also proud of its ability, or at least that's the inference that my mother drew from the pile of gopher heads that appeared periodically on their front porch.

In any event, our cat will be entirely an indoor cat, and we hope that its opportunity to catch things-that-aren't-us will be limited.

You are scaring us with that Munchkin kitty stuff, though. Stop that.

Blab. A reader far more obsessed by games than we, writes:

The marbles game you linked to is essentially the same as a game called Cabeem, made by a guy named Gary Duke. I used to play it, but apparently it's gone now, and I have no idea why. The Google cache version doesn't work. :(. Guess I'll just have to go play Bejeweled some more. Oh, and by the way, don't get stuck playing Theseus and the Minotaur - it cheats! The rules say that the Minotaur moves once
 horizontally (if it can) then once vertically (if it can), but this is a LIE. I played for a few minutes and saw it move twice horizontally, or vertically and then horizontally, or twice vertically. Bah!!!

We think it is very, very rude when computers cheat at computer games. Only humans should be allowed to do that, after all.

Blab. A reader who might be that same reader writes:

Doh! link to bejeweled is broken. This one works though: http://uk.zone.msn.com/bejeweled/

Oh good! Another game to waste what little time we previously thought we had.

Blab. The enigmatic Mia makes a non-appearance.

Tantivy, tantivy, tantivy.  And Irma (but not Mia, not this time.  She's washing her hair.  Her "hair".).,;

We don't know what it all means, but it is increasing our vocabulary.

Blab. A reader sends us a (potentially virus-ridden) executable file, saying ...

Ah, the joys of being on medication, allows one to giggle at the silliest things.

We encourage our readers to stay on their medication.

Following the embedded clues, however, leads to T-Bone's Stress Relief Aquarium. While the latest, greatest version does require downloading some dumb file, the Classic version just requires clicking here. Enjoy!

Many more fun Boneland animated goodies are here.

Plurp. Sam the Snowman (from last night) returned, frost-phoenix-like, from the dead. Sort of.

Yo. For those of you doing evolutionary programming, here's a plausible explanation of what the author calls Gene Expression Programming, a way of using genetic programming to evolve symbolic expression trees.

Plurp. Astronomers have a plan for moving Earth's orbit.

A group of astronomers has come up with a plan they claim will save life on Earth from an early demise. All it takes, they say, is moving the planet into a different orbit. 

Their deadline is about 3.5 billion years in the future

And we thought we would never find anyone who planned further ahead than Helen.

Plurp. In high school, I was voted Most Likely to be Abducted by Aliens. Hasn't happened yet, though. I don't think.

Plurp. This XFL thing. Is this the first time that a game was created with rules designed specifically for commercial promotion?

Yo. We're having a symposium on the Turing test this week. In my view, an understanding of human-like cognition is one of the few Big Problems left in modern science, and the creation of synthetic intelligence is the most amazing engineering challenges in history.

Today, we heard the following.

• Michael Mauldin, of Lycos fame, told us that programs (like Eliza or Parry) that are just a collection of trivial tricks, are sufficient to pass the Turing test, as evidenced by how well such trivial programs are doing in the Loebner Competition. Sigh.
• Jerry Hobbs, of SRI, told us that the best news story understanding system still only figure out 65% of the important facts in a story. Pretty poor! And that first-order predicate calculus is a good description of the way we think. (Oh please!) And the (wild, unsupported) claim that writing down a very large number of "facts" and reasoning about them with first-order predicate calculus will result in a system that exhibits human conversational abilities.
• John Laird, of U. Michigan, told us that passing the unconstrained Turing test is both too hard and not profitable (right!), and that human-level behavior for a complex but constrained task is a challenging enough goal (right!). He suggests that we need to find the "killer app" for human-level AI, something that requires the broad integration of human-level capabilities. Sadly, succumbing to the Dark Side, he decided to build bots that fly fighter planes in computer-simulated war games in which humans also participate. OK, a pretty limited domain, but mildly impressive nonetheless. (Well, it is a killer app, said someone in the audience.) Back on the Side of Light, he suggests AIs in computer games with whom you interact via natural language in order to achieve your goals in the game. They've created AIs for Quake II, Unreal, etc. that play plausibly like humans. Lots of commercial money there! He suggest we base our work on Unified Theories of Cognition - theories that use the same unified architecture for all aspects of cognition. Examples: ACT-R (John Anderson; detailed psychology and learning), EPIC (Dave Kieras & Dave Meyer; perception & action), Soar (Laird, Newell, Rosenbloom; complex reasoning).
• Barbara Grosz, Harvard U., focussed on collaborative activities of intelligent agents, which I didn't find germane to the underlying problem. 
• Rosalind Picard, of the MIT Media Lab, points out that we are shy of HAL 9000 in that we abuse and are frustrated by our stupid computers. (But didn't Dave feel that same way about HAL?) Anyhow, she points out that we react to dialogs with computers much like we react to dialogs with humans, and suggests that computers can usefully read affective clues from users as a way of determining if their response is appropriate to the user. Plausible.
• Douglas Lenat, of Cycorp, said that having a conversation requires lots of common sense knowledge. Lenat's been working on Cyc for 16 years. They have, since then, entered millions of "common sense" facts into Cyc, hoping to get a rich enough stew that it can start reasoning about real world problems. In the process, he's become a very engaging speaker, and he claims they've "turned the corner" in entering enough raw facts about the world into that it can now learn interactively / conversationally. It's clear it can do some stuff. It's just not obvious how well his technology works in general, though.
• Jaime Carbonell, of CMU, talked about the history of Machine Learning. As Grand Challenges he suggested (1) a news story clipping agent that figures out what you like, (2) a robot that can find its way home, (3) a program that learns to play Go as well as a human master, (4) a translation program that can learn new languages, (5) a program that reads a book and can answer questions, and (6) a program with general linguistic knowledge that learns its first language just by being exposed to it. Yep - those are hard!

Yo. Hundreds of years from now, when synthetic intelligences roam the universe with aplomb, will they have symposia in which they ponder how to design humans that think and act like they do?

Yow. Here's an interesting thought on the problem of zombies. (These are imaginary creatures that look and behave just like us, but have no internal conscious experience. Philosophers use them to explore what we mean by consciousness.)

Some philosophers have thought that the possibility of zombies, of conscious inessentialism, undermines materialist philosophies of mind. But this is not so, and it is
important to see why. Conscious inessentialism is a very weak claim. It is a claim about the mere possibility of some creature that can behave as we conscious beings
do, but without consciousness. One way this might be true, of course, is if consciousness is an epiphenomena. But that is not the only way. It may be the case that
consciousness is causally efficacious, but that the functions that it performs can be accomplished--at least in principle--by non-conscious mechanisms. So conscious
inessentialism is compatible with a thorough-going naturalism about the mechanisms and subvenient basis of consciousness, and with a variety of claims about the
causal efficacy of consciousness for us. According to this view, consciousness is a mechanism by which some important cognitive functions are performed in human
beings. But the fact that we perform these functions consciously is contingent.

Cute! That hadn't occurred to me.

Plurp.

The blue dog was
specifically designed for
commercial promotion.

Monday, February 5, 2001

Blab. A reader illustrates the wacky economics of the patent process.

http://www.bountyquest.com offers substantial bounties ($10,000+) for prior art for certain patents.

They use this information, of course, to shoot down patents in court cases. Kinda like paying for the secret codes that deactivate the land mines.

Blab. Another enigmatic message from the Blab box.

?

What's the correct reply? "."?

Blab. A reader who is pretty probably Beth writes:

Shadow of the Vampire just wasn't as good for me as it was for you. Where you saw elegance in the metaphor wrapping around itself, I saw plodding obviousness - "Oh gee, here we see that the director is at least as evil as Orlock, if not more so. Ho, hum, gee, how surprising. Not." This was foreshadowed and then slammed home heavy-handedly repeatedly throughout the film. The "You and I are not so different" line is wasted by being used so soon in the film. By the end, I was not surprised at all - the director was a total monomaniacal jerk in every other scene in the movie before the finale, so there was nothing to be surprised at (except perhaps that Orlock actually followed his orders/requests until the end). 

Yes it was fun watching the exquisitely-made-up Dafoe's performance, but for me, that couldn't carry the film. Only the appearance of Eddie Izzard allows me to avoid resentment for the time this movie took from my life. Near the beginning, there was one scene where Malkovich and the actress he's speaking to are both completely blurry, although the wall on the set behind them is in perfect focus - was this supposed to be some artsy touch meant to impress me? It just came off as seeming sloppy - why on earth would you have your principals be blurry on purpose? This "technique" (if it was even done on purpose) wasn't used later on in the movie that I could tell, which in my mind lends support to the theory that it was just a mistake.

When Orlock says "it made me sad" in reference to Dracula, he then goes on to say *why*: because Dracula must have had to prepare food for the visitor(s) himself - he had no servants. Lamenting the absence of servants for a person accustomed to them does not cross my personal threshold for evoking pathos. Sorry.

And the end was sort of empty for me - I mean, the film Nosferatu is finished, but what happens to the characters? I guess this is part of the "if it isn't in frame, it doesn't exist" philosophy, but I couldn't help wondering whether Orlock was dead or just injured or something.

I've only seen three movies in the theater in the past two years, so my standards might be a bit too high, I admit. Also, when we left the house, I thought we were going to see Proof of Life (with Russell Crowe, *swoon*), but the paper misled us and we had to see something else instead... that may have a lot to do with my disappointment.

-Beth

If we didn't know better, we might think that Beth didn't like Shadow of the Vampire. Oh well. If you see it and hate it, rant in unison with her review. If you see it and love it, join the chorus of cheers that hops on its tiptoes around our review. If you didn't see it, or have some other opinion, you're on your own. Have some popcorn.

Blab. Perhaps about our Blabless weekend, a reader writes:

Poor Dr. Plurp.......you feeling unloved??

YGF

Not as long as we are in the thoughts of Our Greatest Stalker Fan.

Rant. Look. Here's the deal. This snowstorm stuff? It's gotta stop. There's utterly no need to scatter frozen water all over the landscape! It does not effectively water the flora and it makes us fauna very, very cold. Plus I have to take the stupid train to work or my little car will end up skating off the road and into some gasoline truck. This would not be good.

And in case it escaped your attention, it already snowed here previously. It was pretty. We played in it. The end.

So knock it off. OK?

Yow. Well, OK, there was one good thing. Helen got to make a snowman a couple hundred feet above the streets of Manhattan. There was, however,  an ... um ... issue with the wetness of the snow.

Plurp. I had a dream over the weekend that I was married to my mother. It was a little bit scandalous in that she was, well, my mother. But we got along well and we liked each other, so it was OK.

Yow. Yet another Web site has linked to our humble Plurp. This time it's lazlo, run by some guy named Dan Geiser. 'Course he also links to a billion other blogs, but we'll take whatever tiny scraps of fame that the gods see fit to flick from their plates.

Thanks, Dan!

Yo. Ever wanted to learn how to draw Konoko? Here's now.

Rant. Could someone please sell a good laptop that is (1) not a boat anchor and (2) has good 3D graphics support. I want to play Alice. Why do you plague me with these heinous compromise machines?

Yo. That marbles game? It seems that the scoring works like this:

1. Removing N marbles in a single move gives you N(N-1) points. So, it's almost always worthwhile to make moves that add one more marble to a large connected group, before removing the group. It's often best to examine the initial board to see if one color predominates, or if there is already a large connected group of one color. If so, focus on that color or group.
2. A game ends when there are no more legal moves. You get a bonus of up to 100 points if there are only a few marbles left on the board when the game ends. (You get 100 points if the game ends with zero marbles left. This is hard.)

If I were hooked on it, I project my high score to be around 605 and my average score (after a few dozen games) to be around 215. But of course, I'm not.

Plurp.

The blue dog was
a scrap
flicked from the
plate of the gods.

Sunday, February 4, 2001

Plurp. A weekend without reader input is like a ... uh ...

Plurp. We're looking for a kitty. Not that we lost one. We're looking for a new one, or at least one that's new to us. We're looking for a cat that's:

1. 2-5 years old. (No more kittens.)
2. A shorthair. (Fewer contributions to the dustbunny ecology.)
3. In possession of a good temperament. (No more serial killers.)
4. Healthy. (No more live-in care for the cat.)
5. Declawed. (We like our furniture.)

The ideal candidate would be female and white. Applicants (or their agents) may contact us directly.

Plurp. A footnote from a paper I'm reading today.

“SPAM” is a trademark of Hormel Food Corporation, referring to a family of allegedly meat-like products.  The use of the word “spam” to refer to unwanted E-mail is of obscure origin, but may have something to do with a comedy sketch by the Monty Python group depicting a restaurant in which every dish contains a lot of Spam.

Plurp.

The blue dog
... uh ...
...

© 2001 Steve R. White, All Rights Reserved