We here at The Censorware Project hate to be controversial, so we'll express no preference concerning our favorite chocolate, let alone our favorite color M&M. But one which many seem to like did attract our attention. You see, for no immediately apparent reason, X- Stop blocks the worldwide web site of Godiva Chocolatier.

We scoured the site looking for why. Could it be because of their all too racy soap opera, "Love & Chocolate In Europe", which begins:

"As the world turns each day of our lives, we tend to feel more restless than young. And why? The four letter word that all our children know better than we do: love. At Godiva Online, the web we choose to weave is always tangled with romance (and a healthy dose of back stabbing). You won't want to miss a single episode of this tale of sweet deceit and culpable chocolate."

Physics toys? You betcha, and with the holidays just around the corner, they make for some darn good gift ideas too. (Disclaimer: none of us have any financial interest in the company, but one of us owns, and loves, one of their toys.)

"FASCINATIONS® Mission Statement

"....producing creative gifts and toys since 1985. Our goal is to create products that amaze and captivate the imagination. Upon witnessing our LEVITRON® floating in space or our AstroBlaster^tm rebounding to five times the drop height or our Liquid Storms^tm's trapped swirling dynamic fluid, we hope people will react with amazement and ask "what makes that work?" or "how is that possible?". Our products are designed to illustrate just a few of the many magical aspects of our world."

But we didn't care, we're adventurous if nothing else, so we reached into our pockets to pull out our credit cards from our Duckskins Duct Tape Wallets, and ...

Nothing. Those bad boys at X-Stop took away our wallets too. So much for contributing to the economy.

Oh well, must be time to settle in to read one of our most favorite magazines, Redbook. There shouldn't be any problem with that, should there? Heck, we just saw it on the newsstand right next to People and TIME. Oops. Forgot to turn X-Stop off.

There are a couple of serious lessons to be learned here. First, the real reason why Fascinations and Godiva are blocked is because, long ago and far away, the domain names belonged to porn sites. Censorware vendors can add to their blacklists as much as they want, but unless they take the time and effort to periodically purge their lists (and few do), they risk blocking content which even they never intended to block. The content of a book, presumably, remains static for all time. The content of a website can be, and often is, changed dramatically on a regular basis.

Indeed, in January 1998, Michael Bradshaw, then the CEO of the company which makes X-Stop, proudly proclaimed that X-Stop had purged its blacklist of old, wrongful entries. Declan McCullagh apparently had gotten ahold of the X-Stop blacklist and had fed it into the Netly News Censorware Search Engine (which is sadly no longer in existence). In a note to McCullagh published on the fight-censorship mail list, Bradshaw wrote:

"Now that you show the world that we have a library of exact url's that are all porn, why not give credit where credit is due? All you are doing is helping the lame walk, by giving them our technology. Let them go earn these sights, not just copy them.

"Every search we go thru shows that we are from twice as good and complete to 10 times better....

"Stop sputtering about the single event of the few errors found one time! Why toss X-STOP, who hiccuped once, in the same ward with the terminally ill. Thanks Declan......Mike Bradshaw"

Now, some might say that being banned is just the price one pays for acquiring a "tainted" domain name, but such an argument would carry little weight in a court, particularly where, as here, the sites have been under their current ownership for quite some time indeed. Our readers know that censorware vendors blacklist domain names, but many perfectly intelligent site owners don't. We've had some correspondence with the Fascinations people about their ban, they knew about content blocking, they knew about NetNanny, but they had no clue about domain name blocking. And they're physicists.

The second lesson is drawn from the block on Redbook. Look at the keywords which Redbook uses to describe their site to search engines:

<META name="keywords" content="redbook magazine sex marriage husbands wives">

X-Stop honed in on "sex", with utter disregard for the context, and apparently without a clue as to just how acceptable Redbook magazine is. This, of course, is a classic illustration of the problem of relying on ‘bots to do your work for you. In the absence of a quantum leap in artificial intelligence, no machine can accurately gauge the context, instead of just the words. Whatever X-Stop or the other censorware vendors may tell you, this is the cold, hard truth.

People tend to think computers can do anything. They're great at performing boring, repetitious tasks such as multiplication or welding a certain seam on the frame of an automobile.

This is: a) The "Mud Room" at Log-On Data Corp. or b) An especially stupid-looking stock photo.

But they're terrible at making decisions like, "Is this offensive?". Computers can't look at pictures; the current state-of-the-art in analyzing pictures is almost sufficient to get a computer to distinguish between a picture of a person and picture of a field of daisies with moderate accuracy. It's no accident that the National Security Agency employs hundreds of highly-trained and highly-paid people to scrutinize satellite reconnaissance photos. Computers simply can't do it.

And when it comes to text, computers can't determine context, or interpret sentences. All they can do is look at certain words that may appear in the text of the webpage and see if those words are in their ban-list. This comparison might happen when the user is running the program (people call these keyword filters); or it might occur in advance, at the software company's headquarters (which is the way X-Stop works); but either way, the procedure is the same and the results are the same. You can get a little bit sophisticated - for example, you could say, "Ban all pages with the word 'breast' in them unless the phrase 'chicken breast' or the phrase 'breast cancer' is in the page." That would eliminate two sources of error in your classifications. But the web is vast; no matter how many exceptions you build into your rules, they won't be enough.

There could literally be two web pages with the exact same text with completely different pictures. These would be treated identically by any computer examiner, but perhaps one has disgusting photos and the other does not. The computer can't tell the difference. The computer can't tell the difference between Redbook pictures and Hustler pictures - and both webpages have the word "sex."

Log-On Data's Mudcrawler: The first true AI? We doubt it.

Artificial intelligence -- computers that read like humans do -- isn't coming any time soon. When it does, it will rock the world. AI is a development on the scale of the invention of the wheel, or the use of fire, or the harnessing of nuclear energy. When that day comes, all of this work will be moot. There truly will be an entity both fast enough and possessing the intelligence to evaluate the entire Galaxy Wide Web, and make decisions about offensiveness, hatefulness, obscenity, and the ineffable quality that distinguishes art from junk, good writing from bad, funny (Norm MacDonald on Saturday Night Live's Weekend Update) from unfunny (Colin Quinn). When that day comes, we will have to throw out all of our paradigms about what computers can and cannot do. When that day comes, when computers can replace not only a human's hands but also a human's brain, the world is in for an upheaval the likes of which has never been seen.

But it hasn't come. And no $24.95 software is going to even come close to what billion-dollar research projects have failed at for thirty years. The first AI won't be named "Mudcrawler".

Next > >