Slashdot Mirror
Artificial Intelligence At The COPA, COPA Commission
The Child Online Protection Act, passed late last year and then struck down early this year, is still under appeal. Colloquially it's known as "CDAII." Part of what the Act does is establish a Commission that meets every so often -- the Commission's website has details on its mandate and so on.
(Update, a few minutes later: make that "injunctified," or whatever one says for a law against which an injunction has been applied, instead of "struck down." Sorry; IANAL.)
Speaking at the Commission meeting yesterday and today were the CEOs of several major censorware companies. Among them was Michael Stephani, whose company Exotrope makes a product called BAIR.
BAIRBAIR checks images as they download onto your computer, and claims to be able to tell the difference between pornography and other types of images. The "AI" in its acronym stands for artificial intelligence, running on supercomputers.
When the Wired story on BAIR came out last month (a story "borrowed" from Peacefire -- I'm not going to get into it), Wired quoted the company as saying "they plan to fix the errors within the next month." What errors?
"BAIR incorrectly blocked photographs of Yellowstone, the Baltimore waterfront, Snoopy, boats, sunsets, dogs, vegetables and even a Wired News staff meeting.
"It rated as acceptable for minors -- even on the most restrictive setting -- explicit images of oral sex, anal sex, group sex, masturbation, and ejaculation."
That was one month ago. How's BAIR doing now?
Peacefire retested the same 50 pornographic images that they'd used last month (which presumably BAIR's programmers would have paid extra-special attention to). Their new report finds that, instead of zero, the number of blocked images is now: 34. I've got a great slogan for them: "now your children can only see 32% of the web's oral sex, anal sex, group sex, masturbation, and ejaculation."
One's respect for these programmers is dampened a little, though, because there's more to Peacefire's report. It seems, in a random sample of 50 photos of people's faces, BAIR blocked ... how many? ... 34.
Maybe that slogan should be: "now your children can only see 32% of the web," period.
It's wonderful to live in a world where artificial intelligence offers limitless possibilities. Its website suggests that "Because Artificial Intelligence can be taught to recognize a variety of patterns," -- oh, OK -- "our BAIR can be taught to evaluate other categories such as violence or illegal activities. The BAIR is currently undergoing training in these areas to provide additional filtration selections."
ClickSafeRichard Schwartz, CEO of ClickSafe, also spoke yesterday at the COPA Commission meeting. Just for kicks, Peacefire decided to try out their spiffy AI software too.
Insert marketblurb here: "...by combining cutting-edge graphic, word and phrase-recognition technology, ClickSafe has achieved accuracy rates of over 99% (according to recent sample tests). ClickSafe can precisely distinguish between appropriate and inappropriate sites (e.g. sites related to issues such as breast cancer will not be blocked)."
What Peacefire did was test this software against the website of the COPA Commission itself, and related sites such as those of speakers or Commission members. They found that blocked pages included:
- The Child Online Protection Act itself, in original and amended form;
- The COPA Commission FAQ;
- Biographies of Commission members Stephen Balkam and John Bastian;
- Bio of Commission member and famed anti-porn crusader Donna Rice Hughes, as well as AppendixA from her book Kids Online: Protecting Your Children in Cyberspace;
- A list of technologies the Commission examines;
- The scope of what the Commission is called upon to do;
- A service agreement from a little company called Network Solutions, whose rep chairs COPA's meetings;
- "About the ICRA" (the makers of RSACi, "a simple, yet effective rating system for web sites which both protected children and protected the rights of free speech");
- Bible study tools: "We hope these free resources foster a desire for Christians to learn more about the Bible, deepening their relationship with God" unless they're using censorware;
- The American Family Association (a conservative Christian group that is trying to force censorware into public libraries, including those surrounding the Slashdot Geek Compound);
- The ACLU, the EFF, and the Center for Democracy and Technology;
and so on.
When I spoke with Bennett about this, he commented that the strange thing was that these flaws are so easy to find; you'd think someone would have run these simple tests already. If anyone reading wants to get their name in Slashdot (and other news media too), censorware is a gold mine of untested misinformation. Buy a product, design a solid unbiased test for it, run the test, and send us what you find. Repeat until the whole world has a clue.
The COPA Commission MeetingThe following is an account of yesterday's COPA Commission meeting, by Waldo Jaquith. Keep in mind that this meeting's purpose, according to the Scope & Timeline Proposal which is blocked by ClickSafe, is to study filtering and blocking software to learn what to recommend in its report to Congress late this year.
Folks,
For more information on the COPA Commission, see http://www.copacommission.org/. (Unless your network has ClickSafe installed, in which case you shouldn't bother.) There is an agenda for this meeting, and there are bios for most people, as well as the prepared speeches for many of the below folks. I've tried to be objective.
Oh, screw that. There's nothing objective about it. But I've tried to give useful facts, quote accurately, etc.
The whole affair, which was scheduled to start at 9:30am, didn't actually start until 10:15am. Which was good, because I didn't get there until 9:45. Although the event was being held at the University of Richmond's Jepson Alumni Center, the room felt like your basic hotel meeting room. Bad carpet, ugly chairs, poor lighting. There were enough chairs to seat about 100 people, but only 35 people were in attendance. Directly in front of the two columns of chairs was a table with chairs, facing away from the audience. This table was for people asked to testify before the COPA Commission. On the other side of that table was a long table, at which was seated the commission, all sixteen members. The result was that the people testifying, who did most of the talking, could only be recognized by the backs of their heads by the audience.
Chairman Donald Telage called the meeting to order and introduced the first panel, who was to speak for approximately 45 minutes on the topic of client-side filters. This panel included Gordon Ross, the President and CEO of Net Nanny, Mark Smith, the President of BrowseSafe, Susan Getgood, the VP and General Manager of Cyber Patrol, and Richard Schwartz, the CEO of Opportunity-America (ClickSafe.com).
Gordon Ross kicked things off with a tremendously boring ten minute speech about how client-side filters work. The only interesting comment that he made was his belief that "consumers should have the ability to analyze each and every site in the database..." [...because his product Net Nanny is the only one of the 150 censorware packages on the market that allows oversight of its blacklist. -ed] He also kicked off the First Amendment references, which nearly every speaker throughout the day would spend some time talking about, but not really saying very much.
Mark Smith from BrowseSafe occupied the next few minutes, giving a rambling speech in which he discussed censorware as if it were some far-off and idyllic concept.
"Most products focus on either client-side- or server-side-based technology. What would happen if the benefits of each could be brought together to provide the user with a new, more flexible and powerful way of surfing the web? What if every sub domain of every site had been categorized and classified by its content? Wouldn't you agree that everyone could benefit from that combination of technology? Of course you would? Now let's walk across the street to the front porch of the family of the home and try to view it from the parent's perspective. What if parents were able to determine what the child sees? What would it be like if e-mail, instant messaging, chat and other computer tools could be also controlled?"
Then, although the topic was client-side filters, he rambled on for several minutes about PlanetGood, a website that was probably unfamiliar to many in the room. He used the site's name in every single sentence for several minutes. And, naturally, he closed talking about "our forefathers" and "these inalienable rights that our forefathers entrusted to us and many of them died for."
Susan Getgood from Cyber Patrol kept things short and sweet, and took the "I'm a new mother and want to protect my children" approach. She muddled the definition of censorship somewhat, saying that "[s]ome critics confuse censorship, which is imposed by the government, with technology that a family or school can choose to use and then set to implement an individual policy." Our school system isn't a part of the government?
Richard Schwartz of ClickSafe.com touted his product nearly as much as Mark Smith promoted the mysterious "PlanetGood." He also described a system that his company has developed that sounds very much like Exotrope's BAIR. "Fleshtone has a very unique set of features [...] Through a combination?of a set of sophisticated algorithms it can establish if something is pornographic. [...] Justice Potter Stewart lives within our system, because he knows it when he sees it. It works, it's been tested out, it's over 99% effective." "We can distinguish between chicken breast and sexy breast." "A consortium of Portuguese and Australian pornographers had been hijacking people off of different sites, including the Harvard Law Review site into their pornographic sites. And then you have to reboot your computer in order to get out."
After the four had testified, we moved into the commission Q&A session. (No questions would be allowed from the audience.) A few interesting questions, answers, and comments cropped up during this portion.
Richard Schwartz, only half kidding, proposed a tax on Internet pornography.
Commissioner Gregory L. Rohde asked Richard Schwartz if his image filter could tell the difference between art and pornography. Astoundingly, Schwartz replied that it could.
Commissioner Jerry Berman asked if there were any plans to create an organization that could provide objective reviews of censorware products to help parents decide what to buy. Gordon Ross said that this had been tried a few years back with SIFT (?), and that it didn't work out.
After a short break, we began the second panel, which addressed server side filtering. Testifying was Kevin Fink, N2H2's CTO; Sunil Paul, Chairman of Brightmail; Stephen Boyles of Library Guardian (Swifteye); Michael Stephani, President and CEO of Exotrope; Ginny Wydler, Director of Standards and Policy at AOL; and Tim Robertson, CEO of FamilyClick.
The first person to say anything interesting was Michael Stephani, who made some fairly interesting claims. He said that their blacklist of sites included four million sites, and that their image-recognition software, BAIR, is 99.8% percent effective. Stephani bragged that it blocked 1 out of 6 general images and 96 out of 100 pornographic images. He pointed out (perhaps rightly) that image filtering is the only real way to filter out pornography, and also that client-side filtering would so go the way of the dodo, given the proliferation of Internet appliances. It wasn't long before he got all 'God bless America' and 'think of the children,' and eyeballs could be heard rolling throughout the room.
As Commissioners asked questions of the panel, Chairman Donald Telage admitted that he wasn't aware that client-side filters were able to use a blacklist. He was under the impression that they could only filter. I had flashbacks from the Napster hearings last week ("Can't you track their intellectual property address?")
Out of the blue, Karen Talbert asked the panel for a show of hands regarding their respective products' ability to work with high-speed connections. Obviously, everybody's hands went up.
How do these people get on the commission?
When given half a chance, Stephani got all "think of the children, my god, won't somebody think of the children?" again. He also bragged that Exotrope has a new, not-yet-released product that filters IM [AOL Instant Messaging -ed.] and even detects innuendo. Stephani said that they just got a contract to install this program on 30,000 school servers. Continuing his spectacular Old Faithful of shit, he cheerfully envisioned a time in the future when there would be "photonic switches" that would maintain a complete blueprint of everything that every user had ever done on-line. Christ, that's frightening. Stephani said that they'd spent $6.5MUS developing BAIR, and went on to point out the coincidence that Peacefire released the report showing that BAIR was 0% effective on the same day that their servers went down. Perhaps he was implying that Peacefire members hacked the server, perhaps that we were taking advantage of them, or perhaps he was just laughing at the circumstances.
There was no promised audience Q&A. That's probably because the whole event ran well over when it was supposed to end. Lacking a better approach, I rushed up to the ebullient Stephani with a copy of the newest BAIR report in hand. Although he was already talking to a reporter, he stopped when he saw my nametag ("Waldo L. Jaquith, Peacefire") and looked a little surprised. He, as well as his sidekick PR guy, enthusiastically introduced themselves. We talked for a few minutes, during which time I said that BAIR appears to suck less than many other censorware programs. But I was still fundamentally opposed to all of them. Between this and the revised report, Stephani was my new best friend. Several other people came forward to read nametags and shake hands, but I continued to talk to Stephani and the reporter, Drew Clark from Technology Daily.
Ten minutes later, when I walked out, I felt a little baffled. Stephani behaved towards me as if Peacefire had just given him the most glowing review that BAIR had ever gotten. This, despite my repeatedly pointing out that Peacefire is fundamentally opposed to filters, always will be, and BAIR is simply rather effective at performing the task that we hate.
I was disappointed that a few major points were never brought up during the discussions:
- Server-side censorware (especially that which is housed with each website) will always be a severe privacy violation, because it needs data on the user in order to establish what information to provide.
- Client-side censorware is doomed to fail because children know more about computers than their parents. The parent has to trust that little Suzy won't uninstall Cyber Patrol. But if Suzy can be trusted, why bother with Cyber Patrol?
- Internet censorship is impossible. The Internet is so large that it's a waste of time, so let's all stop. Gated community models, like AOL, Compuserve and such, are a far better way to provide a "safe" experience for kids.
- The concerns about children's wellbeing presented during the meeting mirror those that parents, since the beginning of time, have always had for their children. How can I keep my child safe when I'm not watching him? How do I know what my child is doing if I'm not around? How do I keep my children from hearing / seeing / saying bad things? Censorware makes no more sense than installing a v-chip in little Suzy's head. Get over it.
In a nutshell, I'm not sure what, if anything, was established at this meeting. It's clear that most of the Commissioners knew every little to start off with, and their opinions are being formed on what amounts to a series of sales pitch sprinkled with god-and-country references, a la mega blowout carpet sales around Independence Day. I'm glad COPA was struck down. Let's get on with our lives.
Best,
Waldo
AI Bots that can identify and snarf porn for us!
:)
What will those great Censorship people think of next?
---
pb Reply or e-mail; don't vaguely moderate.
pb Reply or e-mail; don't vaguely moderate.
We have quite a few stories at LISNews.com on Censorship and Filtering that can give you some idea on what's going on around the country in these areas as well. Mostly from the popular press, shows who is for and against and where this is being used and fught against.
Think about it.
First, we have the idiot box (I'll bet "boob tube" is probably filtered out by default). Put your child in front of it, point their heads at the shiny part, and walk off. Originally, this worked out pretty well, then TV started getting sexy. That just had to go.
Nevermind watching TV with your kids, probably the most minimal form of parenting possible. Nevermind that you don't want your kids to see sex (hopefully, they will grow up and have sex), but it's okay to watch gods know how many murders per day (hopefully, they will grow up and not murder people).
No, for the 15% of the children who were just a little too active and intelligent to just sit in front of the television, let's give them something interactive. Here comes the Web. Same principle applies here. Put the kid in front of a computer, let little Timmy click away, and, again, stop interacting with the child.
Minimalist parenting arises from a mostly Republican morality and a Democratic sense of "we know what's best." The worst of both parties has collided to create parents who would like to put a childproof cap on the world, kid-safe, mother-approved, no small parts to swallow, no sharp edges. Just have them, take the baby pictures, throw some clothes on them, and then let them wander about the big, wild world while you go off and have your lattes and shop frantically in your SUVs. Your children will be protected automagically, just as easy as procreation itself.
Parents have pretty much abdicated all interaction with their kids, and tools like this help it happen.
You seem to have this impression that the United States is filled up primarily with prudes, moralists, and thought police that are hell-bent on making sure that people aren't exposed to any material that might make their penis hard or make them question their personal relationship with Jesus. In reality, this is not the case. The advocates of censorware and book-burning and sodomy laws and things like that are in a small (but very vocal) minority. They have taken it upon themselves to protect "the majority" from all these things that they find evil, immoral, or otherwise unwholesome, but the fact of the matter is that most of "the majority" could care less.
.. it provides talk show hosts with a never-ending supply of joke fodder. Most folks I know pity the "moral right" and the sad, angry little world in which they live. Don't try to paint all Americans with the same brush.
In short, you're trying to pigeonhole Americans with this carefully-cultivated mental image that is probably the result of reading too many religious trolls on Slashdot. Very few Americans are "bible thumpers." Very few Americans are content to let the "moral right" tell them how to live their lives and behave in their bedrooms. Very few Americans have "hangups about sex." As a matter of fact, the "moral right" in America serves a very useful purpose
We're going down, in a spiral to the ground
These moralistic bastards honestly believe that they can do a better job of parenting than I can. I beg to disagree.
Will their software gradually "ease the reins" while little Johnny grows older and starts to explore more complicated and adult interpersonal relationships?
Will their software help little Suzie to understand why some adults choose to look at pr0n, or merely block it?
No thanks, guys. I prefer to parent the old-fashioned way. I take responsibility for controlling what my children can browse, and I am the one who will supervise them, answer their questions, and help them to grow up.
In the mean time, I'm running Squid to proxy and log all HTTP requests. Does this mean that I'm spying on my children? Not really; those are the well-publicized house rules. My children will learn what their mother and I believe to be acceptable content, in the same way that they'll know what comic books and novellas we approve of.
Dewey, what part of this looks like authorities should be involved?
My own tests here on 100 clips of live goat porn showed that the AI blocked 34 of them. I've yet to test it on pictures of naked black women, body painted women, Asian women, body painted Asian women, or any of the above posed with goats. I'll try to get to that later today. I may have to apply to the assorted censorware manufacturers for a grant for a larger hard drive so that I may continue this IMPORTANT RESEARCH, as all images must be kept for archival purposes.
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?