Turing Test transcripts: Is it bot or not?
From the earliest days of Eliza to the present, programmers have dreamed of creating a computer program that could pass the Turing Test -- one that could fool a user into believing that its chat was that of a human being.
Since 1990, a competition for the $100,000 Loebner Prize for Artificial Intelligence has conducted a sort of demolition derby for chatterbots trying to pass themselves off as people: Judges sit down at a set of terminals and chat, trying to guess whether the words appearing on screen have been generated by a person or a program. (In a 1997 article for Salon 21st, Tracy Quan wrote about her experiences as one of the competition's human ringers or "confederates.")
The transcripts of this year's competition, held last month at Flinders University in South Australia, make for amusing and often fascinating reading. Some of the ploys are obvious -- if a typo throws your conversational partner, you can bet it's a program, right?
But the more you read, the more difficult it is to put your finger on exactly what it is that makes conversation feel human. Was that non sequitur a philosophical leap -- or a line of dialogue pulled at random from a database? The most successful chatterbot in the competition, a program by Robby Garner, achieved only a 10 percent score on the Turing scale (50 percent would be indistinguishable from a human being). But some of the human "confederates" didn't score all that well in the final results tally, either.
-- Scott Rosenberg
SALON | Feb. 26, 1999
- - - - - - - - - -
Senate to nation: Get serious about Y2K
Don't retire those millennium-bug worries yet. While a good percentage of pundits are downplaying fears of a massive technology disaster when computer clocks hit midnight on Dec. 31, 1999, a new draft report from a special U.S. Senate committee studying the problem uses some dire language: "This problem will affect us all individually and collectively in very profound ways ... It will indeed impact individual businesses and the global economy. In some cases, lives could even be at stake.''
Those words come from an introductory letter to the committee's findings, which were obtained by the San Jose Mercury News. The letter, from committee Chairman
Robert Bennett, R-Utah, and Vice Chairman Christopher Dodd,
D-Conn., goes on to say: "The committee has no data to suggest that the United States will experience nationwide social or economic collapse, but we believe that disruptions will occur that in some cases will be significant. The international situation will be more disturbing. Those who suggest that it will be nothing more than a 'bump in the road' are simply misinformed.''
According to the Mercury News, the report counsels moderation, suggests stockpiling small amounts of extra food and water and urges consumers to "keep copies of financial statements." Hell, if the crisis gets bad enough, we can always burn our tax returns -- and Senate reports -- to stay warm.
-- Scott Rosenberg
SALON | Feb. 25, 1999
- - - - - - - - - -
Pentium III serial numbers hacked
On Monday, c't, a German technology magazine, revealed that it had found a way to read the serial number of Intel's new Pentium III chip without the owner's knowledge or consent.
Ever since privacy advocates raised an alarm about the new chips' serial numbers, which can be read by Web sites, Intel has assured the public that Pentium III owners would be able to use a software tool to turn the feature off and on and protect their privacy.
But c't's chip specialist, Andreas Stiller, found a way around Intel's safeguard. Stiller loaded an Active X "Trojan horse" (a disguised, malicious
security breaking program) onto a remote PC over the Internet. He then circumvented Intel's software tool by abusing a feature called Advanced Configuration and Power Interface (ACPI) -- a power-conservation standard created by Intel, Compaq and Microsoft.
"I switched the computer into 'Deep Sleep' mode, and rebooted the machine, then read the serial number before Intel's software tool was started," says Stiller.
The problem, it seems, is that the processor's serial number is in the "on" position by default; it's only Intel's software that blocks the number. Seth Walker, a spokesman for Intel, responds: "Don't think of it that way. The number is just there, it's not 'on.'"
In fairness to Intel, if someone manages to load a "Trojan horse" on your computer, then access to the chip's serial number is probably the least of your worries. Still, the report won't make Intel's job any easier as it tries to dispel fears and reassure PC users that their personal information is safe from prying eyes.
What can users do to protect their privacy? Intel is not just providing the software tool but also advising computer manufacturers to switch the serial number off in the BIOS (the first software instructions a computer loads when it boots up). The proud owners of new Pentium III PCs can then enable the serial number function using a custom piece of software from the manufacturer. But not all manufacturers will disable the serial number in BIOS, and once enabled it will be very difficult to turn off. Finally, Intel's Walker says, "We also advise users to choose carefully which Web sites they spend their time on." When it comes to privacy, it sounds like Intel's stance is "caveat surfer."
The company has vowed that it will not be keeping a database of the serial numbers -- although Intel vice president Mike Aymar admits that "we may be able to tell approximately when and therefore to whom the processor was sold."
So why did Intel introduce the serial number in the first place? To help corporations track and manage their PC inventory, and to provide another level of security for online banking and e-commerce applications. Banks will be able to use the serial number, together with user names and passwords, to verify an individual's identity.
Privacy groups such as the Electronic Privacy Information Center believe that the U.S. government had a hand in Intel's decision.
"We have repeatedly asked Intel if the NSA or the FBI requested them to include the serial number," says Dave Banisar, policy director for EPIC. "Their only response is that their largest customers have requested the serial number."
Of course, Banisar points out, the U.S. government is one of Intel's largest customers.
-- Niall McKay
SALON | Feb. 24, 1999
- - - - - - - - - -
Nuremberg Files anti-abortion site back online -- in Holland
It may be ironic that the Nuremberg Files, a Web site that packaged its anti-abortion militancy with banners of dripping blood, has been resurrected by "a liberal, smoking, cursing, bisexual, pro-abortion writer." But free speech issues, shrugs the writer in question, Karin Spaink, make for "strange bedfellows."
Earlier this month, the Nuremberg Files were taken down by the site's service provider, after the site was fined $100 million for promoting violence against abortion doctors by listing their personal information and "crossing out" doctors who had been murdered. Now Spaink,
a Dutch writer, has decided to put the site back online in her own Web space, hosted by the controversial Dutch service provider xs4all. The revived Nuremberg Files are now prefaced with a lengthy introduction by Spaink, explaining why she decided to make the material available again.
"While I strongly hold that every woman should have an abortion if she needs one," the site explains, "I do not think that other opinions about the subject should be outlawed or fined, no matter how harshly they are put." Adds Spaink, "I'd like to have a new debate about this issue; it's part and parcel of free speech. You have to realize that when you offer free speech, the stupid people and the nasty people and the bastards have the right to that speech too."
Free speech online is an issue near to Spaink's heart. Not only has she won a lawsuit again the Church of Scientology after they removed her personal Web site for "copyright violations," but she is the president of the group Contrast, which provides "online asylum" for banned political Web sites, such as the leftist German newspaper radikal.
Because the site has been up for less than 24 hours, Spaink says she hasn't received any reactions from the Net public at large, although she hopes that the context in which she has now placed the site will encourage useful debate. But lest anti-abortion activists try to use the new Nuremberg Files site to continue violence against the doctors listed therein, she warns, "You can never be sure that I haven't tampered with the names and addresses, and if you use the names for sick purposes, be advised that you might end up killing one of your affiliates."
-- Janelle Brown
SALON | Feb. 23, 1999
- - - - - - - - - -
Girl game pioneer Purple Moon shuts down
On the front door of Purple Moon's Web site last week, a cartoon girl named Miko exclaims, "Everyone has gone weird around here ... rumors flying, boys behaving totally strange, band members freaking out ... GET A GRIP, people!"
The screen was, perhaps, an unwitting commentary on the sudden demise of Purple Moon itself, the Paul Allen-backed company that proposed to lure girls to computers by offering them software with "a high cootie factor." Brandishing sheaves of research on girlish behavior by interface design pioneer Brenda Laurel, Purple Moon has produced a handful of games and an online community in the last two years -- all built around girls' relationships and friendships. As Laurel put it in an interview in the book "From Barbie to Mortal Kombat," the games focused on "emotional rehearsal for social navigation."
Purple Moon announced Friday that it was shutting its doors and firing all its employees, citing market competition. Purple Moon's signature character -- an anxious teen named Rockett -- couldn't, it seems, hold a candle to her buxom blond rival, Barbie. Despite the media buzz around Purple Moon, Mattel's popular line of "Barbie software" still firmly holds 63 percent of the "girl game" market, according to IDG.
It might be a case of good-for-you "girl-positive" earnestness falling flat on its face in front of frivolous, mainstream fun. It's also a big blow to the girl games pioneers -- idealistic independent entrepreneurs like Girl Games and Her Interactive, Purple Moon and Rhinestone Publishing -- who have hoped to weave pro-woman principles into entertainment software and bring girls up to speed on the technological revolution to boot.
"The whole girl games movement came from an unstable alliance between people out to make money, and therefore subject to market pressure, and people out to do good for girls and technology," says Justine Cassell, MIT Media Lab professor and editor of "From Barbie to Mortal Kombat." "Certainly Purple Moon were the most successful of the nonstandard paradigms, the nontraditional startup; they were really pushing a different kind of game ... But, who made the big inroads? It was Mattel, and they started out with 98 percent of the market."
Many of the independent girl game publishers have banded together in the past both to build the category and to create a united front in their competition against the Mattels of the world. Sadly, it seems, even independent companies with Paul Allen's cash behind them can't compete in an increasingly consolidated gaming industry that measures success by millions in sales. Even one of the biggest producers of girl-positive games -- the Learning Company -- was recently bought out by Mattel.
But as Cassell looks at it, even Purple Moon's demise isn't a total failure for do-good girl gaming: "The biggies aren't going to forget what Purple Moon learned, and I don't think we can go back at this point. Parents are so aware in 1999 of the fact that what used to be called the games market is actually the boys' games market. Neither girls nor their parents are going to put up with boys' games in pink boxes. And that's a success story."
Shares