Seven hot technologies
November 30, 2003 3:46 PM Subscribe
Seven hot technologies that we'll soon see on the market, according to MIT's Tech Review magazine. The spam blocker sounds like it might work. But the babelfish?
I'd like to know how the spam blocker prevents a spammer from getting around the math with some software tricks.
posted by punishinglemur at 4:01 PM on November 30, 2003
posted by punishinglemur at 4:01 PM on November 30, 2003
The change they are proposing is a change at the protocol level, so its much more of a challenge to implement, but it could probably be built ontop of current standards.
To get this to work, it seems straightforward. Say Bob sends Jane an email. Jane then send's bob's computers a small math problem to solve. Bob's computer sends Jane the answer. If the answer is right, Jane;s computer tells her she has a new Email rom Bob. This problem takes a computer like ~ 10 seconds to solve. So, a normal person could send hundreds of emails and not notice. However, a spammer who wants to send millions of emails would have a big problem because he would need ot buy a lot of computers to do that.
The trick is that each email generates a unique math problem that has to be solved. So, it costs the sender to send email a lot more than it does now.
posted by nads at 4:24 PM on November 30, 2003
To get this to work, it seems straightforward. Say Bob sends Jane an email. Jane then send's bob's computers a small math problem to solve. Bob's computer sends Jane the answer. If the answer is right, Jane;s computer tells her she has a new Email rom Bob. This problem takes a computer like ~ 10 seconds to solve. So, a normal person could send hundreds of emails and not notice. However, a spammer who wants to send millions of emails would have a big problem because he would need ot buy a lot of computers to do that.
The trick is that each email generates a unique math problem that has to be solved. So, it costs the sender to send email a lot more than it does now.
posted by nads at 4:24 PM on November 30, 2003
seems like it would actually have to be the mail server asking the math problem for it to work, but how would you stop the spammers from using a rogue mail server?
posted by rhyax at 4:30 PM on November 30, 2003
posted by rhyax at 4:30 PM on November 30, 2003
re: bablefish .. this would be such a revolutionary device, why don't we have it? It is like the flying car. The article says there are no plans to expand this device further. One would think there would be a government initiative to create one if nothing else to help gather intelligence in Iraq, I wonder if certain partys are afraid of such a device and work to keep it from happening. Or is it it really that difficult to create.
posted by stbalbach at 4:36 PM on November 30, 2003
posted by stbalbach at 4:36 PM on November 30, 2003
Say that solving the problem takes 10 seconds for an average computer; a single computer working round the clock could send out only 8,600 messages. To stay in business, spammers would have to invest heavily in new hardware.
What rhyax said. But also the email world is not cut so clearly between 1-1 messages and broad spam. What about the legitimate companies that publish on-line newsletters, or servers that host and distribute email discussion lists?
posted by vacapinta at 4:40 PM on November 30, 2003
What rhyax said. But also the email world is not cut so clearly between 1-1 messages and broad spam. What about the legitimate companies that publish on-line newsletters, or servers that host and distribute email discussion lists?
posted by vacapinta at 4:40 PM on November 30, 2003
re: the babelfish, my gf is a linguistics major and hopes to devote her career to just such challenges. Language is awesome and wondrous but there is a science to describing its workings. I have high expectations for the next 50 years in terms of machine translation.
posted by scarabic at 4:45 PM on November 30, 2003
posted by scarabic at 4:45 PM on November 30, 2003
stbalbach, there are three hard problems involved with babelfish.
(1) Speech-to-Text -- You have to first convert someone speech to text before you can translate it. Almost no one uses dictation software because it still isn't there yet. Although it definitely has gotten better since the software first started appearing in the mid 90's.
(2) Text-to-Text translation -- this is also tough and requires a lot of linguistic insights. IF you want to see how bad this is right now, use google translation service on a foreign web page.
(3) Text-to-Seech - This I think is probably the farthest along in terms of something acceptable to hear. I think AT&T has a sample program you can test over the web.
posted by nads at 4:56 PM on November 30, 2003
(1) Speech-to-Text -- You have to first convert someone speech to text before you can translate it. Almost no one uses dictation software because it still isn't there yet. Although it definitely has gotten better since the software first started appearing in the mid 90's.
(2) Text-to-Text translation -- this is also tough and requires a lot of linguistic insights. IF you want to see how bad this is right now, use google translation service on a foreign web page.
(3) Text-to-Seech - This I think is probably the farthest along in terms of something acceptable to hear. I think AT&T has a sample program you can test over the web.
posted by nads at 4:56 PM on November 30, 2003
A babelfish device is second on the list (after workable fusion power) on my list of tech things I really really really wish we had. IMHO, it's worth ten Land Warrior, Objective Force Soldier, MOAB, and Joint Strike Fighter programs. This is not just from a peacenik standpoint either -- good language translation is vital to intelligence-gathering, which in turn is vital to combating the security threats that the modern world faces.
posted by moonbiter at 5:22 PM on November 30, 2003
posted by moonbiter at 5:22 PM on November 30, 2003
The spam blocker is basically a twist on HashCash. And no, it wouldn't end spam in itself, it would have to be used as a part of an overall strategy. You assume that all e-mail is spam unless proven otherwise, and begin looking for evidence it was sent by a real human. If you find such evidence (e.g. the solution to the math problem) then you let the mail through. You have several methods of looking for evidence of non-spamminess and you apply these in some order related to effectiveness and/or transparency. For example:
0) If the sender's on your whitelist or in your address book, let it through. (You whitelist all the mailing lists you're on, that way the mailing list people don't have to do the math problem thing.) If not, continue to...
1) If it has the math problem thing or HashCash or the sender has posted a bond, or whatever other clever legitimacy proofs you decide to accept, it's not spam, let it through. Else continue on to...
2) Statistical filter; if it's clearly non-spammy, let it through. Else continue to...
3) Challenge-response system. These are annoying to legitimate senders, many of whom will refuse to reply to the challenge just on principle, so they should be used only as a last resort and only on messages that are fairly ambiguous to the statistical filter. Some good netizens may instead manually look through more suspect messages to avoid subjecting legitimate senders to this kind of abuse.
Anything that still seems to be spam can get tossed or placed in quarantine, depending on how spammy the statistical filter finds it and how confident the user is in its judgment.
posted by kindall at 8:59 PM on November 30, 2003
0) If the sender's on your whitelist or in your address book, let it through. (You whitelist all the mailing lists you're on, that way the mailing list people don't have to do the math problem thing.) If not, continue to...
1) If it has the math problem thing or HashCash or the sender has posted a bond, or whatever other clever legitimacy proofs you decide to accept, it's not spam, let it through. Else continue on to...
2) Statistical filter; if it's clearly non-spammy, let it through. Else continue to...
3) Challenge-response system. These are annoying to legitimate senders, many of whom will refuse to reply to the challenge just on principle, so they should be used only as a last resort and only on messages that are fairly ambiguous to the statistical filter. Some good netizens may instead manually look through more suspect messages to avoid subjecting legitimate senders to this kind of abuse.
Anything that still seems to be spam can get tossed or placed in quarantine, depending on how spammy the statistical filter finds it and how confident the user is in its judgment.
posted by kindall at 8:59 PM on November 30, 2003
Speaking as a professional translator, I'm not too worried about being displaced by Babelfish technology. Translation is, IMO, a computationally intractable problem. Not everyone speaks a textbook version of their language. Strike that--nobody does. Much of the meaning of our speech is contextually determined--by previous utterances or the situation at hand. Machine-translation researchers are aware of this--they've got a hierarchy of problems involved in translation, and last I checked in on it, they were only a couple rungs up.
Machine translation was thought to be a perfect problem for computers back in the 50s. A huge amount of work has gone into MT since then (notably, a big MT project in Japan during the 80s), and the results are, well, I guess they're interesting.
posted by adamrice at 6:49 AM on December 1, 2003
Machine translation was thought to be a perfect problem for computers back in the 50s. A huge amount of work has gone into MT since then (notably, a big MT project in Japan during the 80s), and the results are, well, I guess they're interesting.
posted by adamrice at 6:49 AM on December 1, 2003
how about instead of making the computer solve the math problem, we make the person sending the mail solve it? that way we take out spammers and people on AOL in one fell swoop.
posted by mcsweetie at 7:07 AM on December 1, 2003
posted by mcsweetie at 7:07 AM on December 1, 2003
adamrice:
Much of the meaning of our speech is contextually determined
article:
The system works with phrases likely to be used in specific situations, namely, travelers ordering in a restaurant, navigating a city, or seeking emergency medical care.
Bad journalism rears its ugly head again. This is no babelfish, no universal translator by far, just fairly standard context-specific translation technology.
posted by badstone at 9:49 AM on December 1, 2003
Much of the meaning of our speech is contextually determined
article:
The system works with phrases likely to be used in specific situations, namely, travelers ordering in a restaurant, navigating a city, or seeking emergency medical care.
Bad journalism rears its ugly head again. This is no babelfish, no universal translator by far, just fairly standard context-specific translation technology.
posted by badstone at 9:49 AM on December 1, 2003
« Older The Ovid Collection | The real Michael Jackson Newer »
This thread has been archived and is closed to new comments
posted by keswick at 4:01 PM on November 30, 2003